Decision coaching for people making healthcare decisions

Abstract

Background

Decision coaching is non‐directive support delivered by a healthcare provider to help patients prepare to actively participate in making a health decision. ‘Healthcare providers’ are considered to be all people who are engaged in actions whose primary intent is to protect and improve health (e.g. nurses, doctors, pharmacists, social workers, health support workers such as peer health workers).

Little is known about the effectiveness of decision coaching.

Objectives

To determine the effects of decision coaching (I) for people facing healthcare decisions for themselves or a family member (P) compared to (C) usual care or evidence‐based intervention only, on outcomes (O) related to preparation for decision making, decisional needs and potential adverse effects.

Search methods

We searched the Cochrane Library (Wiley), Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE (Ovid), Embase (Ovid), PsycINFO (Ovid), CINAHL (Ebsco), Nursing and Allied Health Source (ProQuest), and Web of Science from database inception to June 2021.

Selection criteria

We included randomised controlled trials (RCTs) where the intervention was provided to adults or children preparing to make a treatment or screening healthcare decision for themselves or a family member. Decision coaching was defined as: a) delivered individually by a healthcare provider who is trained or using a protocol; and b) providing non‐directive support and preparing an adult or child to participate in a healthcare decision. Comparisons included usual care or an alternate intervention. There were no language restrictions.

Data collection and analysis

Two authors independently screened citations, assessed risk of bias, and extracted data on characteristics of the intervention(s) and outcomes. Any disagreements were resolved by discussion to reach consensus. We used the standardised mean difference (SMD) with 95% confidence intervals (CI) as the measures of treatment effect and, where possible, synthesised results using a random‐effects model. If more than one study measured the same outcome using different tools, we used a random‐effects model to calculate the standardised mean difference (SMD) and 95% CI. We presented outcomes in summary of findings tables and applied GRADE methods to rate the certainty of the evidence.

Main results

Out of 12,984 citations screened, we included 28 studies of decision coaching interventions alone or in combination with evidence‐based information, involving 5509 adult participants (aged 18 to 85 years; 64% female, 52% white, 33% African‐American/Black; 68% post‐secondary education). The studies evaluated decision coaching used for a range of healthcare decisions (e.g. treatment decisions for cancer, menopause, mental illness, advancing kidney disease; screening decisions for cancer, genetic testing). Four of the 28 studies included three comparator arms. 

For decision coaching compared with usual care (n = 4 studies), we are uncertain if decision coaching compared with usual care improves any outcomes (i.e. preparation for decision making, decision self‐confidence, knowledge, decision regret, anxiety) as the certainty of the evidence was very low. 

For decision coaching compared with evidence‐based information only (n = 4 studies), there is low certainty‐evidence that participants exposed to decision coaching may have little or no change in knowledge (SMD ‐0.23, 95% CI: ‐0.50 to 0.04; 3 studies, 406 participants). There is low certainty‐evidence that participants exposed to decision coaching may have little or no change in anxiety, compared with evidence‐based information. We are uncertain if decision coaching compared with evidence‐based information improves other outcomes (i.e. decision self‐confidence, feeling uninformed) as the certainty of the evidence was very low.

For decision coaching plus evidence‐based information compared with usual care (n = 17 studies), there is low certainty‐evidence that participants may have improved knowledge (SMD 9.3, 95% CI: 6.6 to 12.1; 5 studies, 1073 participants). We are uncertain if decision coaching plus evidence‐based information compared with usual care improves other outcomes (i.e. preparation for decision making, decision self‐confidence, feeling uninformed, unclear values, feeling unsupported, decision regret, anxiety) as the certainty of the evidence was very low.

For decision coaching plus evidence‐based information compared with evidence‐based information only (n = 7 studies), we are uncertain if decision coaching plus evidence‐based information compared with evidence‐based information only improves any outcomes (i.e. feeling uninformed, unclear values, feeling unsupported, knowledge, anxiety) as the certainty of the evidence was very low.

Authors' conclusions

Decision coaching may improve participants’ knowledge when used with evidence‐based information. Our findings do not indicate any significant adverse effects (e.g. decision regret, anxiety) with the use of decision coaching. It is not possible to establish strong conclusions for other outcomes. It is unclear if decision coaching always needs to be paired with evidence‐informed information. Further research is needed to establish the effectiveness of decision coaching for a broader range of outcomes.

Plain language summary

Decision coaching for people making healthcare decisions

Background

There is a need to better involve people who are making healthcare decisions about treatments (e.g. surgery) or screening (e.g. tests to tell if there is a health problem). Quality decisions are made when people know the best available evidence on options and can share what matters with their healthcare provider(s).

Decision coaching supports people to prepare for making a health decision. It is provided by healthcare providers who are trained or use a protocol for decision coaching (e.g. nurses, doctors, pharmacists, social workers, health support workers such as peer health workers). 

We wanted to find out if decision coaching helps people to prepare for making healthcare decisions.

 

What did we do?

We are a team of patient partners, healthcare providers, teachers and researchers from seven countries.

We looked for studies that tested decision coaching with people (adults and children) to prepare them for making a healthcare decision about treatment or screening for themselves or a family member.

We included studies if people who received decision coaching were randomly put into study groups (e.g. using a computer to decide who goes in which group). Studies where people are randomly put into groups are the best way to compare findings between groups and give results we can rely on when we look at the effects of an intervention like decision coaching. 

 

Search strategy 

To find studies, we searched eight online data banks. We asked experts and authors of studies about decision coaching. We included studies published up to June 2021. 

 

What we found 

We found 28 studies about decision coaching used alone or with high‐quality patient information based on research (called ‘evidence‐based information’). There was a total of 5509 adults in the 28 studies. None of the studies included children, and only one study included people (parents) making decisions for someone else. The studies tested decision coaching with a range of healthcare decisions such as treatment decisions related to cancer, menopause, or mental illness; cancer screening decisions, and genetic testing. 

Some of the studies looked at decision coaching with, or compared to, disease‐specific information or evidence‐based information such as patient decision aids (booklets, videos, online tools that: make the decision clear, provide options and the pros and cons, and help people be clear on what matters to them).

 

What does the evidence show?

People who received decision coaching compared with evidence‐based information only: 

‐ may have little or no change in knowledge (406 patients, 3 studies);

‐ may have little of no change in anxiety (242 patients, 1 study).

 

People who received decision coaching plus evidence‐based information compared with usual care: 

‐may have improved knowledge (1073 patients, 5 studies)

 

Our confidence in the results

We have low confidence that decision coaching plus evidence‐based information improves people’s knowledge compared to usual care. We have low confidence that decision coaching may have little or no effect on knowledge and anxiety compared to evidence‐based information. We are less confident about our other findings, as the certainty of the evidence is very low and there were important outcomes that were not reported by the included studies. Many studies had a small number of people taking part and this means that the results of this review might change with more studies.  

 

What this means

Decision coaching may improve peoples’ knowledge to help them prepare to make healthcare decisions when used with evidence‐based information. The review did not detect any adverse effects with the use of decision coaching. 

Background

Description of the condition

Healthcare systems around the world call for improved patient engagement and increased efforts to support those making healthcare decisions (see ACSQHC 2021; Chow 2009; Frosch 2011; Härter 2011; NICE 2021; Oslo University Hospital 2021). Shared decision making (SDM) is a process by which people who are personally experiencing a health issue ('patients') (SCPOR 2019), and their clinician(s) work together to make decisions about screening, treatments, or management of chronic conditions. Family members or significant others, or both, may also participate in SDM. SDM is based on the best available evidence on healthcare options and the patients' informed preferences (Coulter 2011; Makoul 2006). SDM has been shown to improve patient outcomes and experience, improve the experiences and effectiveness of health professionals in their communication with patients, and may optimise costs in healthcare (Légaré 2012; Légaré 2014; Légaré 2018). Despite the benefits and effectiveness of SDM, there have been low levels of uptake in clinical practice (Couet 2013; Coulter 2018). 

To facilitate patient involvement in decision making, healthcare providers trained or using a protocol in decision coaching may use different interventions with patients including decision coaching, patient decision aids, and question prompts (Stacey 2017a). 'Healthcare providers' refers to both healthcare professionals (e.g. audiology, speech language pathology, dentistry, medicine, nursing, midwifery, occupational, physiotherapy) and health support workers (e.g. peer support worker, lay health worker) (Jull 2019; WHO 2010). We consider healthcare providers to be all people who are engaged in actions whose primary intent is to protect and improve health (WHO 2007). Interventions for patients only have the potential to be effective at facilitating patient involvement in decision making (e.g. patient decision aids); however, interventions targeting patients and healthcare providers (e.g. training) have the greatest potential for facilitating SDM (Légaré 2018). 

Description of the intervention

Decision coaching is non‐directive support delivered by a healthcare provider to help patients prepare to actively participate in making a health decision (Jull 2019; Rahn 2021; Stacey 2012). The goal is to improve the decision‐making process and to ultimately achieve informed, values‐congruent decisions. Decision coaches support patients to: a) identify their decision‐making needs; b) help them understand information that is evidence‐based on options (including watchful waiting), benefits and harms; c) help them clarify their values for outcomes of options (what matters to them); and d) encourage them to communicate their preferences to others (e.g. family, clinicians). Decision coaches may also provide patients with other evidence‐based information such as patient decision aids that are paper‐based or online tools that may be used alone or in a consultation with a healthcare provider (Jull 2019; Rahn 2021; Stacey 2017b). 

 

Decision coaching is provided by interacting with patients using face‐to‐face, telephone or other communication media (i.e. not automated electronically) (Jull 2019; Stacey 2012). Decision coaching aligns well with the competencies of most healthcare providers who are expected to provide information that is evidence‐based and to support patient’s chosen level of participation (Joseph‐Williams 2014; Politi 2013). Decision coaching is not routinely taught in training programmes preparing healthcare professionals for licensure; healthcare providers require additional specific training focussed on the knowledge and skills for providing decision coaching or a specific protocol (Stacey 2006; Stacey 2009). 

How the intervention might work

Decision coaching aims to support adults and children to participate in the process of decision making to improve the quality of the decision‐making process and, consequently, the quality of the decision. This is consistent with decision coaching and other interventions described by the Ottawa Decision Support Framework (O'Connor 1998; Stacey 2020). According to this Framework, decision making can be adversely affected by unmet decisional needs: decisional conflict (e.g. uncertainty about the best course of action); inadequate knowledge; unrealistic expectations; unclear values for outcomes of options; inadequate support or resources; complex decision type; urgent timing; unreceptive stage of decision making; polarised leaning towards an option; and participants' characteristics (Stacey 2020). The main assumption of the Framework is decision‐support interventions (e.g. counselling, patient decision aids, decision coaching) that address patients’ decisional needs to improve the quality of the decision and the decision‐making process, which may favourably affect implementation of the chosen option and appropriate use of health services. However, decision uncertainty is typically addressed after consultation with the clinician. 

Why it is important to do this review

The Cochrane Review of interventions to improve the use of SDM reported low certainty of evidence (Légaré 2018). When interventions to improve the use of SDM included decision coaching, it was delivered as part of a multi‐faceted intervention which makes it difficult to determine the effect of decision coaching and the effect on outcomes beyond SDM is unclear. Another review, focussed on the broader concept of coaching, showed that 17 studies of various designs indicated limited evidence suggesting improved patient‐physician communication in specialists’ consultations (Alders 2017). Other reviews have captured aspects of decision coaching when conducted within randomised controlled trials evaluating patient decision aids, yet none have specifically focussed on decision coaching as an intervention (Rahn 2021; Stacey 2012; Stacey 2017b). For example, the Cochrane Review of patient decision aids reported moderate certainty of evidence for the effects of patient decision aids on patient participation in decision making and multiple other outcomes (e.g. decision quality, decision‐making process) (Stacey 2017b). Some patient decision‐aid studies included decision coaching. A recent scoping review of decision coaching provided alongside a patient decision aid compared to usual care indicated that decision coaching with a patient decision aid improves knowledge and there were mixed results for other outcomes with no harms reported (Rahn 2021). A recent study about implementing SDM in clinical practice by healthcare providers indicated the need to focus on the use of behaviour change approaches with healthcare providers and with patients, for more effective SDM implementation (Agbadjé 2020). 

In summary, we convened an interdisciplinary, international team who are active in decision coaching. There have been no previous reviews conducted to explicitly measure the effectiveness of decision coaching as an intervention to facilitate patient participation in decision making. Given the human resource requirements to provide decision coaching, our team viewed it as important to determine the unique contribution of decision coaching as a healthcare intervention.

Objectives

To determine the effects of decision coaching (I) for people facing healthcare decisions for themselves or a family member (P) compared to (C) usual care or evidence‐based intervention only, on outcomes (O) related to preparation for decision making, decisional needs and potential adverse effects.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs) with any kind of allocation method, e.g. individual, cluster, or stepped‐wedge. We excluded all other study designs and commentaries. There were no language restrictions.

Types of participants

We included studies conducted with adults and children preparing to make a healthcare decision regarding treatment or screening for themselves or a family member (substitute decision maker). Adults were defined as 18 years of age and older. Children were defined as under 18 years of age. We excluded any studies with someone who was making a lifestyle decision only, a decision about participation in research, or advanced care planning. We excluded studies with simulated patients and/or with patients making hypothetical decisions.

Types of interventions

We considered studies evaluating the decision‐coaching intervention that met the following minimal criteria:

• delivered by a healthcare provider trained or using a protocol in decision coaching;

• aimed to prepare the people to participate in making a healthcare decision;

• delivered non‐directive support to people in the process of decision making; and

• delivered individually, person‐to‐person, by telephone, or via the internet.

Different terms may be used in place of 'decision coaching', such as 'health counselling', 'support', and 'navigation' (Stacey 2013). We also included decision‐coaching interventions paired with evidence‐based information (e.g. patient decision aid) (Stacey 2012).

We excluded studies that described healthcare providers who were: a) making the decision with, or on behalf of, the person; b) not prepared for decision coaching with training or a protocol; c) recommending or being directive about a specific option; d) not described as having direct interests in providing decision coaching; e) providing decision coaching to groups; or f) automated decision coaching.

We included any comparison (e.g. usual care or other) except those where decision coaching was provided to both groups in the study:

• decision coaching compared with usual care;

• decision coaching compared with evidence‐based information only;*

• decision coaching plus evidence‐based information* compared with usual care; and,

• decision coaching plus evidence‐based information* compared with evidence‐based information only*.

*evidence‐based information includes patient decision aids.

Types of outcome measures

We took the following steps before including outcomes in the Results and Summary of findings tables.

If the outcomes of interest were reported in the article with only one measure (e.g. Decisional Conflict subscales), then no further action was needed. If reported with more than one measure (e.g. knowledge), we had planned that the outcome measures for each article would be listed and two review authors would independently make decisions about what is most relevant to patients; any differences were resolved by the involvement of a third author. For outcomes with more than one measure (e.g. knowledge) that could not be included in meta‐analysis, we reported results descriptively.

Outcomes of interest to the knowledge users and researchers on the team included active preparation for health decision making, decision‐making process outcomes and decision quality (e.g. informed, values‐congruent chosen option). As the main aim of decision coaching is to prepare people to participate in decision making, we identified primary outcomes that include preparation for active participation in making a health decision, resolution of modifiable decisional needs of relevance to decision coaching and adverse effects. 

We included data for all outcomes reported in each study that mapped to the primary and secondary outcomes of this review.

Primary outcomes

• Preparation for active participation in making a health decision (patient‐reported). Outcomes were reported as per the Preparation for Health Decision Making scale (Bennett 2010). Preparation for active decision making includes: recognise that a decision needs to be made, feels prepared to make a decision, is helped to consider advantages and disadvantages of the decision, considers which advantages and disadvantages matter the most to the patient, organise thoughts about the decision, think about how involved they want to be in the decision, prepare to talk with a healthcare provider about the decision, and identify questions to ask. Other indicators of preparation for active participation in making a health decision include decisional self‐confidence and patient involvement in decision making (patient, observer‐reported) including perceived involvement, preferred level of involvement, participated at preferred level of involvement.

• Resolution of modifiable decisional needs relevant to decision coaching (patient‐reported). Example outcomes, according to the Ottawa Decision Support Framework (Hoefel 2020; Stacey 2020), include: uninformed, unrealistic expectations; unclear values; and inadequate support.

• Any reports of adverse effects on patient or decision coach (patient, observer‐reported), that includes worsening effects such as increased decision regret and anxiety (meaning lower remorse or distress over a decision (Brehaut 2003)).

Secondary outcomes

• Satisfaction with decision coaching (patient‐reported)

• Decision quality (e.g. informed choice, values‐congruent chosen option)

• Health systems resources (e.g. length of time, cost)

• Quality of life (patient‐reported)

• Quality of decision coaching (measured by an observer)

Search methods for identification of studies

Electronic searches

We developed and executed a comprehensive search strategy with an information specialist (AP), that was peer reviewed by a second information specialist (TR), according to the Peer Reviewed Electronic Search Strategy (PRESS) guidelines (McGowan 2016). After screening for included studies, we identified one additional study that should have been included. We worked with the information specialist (AP) to update our search strategy and conducted a new search in May 2020 with the additional key words (knowledg* or power* or empower*).

We searched from database inception to June 2021 in the following databases: the Cochrane Library (Wiley), Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE (Ovid), Embase (Ovid), PsycINFO (Ovid), CINAHL (Ebsco), Nursing and Allied Health Source (ProQuest), and Web of Science. We conducted the original search strategy for MEDLINE (Ovid) and tailored it to other databases (Appendix 1; Appendix 2; Appendix 3; Appendix 4; Appendix 5; Appendix 6; Appendix 7). The search strategy had the Cochrane RCT classifier added to limit to RCTs only. Conference abstracts and theses were excluded. We did not restrict the search by language. 

Searching other resources

We searched the following grey literature sources:

• Forward and backwards citation searching of included studies were used. Google Scholar was used to identify articles that have cited each of the included studies (forwards searching) and reference lists of included studies were screened for other relevant articles (backwards searching).

• Google Scholar using advanced search (with key words combined scanning the most recent 50 citations, as per guidance from the information specialist (AP))

• Contacting experts on shared decision making and decision coaching through the International Shared Decision Making Facebook group and authors of included studies to identify other relevant studies.

Data collection and analysis

According to the Canadian Institutes of Health Research (CIHR), integrated knowledge translation (KT) is a research approach that involves knowledge users as research team members who are active throughout the entire research process, from defining the research question to applying the findings (CIHR 2015). Our team consisted of a study executive including a patient partner (JJ, DS, SKo, MS) and international collaborative research group, that has been engaged in the entire systematic review process so that the outputs of the research are more likely to be applied in practice and policy (Bowen 2013; Van de Ven 2006). For each step of the systematic review methods, we invited team members to actively participate. We also used formal and informal group consensus methods to make decisions about the conduct of the review (Pollock 2019).

Selection of studies

We applied the Cochrane RCT Classifier to the database search results. The Classifier assigned a probability (from 0 to 100) of being a randomised trial to each citation retrieved. Those citations with Classifier scores of nine or less were excluded from dual reviewer screening but were screened by a single reviewer (titles and abstracts) to ensure that no studies were misclassified and wrongly excluded from the search outputs.

For other studies assigned a probability of > 9 of being an RCT, two members of the author team (DS, JF, KL, LB, AD, SKi, SKo, JJ, AR, JK) randomly allocated and independently screened titles and abstracts against inclusion and exclusion criteria using the online systematic review software Covidence 2021. After resolving discrepancies, we retrieved the full text of any studies identified as potentially relevant. Two members of the review author team (DS, JF, JJ, AR, SKi, KL, LB, AD, SKo, JK) were randomly allocated to screen full‐text studies. Discrepancies were resolved by discussion with a third review author who did not participate in the screening to reach consensus. Finally, two reviewers confirmed the eligibility of the decision‐coaching intervention in all included studies and a third reviewer helped reach consensus on discrepancies (JJ, DS, JF).  Findings were reported in a PRISMA diagram (Liberati 2009).

All excluded full‐text studies were listed with reason(s) (see Characteristics of excluded studies). When there were multiple publications on the same study, reports were collated such that each study (rather than each report) was the unit of analysis. Concurrently, one of two members of the author team (JZ, DS) reviewed all citations excluded when the RCT classifier was added to screen for any RCTs that may have accidently been removed, as per guidance from the information specialist (AP).

Data extraction and management

We used the Cochrane Consumers and Communication Review Group Data Extraction Template (available at: cccrg.cochrane.org/author‐resources) and included criteria from PRISMA Equity (Welch 2015), the Template for Intervention Description and Replication (TIDieR) (Hoffmann 2014), and Guidance for Reporting Involvement of Patients and Public (GRIPP2) (Staniszewska 2017), to extract information on the general study information, and methods. In addition to TIDieR, data were extracted on the decision‐coaching intervention for eight elements: training/protocol; facilitates progress in decision making; focus on specific decision; decision‐making needs assessment/discussion; provides information/discusses options; clarifies values; encourages communication; and screens for implementation. Data on characteristics of participants were extracted using ‘PROGRESS‐Plus’ (place of residence, race/ethnicity/culture/language, occupation, gender/sex, religion, education, socioeconomic status, social capital) and Plus refers to additional categories such as age, sexual orientation and disability which may influence opportunities for health of individuals and populations (Evans 2003; O'Neill 2013; Oliver 2008; Oliver 2012). 

To determine patient and public involvement as partners on the research teams, we used the Guidance for Reporting Involvement of Patients and Public (GRIPP2) (Staniszewska 2017). While not specific to systematic reviews, many of the recommended items have been identified by Cochrane as applicable to systematic reviews (Cochrane Training 2021).

Four members of the review team (JJ, AR, DS, JZ) pilot tested the standardised data extraction form in MS Excel with three studies. Any discrepancies were resolved by discussion that included other team members (SKo, MS, MC) until consensus was reached.

Two members of the review author team (JZ, DS, AR, JF, JJ, SKo) independently extracted data from included studies using the pilot‐tested data extraction form. One review author (JZ) extracted 100% of the data and a second author extracted data on the characteristics of the intervention(s) and outcomes. Any discrepancies were resolved by discussion until consensus was reached, or through consultation with a third review author who did not conduct the data extraction. Review authors did not extract data from their own studies (AR, SKo, JK). 

All extracted data in the standardised data extraction form were transferred into Review Manager (RevMan Web) or Excel tables by one author (DS), and independently checked for accuracy against the data extraction sheets by a second review author (MC).

Assessment of risk of bias in included studies

Two members of the review author team (DS, MC, JF, JJ, SKo, AR, JZ) independently assessed the methodological risk of bias of included studies, in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and the Cochrane Consumers and Communication guidelines (Ryan 2013). These guidelines recommend the explicit reporting of the following individual elements for RCTs: random sequence generation; allocation sequence concealment; blinding (participants, personnel, and outcome assessment); completeness of outcome data, selective outcome reporting; and other sources of bias (such as whether the same healthcare provider delivered the decision‐coaching intervention and the comparator, whether clustering was accounted for in the analysis, and other potential sources of bias reported by the study authors).  We judged each item as being at low, unclear, or high risk of bias, as set out in the criteria provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We provided a quote from the study report and a justification for our judgement for each item in the risk of bias table.

We considered studies to be at high risk of bias if they were judged as having high or unclear risk of bias regarding sequence generation or allocation concealment, based on these factors being particularly important potential sources of bias (Higgins 2011). We considered blinding separately for different outcomes, where appropriate. For example, studies using subjective outcome measures were judged as having high risk of bias when no blinding was used and those using objective outcome measures were judged as having low or unclear risk of bias. Completeness of outcome data (avoidance of attrition bias) was considered separately for different lengths of follow‐up (shorter and longer follow‐up). We considered the outcomes in terms of timing of assessment to determine what data could be analysed together in meta‐analysis. If multiple time points were reported, then we considered data in relation to time points (e.g. knowledge should be measured soon after exposure to the intervention versus knowledge measured at one year).

For cluster‐RCTs, we assessed and reported the risk of bias associated with an additional domain: 'Selective recruitment of cluster participants.' 

In all cases, two review authors independently assessed the risk of bias of included studies, with any disagreements resolved by discussion with a third author to reach consensus.

Measures of treatment effect

For continuous measures, we analysed data using a random‐effects model to calculate the mean difference (MD) and 95% CI. If more than one study measured the same outcome using different tools, we verified that the data could be merged, standardised the findings and used a random‐effects model to calculate the standardised mean difference (SMD) and 95% CI. When data could not be pooled, we reported outcomes and strength of findings according to the original study.

Unit of analysis issues

For cluster‐RCTs or studies with multiple treatment groups, we applied the methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2017). If cluster‐RCTs or studies with multiple treatment groups were included, we checked for unit of analysis errors. Where errors were found and sufficient information was available, we planned to re‐analyse the data using the appropriate unit of analysis, by taking account of the intra‐cluster correlation coefficient (ICC). We also planned to obtain estimates of the ICC by contacting authors of included studies, or to impute them using estimates from external sources. If it was not possible to obtain sufficient information to re‐analyse the data, then we would not include studies in meta‐analysis. In case of a narrative summary, we planned to annotate the unit of analysis error. Of the two cluster‐RCTs, Hamann 2006 did not account for clustering in the analysis of data and Berger‐Höger 2019 accounted for clustering in the analysis of data.

No studies were included in meta‐analysis that compared more than one intervention group to one control group.

Dealing with missing data

For participant data, where possible, we planned to conduct analysis on an intention‐to‐treat basis; otherwise data were analysed as reported. We attempted to contact study authors when there were missing data. We reported on the levels of loss to follow‐up data as attrition bias.

Assessment of heterogeneity

For meta‐analysis findings, we assessed the degree of heterogeneity by visual inspection of forest plots and by examining the Chi² test and the I² statistic for heterogeneity for each outcome (Gentles 2013; Hatala 2005). An I² value of 50% or more was considered to represent substantial heterogeneity. The I² value was interpreted in relation to the size and direction of effects as well as the strength of the evidence for heterogeneity, based on the Chi² test (Deeks 2017). 

Assessment of reporting biases

To minimise the risk of publication bias, we performed comprehensive searches in multiple databases, including searching for unpublished studies in progress. Due to the small number of studies included for each endpoint, we did not use a funnel plot to investigate the likelihood of overt publication bias.

We assessed reporting bias qualitatively, based on the characteristics of the included studies (e.g. if only small studies that indicated positive findings were identified for inclusion) and if information obtained by contacting experts and study authors indicated there were relevant unpublished studies. We considered publication of registered protocols to be an indicator that studies were reported as originally planned. 

Data synthesis

Where studies were considered similar enough (based on consideration of populations, interventions, outcomes), we conducted meta‐analysis of primary and secondary outcomes, using RevMan Web (RevMan Web 2021). When possible, meta‐analysis of outcomes was conducted based on whether the intervention(s) in the included studies were similar (in terms of participants, settings, intervention, comparison and outcome measures) and provided meaningful conclusions from the statistically pooled results.

For the synthesis, the time points post‐coaching ('post‐intervention') were compared between studies. We used the post‐intervention results in meta‐analysis or descriptively. We were able to apply this approach to all the studies, except one study that provided two post‐intervention measures (Brown 2019) where no data were reported and the authors indicated that the measures (e.g. decision regret) were likely used too soon after the intervention to detect differences. 

Due to the variability in the populations, interventions, and different instruments to measure the outcomes (i.e. knowledge) we used a random‐effects model for meta‐analysis. Otherwise, findings were synthesised without meta‐analysis.

Subgroup analysis and investigation of heterogeneity

Where heterogeneity was present in pooled effect estimates, we had planned to explore possible reasons for variability by conducting subgroup analysis. We had proposed subgroup analyses based on the Ottawa Decision Support Framework theory (O'Connor 1998; Stacey 2020), and the expertise of members of the author team who are active in the development, implementation, and evaluation of decision coaching: types of decisions (treatment, screening, diagnostic); method of intervention delivery (in person, by telephone, internet‐based); and characteristics of participants who are recipients of the intervention and who delivered the intervention, i.e. sex and/or gender, education, socioeconomic status, occupation of decision coaches (healthcare providers who are health support workers versus healthcare professionals). We did not conduct any subgroup analyses or investigate heterogeneity due to small numbers of included studies in each of the review's comparisons.

Sensitivity analysis

Sensitivity analyses was planned to assess the robustness of the results, to examine the effect of including compared to excluding studies at high risk of bias. For example, if different decisions are made about the analysis, how much does this affect the results? Sensitivity analyses were also planned to assess the effects of any imputed data on pooled effect estimates. We did not conduct sensitivity analyses due to the small numbers of included studies within comparisons.

Summary of findings and assessment of the certainty of the evidence

We prepared a summary of findings table to present the results of meta‐analyses syntheses and/or narrative synthesis for each primary outcome (e.g. preparation in health decision making), including potential harms, outlined in Types of outcome measures (Ryan 2016). Two reviewers independently used the GRADE criteria to rate the strength of the evidence for each major outcome. We provided a source and rationale for each assumed risk cited in the tables, and used the GRADE criteria to rank the certainty of evidence based on the methods described in chapter 11 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011), using GRADEpro GDT software (GRADEpro GDT 2021; Schünemann 2013). 

Results

Description of studies

Results of the search

The search strategy yielded 12,984 citations (Figure 1), 12,702 were excluded after title and abstract screening, and 234 excluded after full‐text screening (see Characteristics of excluded studies). We included 28 studies, identified 16 ongoing studies (see references to Ongoing studies and Characteristics of ongoing studies) and four studies are awaiting classification (see Characteristics of studies awaiting classification).

1.

Included studies

We included 28 studies with a total of 5509 participants (see Characteristics of included studies). Of the 28 included studies, 26 were RCTs with individuals randomised and two were RCTs with cluster‐randomisation (Berger‐Höger 2019; Hamann 2006) (Table 5). The studies were conducted in the USA (n = 13), Canada (n = 4), Germany (n = 4), UK (n = 4), Australia (n = 1), Japan (n = 1), and the Netherlands (n = 1). The studies focussed on the use of decision coaching to support healthcare decisions about treatment (n = 18) or screening (n = 10). Treatment decisions were about cancer (n = 6), menopause (n = 2), mental illness (n = 2), advanced kidney disease (n = 1), lumbar stenosis (n = 1), birth control (n = 1), menorrhagia (n = 1), diabetic foot (n = 1), multiple sclerosis (n = 1), osteoarthritis (n = 1), and embryo transfer (n = 1). Other healthcare decisions were about cancer screening (n = 4) and genetic testing (n = 6) (Table 5). We reported on funding for each study (see Table 6).

1. Characteristics of included studies (N = 28).

First Author Year	Country	Study purpose	Decision	Setting	Length of follow‐up	Participants randomised
Adam 2019	Canada	To compare effectiveness and acceptability of DECIDE to conventional genetic counselling, using a non‐inferiority model, in families considering genomic testing for their child with early‐onset epilepsy	Genetic testing for epilepsy	1 epilepsy clinic	2 weeks	54 (coaching) versus 52 (PtDA)
Aoki 2019	Japan	To evaluate the effects of a 7‐day SDM program for outpatients with a first episode of mood disorders among university students	Treatment of mood disorders (depression, bipolar)	1 mental health outpatient centre	6 months	35 (coaching + PtDA) versus 53 (usual care)
Berger‐Höger 2019*	Germany	To investigate whether an informed SDM intervention for women with ductal carcinoma in situ comprising an evidence‐based PtDA with nurse‐led decision coaching enhances the extent of mutual SDM behaviour of patients and professionals about treatment options	Treatment of ductal carcinoma in situ	16 certified breast care centres	2 months	37 (coaching + PtDA) versus 30 (usual care)
Bozic 2013	USA	To evaluate the impact of decision and communication aids on patient knowledge, efficiency of decision making, treatment choice	Treatment of hip/knee osteoarthritis	2 academic medical centres	6 weeks after surgeon consult	95 (coaching + PtDA) versus 103 (usual care)
Brown 2019	Australia	To evaluate the effectiveness of a decision support intervention in facilitating the older person with advanced kidney disease to make a treatment choice	Treatment of advanced kidney disease	4 public hospital renal programmes	12 weeks	19 (coaching + PtDA) versus 22 (usual care)
Col 2007	USA	To compare effectiveness of individualised evidence‐based information with standard educational materials on decisions about menopausal treatments	Treatment of menopausal symptoms	2 internal medicine practices and 1 OB/GYN	2 weeks	50 (coaching + EBinfo) versus 45 (EBinfo) versus 50 (usual care)
Davison 1997	Canada	To explore the hypothesis that assisting men with prostate cancer to obtain information would enable them to assume a more active role in treatment decision making and decrease their levels of anxiety and depression	Treatment of prostate cancer	1 community urology clinic	6 weeks	30 (coaching + PtDA) versus 30 (usual care)
Deschamps 2004	Canada	To compare the effects of pharmacist consultation versus a PtDA on women's decisional conflict regarding use of hormone replacement therapy and subsequent satisfaction with the decision‐making process	Treatment of menopausal symptoms	1 family medicine clinic	12 months	67 (coaching) versus 61 (PtDA)
Green 2001	USA	To compare face‐to‐face education and counselling by a genetic counselor with education by an interactive computer program, assessing the effects of each on knowledge of breast cancer genetics and intent to undergo genetic testing	Genetic testing for breast cancer	1 community clinic, recruited through newspapers	post‐intervention	29 (coaching) versus 29 (PtDA) versus 14 (usual care)
Hacking 2013	Scotland	To evaluate the efficacy of the decision navigation intervention on patients’ confidence in making decisions, certainty in decisions made during consultations, and impact on mood and adjustment	Treatment decisions for early stage‐prostate cancer	1 hospital diagnostic clinic	pre‐consultation and 6 months	63 (coaching) versus 60 (usual care)
Hamann 2006*	Germany	To assess an intervention designed to facilitate SDM among acutely ill inpatients with schizophrenia	Treatment for schizophrenia	12 acute psychiatric wards of 2 state hospitals	18 months	54 (coaching + PtDA) versus 59 (usual care)
Hunter 2005	Canada	To determine whether alternative approaches to genetic counselling, (a) individual counselling (b) group counselling, and (c) use of a PtDA with the opportunity to speak with a genetic counsellor, are equally effective in prenatal diagnostic testing counselling in advanced maternal age women/partners	Prenatal diagnostic testing	1 regional genetics clinic	at 24 weeks gestation	126 (coaching) versus 116 (PtDA)†
Kearing 2016	USA	To assess the impact of telephone health coaching in addition to a video PtDA compared to a PtDA alone for patients with spinal stenosis	Treatment of lumbar spinal stenosis	1 orthopaedic surgeon or non‐surgical spine clinic	6 months	98 (coaching + PtDA) versus 101 (PtDA)
Kennedy 2002	UK	To evaluate the effects of information with and without a structured preference elicitation interview, on treatment choices, health outcomes, and costs	Treatment of uncomplicated menorrhagia	Specialist clinic at 6 hospitals	2 years	300 (coaching + PtDA) versus 296 (PtDA) versus 298 (usual care)
Langston 2010	USA	To evaluate the addition of structured contraceptive counselling to usual care on choice, initiation, and continuation of very effective contraception after uterine aspiration	Treatment to avoid pregnancy	1 family planning referral clinic to be seen at 1 private practice	3 months	114 (coaching + PtDA) versus 108 (usual care)
Lepore 2012	USA	To evaluate the efficacy of a decision‐support intervention focussed on prostate cancer testing in a sample of predominantly immigrant black men	Screening for prostate cancer	Insurance beneficiaries of large healthcare workers union, study conducted via telephone	2 years	244 (coaching + PtDA) versus 246 (usual care)
Lerman 1997	USA	To evaluate the impact of alternate strategies for pretest education and counselling on decision making regarding BRCA1 testing among women at low‐to‐moderate risk who have a family history of breast and/or ovarian cancer	Genetic testing for breast and/or ovarian cancer	2 cancer centres in Washington DC	1 month	122 (coaching + PtDA) versus 114 (PtDA) versus 164 (waiting‐list control)
Matloff 2006	USA	To examine if a personalised risk assessment and genetic counselling intervention developed for this pilot study would affect knowledge, risks perception, and decision making in a group of women who had one first‐degree relative with breast cancer, as compared with a control group	Genetic testing	Setting not reported	6 months	32 (coaching) versus 32 (usual care)
McBride 2016	UK	To explore whether the use of an intervention to increase SDM increased decision self‐efficacy and foot‐treatment adherence in patients with a diabetic foot ulcer	Treatment of diabetic foot ulcer	1 diabetes foot clinic	12 weeks	30 (coaching + PtDA) versus 26 (usual care)
Miller 2005a	USA	To develop and evaluate a theory‐based educational intervention designed to increase callers’ understanding of inherited risk, their family history of cancer, and benefits/limitations of genetic testing	Genetic testing	National Cancer Institute’s Cancer Information Service	6 months	279 women randomised to coaching plus PtDA versus usual care (numbers per group not reported)
Mishel 2009	USA	To examine the effects of a theory‐based decision‐making uncertainty management intervention providing newly diagnosed prostate cancer patients with information, communication skills and personally designed prompts	Treatment of prostate cancer	2 cancer centres, 3 community hospitals, 1 Veterans' Administration medical centre	3 months	93 (coaching + written information + video) versus 74 (usual care)‡
Myers 2005	USA	To test the impact of an informed decision‐making intervention on prostate cancer screening use	Screening for prostate cancer	3 primary care practices	11 months	121 (coaching plus PtDA) versus 121 (PtDA)
Myers 2011	USA	To assess the impact of a mediated decision support intervention on primary care patient prostate cancer screening knowledge, decisional conflict, informed decision making, and screening	Screening for prostate cancer	2 primary care practices	120 days	156 (coaching plus PtDA) versus 157 (PtDA)
Rahn 2018	Germany	To test the feasibility of a decision‐coaching programme and recruitment strategies to inform the main trial	Immunotherapy decisions for multiple sclerosis	2 multiple sclerosis centres	6 months	38 (coaching plus EBinfo) versus 35 (EBinfo)
Shepherd 2019	Scotland	To evaluate the effectiveness of Consultation Planning, Recording and Summarising on self‐reported decision self‐efficacy, conflict, and regret, anxiety and depression, in patients with colorectal cancer  for up to 3 care planning consultations and 3 months follow‐up, compared with the usual care experience	Treatment of colorectal cancer	1 clinic in a  cancer centre	3 months after discharge from the clinic	68 (coaching) versus 69 (usual care)
Sheridan 2012	USA	To examine the effects of a prostate cancer screening intervention to promote SDM and determine whether framing prostate information in the context of other clearly beneficial men's health services affects decisions	Prostate cancer screening	2 primary care practices (1 academic, 1 community practice)	9 months	60 (coaching + video + information) versus 70 (video on seat belt safety)
Van Peperstraten 2010	Netherlands	To evaluate the effects of a multifaceted empowerment strategy on the actual use of single embryo transfer after in vitro fertilisation	Treatment embryo transfer	5 in vitro fertilisation clinics (2 hospitals, 3 clinics)	5 weeks	152 (coaching + PtDA) versus 156 (usual care)
Vodermaier 2009	Germany	To examine the impact of a PtDA intervention in breast cancer inpatients in Germany	Treatment for breast cancer	1 gynaecology  department at university hospital	1 week	74 (coaching + PtDA) versus 78 (usual care)

Abbreviations: 
^{BRCA1=Breast cancer type 1 susceptibility protein 
EBinfo=evidence‐based information
PtDA=patient decision aid
SDM=Shared Decision Making}

* Cluster RCT

^† Excluded one comparator arm because of group coaching

^‡ Excluded one comparator arm because the intervention measured family support

2. Study funding sources (N = 28 studies).

First Author Year	Source of funding
Adam 2019	This research was performed as part of the study “Paediatric Epilepsy: Using Genomics to Improve Patient Care and Outcomes,” that was supported by the Canada Excellence Research Chair (MF) and the Alva Foundation (MC). The DECIDE research was funded by a Canadian Institutes of Health Research grant to JMF and a Canadian Institutes of Health Research New Investigator award to NB.
Aoki 2019	This research did not receive any specific grant from funding agencies in the public, commercial, or not‐for‐profit sectors
Berger‐Höger 2019	The German Federal Ministry of Health funded the study within the National Cancer Action Plan (Grant No. NKP – 332 – 054).
Bozic 2013	This work was supported by a grant from the RobertWood Johnson Foundation (RWJF).
Brown 2019	National Health and Medical Research Council; Australian Centre for Health Service Innovation (AusHSI); Queensland Health Nursing and Midwifery Research Fellowship; Chronic Kidney Disease Centre of Research Excellence; Sunshine Coast Hospital and Health Service; Wide Bay Hospital and Health Service
Col 2007	This work was supported by grants from the Robert Wood Johnson Foundation (RWJ # 46461) and AHRQ R01 HS013329‐01.
Davison 1997	Supported by a studentship from the National Cancer Institute of Canada with funds provided by the Canadian Cancer Society to the first author, and by an investigator award from the Medical Research Council of Canada and the National Health Research and Development Program to the second author.
Deschamps 2004	Financial support for this study was provided in part by an unrestricted research grant from Eli Lilly Canada Inc. The funding agreement ensured the authors’ independence in designing the study, interpreting the data, and writing and publishing the report.
Green 2001	This publication was supported by grant number 1R03 CA 70638 from the National Cancer Institute (NCI), and grant number 1 R01 CA84770 from NCI and the National Human Genome Research Institute (NHGRI).
Hacking 2013	Macmillan Cancer Support funded this study in its entirety.
Hamann 2006	The trial was funded by the German Ministry of Health and Social Security (217‐43794‐5/9) within the funding project ‘Der Patient als Partner im medizinischen Entscheidungsprozess’.
Hunter 2005	Not reported
Kearing 2016	The Informed Medical Decisions Foundation, Boston MA and the National Institute for Arthritis, Musculoskeletal, and Skin Diseases #P60AR062799 grant funds were received in support of this work.
Kennedy 2002	Our research was supported by a grant from the UK NHS research and development health technology assessment programme. The health economics research group receives funding from the UK Department of Health. Dr Sculpher received a career scientist award in public health funded by the NHS research and development programme.
Langston 2010	Financial support provided by a grant from an anonymous foundation
Lepore 2012	This research was supported by grant R01 CA104223 from the National Cancer Institute of the National Institutes of Health
Lerman 1997	Supported by Public Health Service grants (RO1MH/HG54435) from the National Institutes of Mental Health and the National Center for Human Genome Research, National Institutes of Health Department of Health and Human Services
Matloff 2006	This study was funded by a grant from the Susan G. Komen Foundation.
McBride 2016	This work was supported by NHS Lothian and NHS Education for Scotland.
Miller 2005a	This research was supported in part by the Department of Defense DAMD17‐98‐1‐8306, DAMD17‐01‐1‐0238, and DAMD17‐02‐1‐0382 grants, the Fox Chase Cancer Center’s Behavioral Research Core Facility (P30CA06927), and NIH grant R01HG01766.
Mishel 2009	The work submitted here was funded by NIH, NINR (5RO1NR08144) and was entitled Decision Making under Uncertainty in Prostate Cancer, Merle Mishel, PI.
Myers 2005	The work reported in this paper was supported by a grant from the Quality Care Research Fund, Aetna Foundation.
Myers 2011	The Decision Counseling Trial was conducted under AAMC/CDC cooperative agreement grant MM‐0554‐03.
Rahn 2018	This study was funded by the German Ministry of Education and Research within the KKNMS (01GI1206).
Shepherd 2019	Coventry University, Grant/Award Number: PhD studentship; Macmillan Cancer Care; NHS Lothian
Sheridan 2012	This research was supported by the Centers for Disease Control and Prevention (CDC, #TS0845).
Van Peperstraten 2010	This study was funded by the Netherlands Organisation for Health Research and Development (grant No 945‐16‐105). All researchers are independent from this source of funding. The study sponsor had no role in the study design, collection, analysis, and interpretation of data, the writing of the article, and the decision to submit it for publication.
Vodermaier 2009	This work was supported by the German Ministry of Health as a pre‐operating study in the focus programme ‘The Patient as a Partner in the Medical Decision Making Process’ under Grant no. 217‐43794‐5/2 (Professor Dr Michael Untch, PI) and by a stipend from the Dr‐Werner‐Jackstaedt‐Stiftung in the Founder Association of the German Sciences under Grant no. S134‐10.021 (Dr Andrea Vodermaier).

Characteristics of participants

Participants’ mean age was 48.1 years (n = 25 studies; 4728 participants) ranging from 18 to 85 years (n = 7 studies; 774 participants) (see Table 7, Characteristics of participants). Participants were 64.1% female and 35.9% male (n = 28 studies; 5373 participants). Reported race and ethnicity included: white (52.1%), African‐American/black (33.4%), Hispanic (6.7%) or other/unknown (7.7%) (n = 16 studies; 3080 participants). Participant education levels were reported as post‐secondary (67.6%) or secondary school diploma or less (32.4%) (n = 21 studies; 3458 participants).

3. Characteristics of participants (N = 28 studies).

Author (year)	Age (mean)*	Male:female (per cent)	Ethnicity	Education <= HS:> HS	Income	Employment	Civil status
Adam 2019	n/r	14:86	n/r	25 (23.8%): 80 (76.2%)	n/r	n/r	n/r
Aoki 2019	22.0	55:45	n/r	0 (0.0%): 88 (100.0%)	n/r	All students	n/r
Berger‐Höger 2019†	58.1	0:100	n/r	45 (72.6%): 17 (27.4%)	n/r	n/r	Married 45 (72.6%) Partnership 3 (4.8%) Divorced 7 (11.3%) Widowed 5 (8.1%) Single 2 (3.2%)
Bozic 2013	63.1	46:54	Non‐Hispanic 41 (73.2%) Hispanic 9 (7.3%) Unknown 24 (19.5%)	15 (12.2%): 108 (87.8%)	< $50K 36 (29.3%) ≥ $50 ‐ 100K 37 (30.1%) > $100K 45 (36.6%) Missing 5 (4.1%)	Yes 56 (45.5%) No 9 (7.3%) Retired 44 (35.8%) Disabled 10 (8.1%) Missing 4 (3.3%)	n/r
Brown 2019	77.9	0:100	n/r	n/r	n/r	Work 3 (7.3%) Pension 30 (73.2%) Self‐funded retiree 8 (19.5%)	n/r
Col 2007	52.3	0:100	White 93 (97.9%) Not reported 2 (2.1%)	67 (46.2%): 78 (53.8%)	n/r	n/r	Married or partnered 108 (74.5%)
Davison 1997	67.9	100:0	n/r	35 (58.3%): 25 (41.7%)	n/r	Full time 12 (20.0%) Part time 5 (8.3%) Retired 43 (71.7%)	Married or partnered 52 (86.7%)
Deschamps 2004	n/r	0:100	White 104 (99.0%) Not reported 1 (1.0%)	20 (19.0%): 85 (81.0%)	n/r	Technical 37 (35.2%) Professional 37 (35.2%) Not reported 31 (29.6)	Married or partnered 90 (85.7%)
Green 2001	43.8	0:100	White 70 (97.2%) African‐American 2 (2.8%)	3 (4.2%): 69 (95.8%)	< $35K 7 (9.7%) ≥ $35‐55K 12 (16.7%) > $55K 53 (73.6%)	Work outside the home 56 (77.8%) Do not work outside the home 16 (22.2%)	n/r
Hacking 2013	66.3	100:0	Caucasian 115 (100%)	Reported as education ≥ 16 years	n/r	Employed (27%) Unemployed/retired (73%)	Living with partner 91 (79.1%)
Hamann 2006†	37.7	55:45	n/r	Reported as education ≥ 10 years	n/r	n/r	n/r
Hunter 2005	36.8	0:100	British/European (83.9%) Asian, Aboriginal, Black, Middle Eastern, Ashkenazi Jewish (5.1%) Unknown (11%)	Reported as means using 8‐point scale	Reported as means using 11‐point scale	Yes 211 (87.2%) Not reported 31 (12.8%)	n/r
Kearing 2016	66.7	52:48	White 164 (97.6%) Not reported 4 (2.4%)	54 (32.1%): 114 (67.9%)	n/r	Retired 83 (49.4%) Not reported 85 (50.6%)	Married 118 (70.2%)
Kennedy 2002	40.3	0:100	n/r	Reported as age leaving full‐time education	n/r	n/r	n/r
Langston 2010	26.2	0:100	Hispanic 195 (88.2%) Not reported 26 (11.8%)	71 (32.0%): 151 (68.0%)	n/r	n/r	Stable relationship 155 (69.8%)
Lepore 2012	55.0	100:0	Black 490 (100%)	309 (63.1%): 181 (36.9%)	n/r	n/r	Married 410 (83.7%)
Lerman 1997	n/r	0:100	White 283 (70.9%) African‐American 106 (26.6%) Other 10 (2.5%)	40 (10.0%): 359 (90.0%)	≤ $35K 79 (20.3%) ≥ $35K‐$50K 67 (17.2%) ≥ $50K 243 (62.5%)	n/r	Married 250 (62.5%) Unmarried 150 (37.5%)
Matloff 2006	49.0	0:100	European (54%) Ashkenazi Jewish (21%) Both Ashkenazi Jewish and European non‐Ashkenazi Jewish (8%) Latin American/Caribbean (2%) Other (4%) Unknown (10%)	11 (17.2%): 53 (82.8%)	≥ $75,000 (58%)	n/r	Married (71%)
McBride 2016	61.1	73:27	White 56 (100%)	26 (46.4%): 30 (53.6%)	n/r	Full‐time 11 (19.6%) Part‐time 4 (7.1%) Retired 31 (55.4%) No 6 (10.7%) Student 1 (1.8%) Other 3 (5.4%)	Married 32 (57.1%) Partner 4 (7.1%) Divorced 4 (7.1%) Separated 3 (5.4%) Widowed 3 (5.4%) Single 10 (17.9%)
Mishel 2009	62.5	100:0	Caucasian 119 (71.3) Black 48 (28.7)	Reported as mean years of education	< $6K 4 (2.6%) ≥ $6K–12K 11 (7.1%) ≥ $12K–24K 21(13.6%) ≥ $24K–48K 45 (29.0%) > $48K 74 (47.7%)	Yes 97 (58.1%) Not reported 70 (41.9%)	Married or partnered 132 (79.0%)
Miller 2005a	46.2	0:100	Caucasian 227 (88.7%) African American 14 (5.5%) Native American 4 (1.6%) Hispanic 3 (1.2%) Asian 2 (0.7%) Other 6 (2.3%)	75 (29.4%): 180 (70.6%)	n/r	n/r	n/r
Myers 2005	52.0	100:0	African‐American 242 (100%)	Reported as < or > 12 years of education	n/r	n/r	Married 152 (62.8%) Not married 90 (37.2%)
Myers 2011	56.5	100:0	White 176 (56.4%) African‐American 128 (41.0%) Other 8 (2.6%)	101 (32.6%): 209 (67.4%)	n/r	n/r	Married 197 (62.9%) Not married 116 (37.1%)
Rahn 2018	37.3	26:74	n/r	26 (40.6%): 38 (59.4%)	n/r	At least part‐time 50 (78.1%) Not reported 14 (21.9%)	n/r
Shepherd 2019	62.1	59:41	White 129 (97.7%) Chinese 1 (0.76%) Mixed 1 (0.76%) Other 1 (0.76%)	65 (49.6%): 66 (50.4%)	n/r	Full time 39 (29.6%) Part time 16 (12.1%) Not reported 77 (58.3%)	Married 92 (69.7%) Divorced 18 (13.6%) Separated 3 (2.3%) Widowed 6 (4.6%) Single 13 (9.8%)
Sheridan 2012	57.6	100:0	White 70 (54.7) Not reported 58 (45.3)	41 (32.0%): 87 (68.0%)	n/r	n/r	Married 78 (60.9%)
Van Peperstraten 2010	31.9	0:100	n/r	7 (2.3%): 297 (97.7%)	€13200 ‐ 21120 6 (2.0%) ≥ €21120 ‐ 33000 37 (12.1%) > €33000 212 (69.5%) Did not want to tell 40 (13.1%)	n/r	n/r
Vodermaier 2009	55.2	0:100	n/r	85 (79.4%): 22 (20.6%)	n/r	Yes 52 (47.3%) No 58 (52.7%)	Married/cohabiting 81 (73.0%)

HS=High School

*Mean age of the sample was calculated using the group means for 16/24 studies

^† Cluster‐RCT

 

Description of the intervention 

In the 28 studies, there were four comparisons:

• decision coaching compared with usual care (n = 4 studies);

• decision coaching compared with evidence‐based information only (n = 4 studies);*

• decision coaching plus evidence‐based information* compared with usual care (n = 17 studies); and,

• decision coaching plus evidence‐based information* compared with evidence‐based information only (n = 7 studies).

*evidence‐based information includes patient decision aids.

Four of the 28 studies included 3 comparator arms (Col 2007; Green 2001; Kennedy 2002; Lerman 1997). The evidence‐based information was typically a patient decision aid (21 studies) or more general evidence‐based information (4 studies).

 

Characteristics of decision coaching

Decision coaching was provided by nurses (n = 10), genetic counsellors (n = 4), psychologists (n = 3), a pharmacist (n = 1), a mix of healthcare professionals (n = 2), or others trained in decision coaching (n = 8). Most studies reported coaching occurred in‐person (n = 16) and/or by telephone (n = 12) (see Table 8). 

4. TIDieR Intervention Summary (N = 28 studies).

Author (year)	Brief name and why	What was DC intervention	Who provided DC	How, where, when, how much	Tailoring or modifying DC	Strategies for intervention fidelity
Adam 2019	Genetic counselling for families about genomic testing	DC used checklist with pros and cons of testing (same as PtDA), discussed talking to family and friends	genetic counsellor	Once for 1 hour DC in person at an epilepsy clinic or on telephone once for 1 hour	n/r	checklist and factual content same as the PtDA
Aoki 2019	DC using 7‐day SDM program to increase patient‐perceived involvement in decision making	Prior to DC, clinician gave options and patient reviewed PtDA. DC discussed options, answered questions, asked patient to give their preference	nurse	After PtDA, one 20‐30 minute DC in person at a mental health outpatient centre or on the telephone between clinician consults	n/r	extent of intervention fidelity was monitored
Berger‐Höger 2019*	DC using 6‐step SDM program to support women with breast cancer	DC defined problem requiring a decision; described SDM, gave PtDA, clarified patients' values, gave option to postpone decision, offered prompt cards	specialised nurses	Once for 47.5 minutes DC in person at 16 certified breast care centres on average 4.3 days before MD consult	n/r	Video‐taped DC and MD consults, prompt cards for nurses, summary information on 4 treatment options included in PtDA
Bozic 2013	SCOPED with DC and PtDA to evaluate impact of these SDM tools	After PtDA, DC helped patient write list of  questions for surgeon. After MD consult, sent patient‐ recorded consult and surgeon's dictated note	health coach	After PtDA, one telephone call (length of time n/r) and DC attended 54 minutes consult at an academic medical centre for which 21 minutes was with the surgeon	n/r	n/r
Brown 2019	Decision‐support intervention using OPTIONS (PtDA)	PtDA then DC used the PtDA to support patient in active, autonomous role in decision making	renal nurse	After PtDA, 1 or 2 sessions in person for 45 minutes at a public hospital renal programme	n/r	format was per training in Ottawa Decision Support tutorial
Col 2007	DC plus EBinfo with personal risk estimates	DC reviewed EBinfo to discuss personal risk estimates and the worksheet, prioritise questions and rehearse question asking	coach with Masters in Public Health	Once for 10‐20 minutes DC in person just before MD consult	n/r	protocol, training, EBinfo, document for goals
Davison 1997	Empowerment intervention to help men assert control over the factors that affect their health	DC asked men what information they need for decision making, reviewed example list of questions, showed where to find answers in PtDA, gave final list of questions to men, encouraged to read PtDA, use question list, and participate in decision making	nurse researcher	Once in person DC at 1 community urology clinic (length of time n/r)	n/r	n/r
Deschamps 2004	Pharmacist consultation	Reviewed risks and benefits of medication based on clinical practice guidelines, charts and graphs to visually show population data, agreed on plan, noted in consult letter to MD	pharmacist	Once for 40 minutes DC in person at family medicine study clinic and encouraged to consult MD within 2 to 4 weeks	n/r	charts and graphs to provide consistency between encounters
Green 2001	Genetic counselling	General information about breast cancer risk and heredity; risks/benefits, limitations, implications of genetic testing; possible outcomes, worries, cancer prevention strategies	genetic counsellor	Once for 60‐100 minutes DC in person at a community clinic	Tailored to individual risk	same information as PtDA group
Hacking 2013	Decision navigation for prostate cancer treatment decisions	Coaches assisted patients in identifying and framing key questions and concerns regarding cancer management options to generate a personal consultation plan	two research assistants	Pre‐consultation by telephone or in person, during the consultation, and post‐consultation coaches gave each navigated patient an audio record of the visit and a typed summary	personalised consultation plan	trained by developer of the intervention (2‐day workshop), audio recorded consultation and typed summary, monthly case reviews
Hamann 2006*	SDM group to inform patients about their treatment options and prepare them for a ‘planning talk’ with their MD	Coaches used the PtDA and assisted patients in working through the booklet and answered questions	nurse managers	Once for 30‐60 minutes DC in inpatient acute psychiatric ward of state hospital	n/r	trained, PtDA, physicians trained
Hunter 2005	Individual prenatal diagnosis genetic counselling	Information about risks and limitations, laboratory results, conditions being tested, patient's values and beliefs re birth of an abnormal child and pregnancy termination	genetic counsellor	Once for 1 hour DC in person at regional genetics clinic	n/r	coaching with same content as the PtDA, tape recorded for analysis
Kearing 2016	DC to help patients with their treatment decision process	DC used Ottawa Personal Decision Guide to identify patient treatment intention, knowledge gaps, values clarification, support needs, and planning	Nurse, genetic counsellor, social workers	Within 2 weeks using PtDA, one DC session on the telephone from the Centre for SDM prior to consult with orthopaedic surgeon at spine clinic	n/r	n/r
Kennedy 2002	Structured interview group to help patients clarify and articulate preferences	PtDA (video, booklet) plus DC with values clarification, elicit preference and  desired involvement in decision making	research nurses	Once for 20 minutes (SD 6.2) in person pre‐first MD consult at hospital	Only in building a rapport	highly structured coaching and questions documented on summary form to share with surgeon
Langston 2010	Structured contraceptive counselling	DC using contraceptive flipchart PtDA and samples, wrote down questions for MD	research co‐ordinator	Once for 20 minutes (SD 8 min) DC in person at private practice	n/r	standardised, write questions for MD on note cards
Lepore 2012	Decision‐support intervention for black men about prostate cancer testing	DC established rapport, provided information, helped clarify testing preferences, prepared them to talk to MD	public health graduate students	A week after PtDA sent, up to two telephone calls of 20 minutes (SD 2.3) + 5 min (SD 6.3 ) and prior to MD consult	tailored to knowledge and beliefs	manual, training, treatment fidelity checks in 44% calls
Lerman 1997	Education plus counselling approach to support genetic testing decisions	After PtDA, DC discussed experience with cancer in family, anticipated impact of positive/negative results, anticipated outcomes if not tested, coping resources/skills, intention to tell family	oncology nurses or genetic counsellor	Post‐PtDA, one 30‐minute DC in person at 2 cancer centres in Washington DC	n/r	trained, PtDA, structured protocol
Matloff 2006	Personalised risk assessment and genetic counselling to women considering menopause options	DC using a counselling script using flipbook on benefits/risks of options, future risks of diseases, personalised letter summarising her data (copied to MD)	genetic counsellor	Two DC sessions (60 + 60 minutes) in person (setting n/r)	tailored to questions, concerns, personal history	standardised counselling flipbook, counselling script, letter summarising results
McBride 2016	Decision navigation using SCOPED to increase SDM for diabetic foot care	DC used semi‐structured interview to produce consult plan for MD consult (forwarded to MD), attended MD consult & encouraged use of plan, audio‐recorded consult, written summary of main points	health psychologist	One week after PtDA, DC telephoned from 1 diabetes foot clinic (length of time n/r) to plan for MD consult in 1 week	personalised consultation plan	training, fidelity tests, PtDA, consultation plan
Mishel 2009	Decision‐making uncertainty management intervention to address the stimuli frame concepts of symptom pattern, event congruence, and event familiarity	DC reviewed patient’s questions on prostate cancer information booklet, review handout 'the competent patient', helped patient identify questions on his concerns, reviewed communication skills	nurse	After information materials and DVD on communication strategies, DC provided 4 telephone calls over 7 to 10 days (averaged 20‐30 minutes) and before the treatment consultation with MD	n/r	training, evidence‐based question and answer prostate cancer information, calls were audiotaped for quality control
Miller 2005a	Educational intervention on risk of inherited cancer and benefits/ limitations of genetic testing	DC reviewed PtDA, detailed family history, discussed benefits and limitations of genetic screening, referred to a high‐risk genetic counselling (if appropriate)	Cancer Information Services Specialist	A self‐initiated telephone call to the National Cancer Institute’s Cancer Information Service for 20‐30 minutes	n/r	Training, written PtDA, intervention protocols for standard and enhanced
Myers 2005	Enhanced intervention (decision education) to enable men to consider decision information	DC discussed PtDA information, personal values for alternative options, clarified personal preference, shared calculated preference score, and encouraged to make decision with MD	health educator	One month after receiving PtDA, DC done on telephone or in person from 3 primary care practices in Philadelphia (length of time n/r)	42% (n = 51) did not receive DC because too difficult to contact or set up DC session	training, preference scoring process and algorithm
Myers 2011	Enhanced intervention (decision education) to enable men to consider decision information	DC discussed PtDA information, elicited factors influencing screening decision, computed decision preference score, discussed score, placed note to MD on chart	nurse educator	After receiving the PtDA, DC for 28 minutes (SD14) in person at 2 primary care practices	n/r	training, fidelity tests, PtDA, monthly meetings
Rahn 2018	Decision‐coaching intervention for people with multiple sclerosis	DC using evidence‐based Wiki information on pros/cons of options, patient expectations, decision, next steps	nurses	DC in person at 2 multiple sclerosis centres (length of time n/r) followed by up to two MD consults where decision was made	n/r	training, structured 6 steps, patient workbooks, coaching guide
Shepherd 2019	Consultation planning with summary and audio‐recording (SCOPED)	DC met to identify a list of questions for patient and MD, DC attended 3 clinic appointments to type notes, audio‐record and create plain language typed summary	research psychologists	DC telephone call pre‐consult and in person at MD consult (x3) at a clinic in a tertiary cancer centre	n/r	protocol, manual, training programme
Sheridan 2012	Counsellor‐delivered DC	DC to answer men’s questions about prostate screening, help clarify their values, prepare to discuss with MD, given summary sheet	health counsellor	After information video, one DC for 8 minutes in person prior to MD consult at a primary care practice	tailored to discussion barriers	training, video information, training for physicians, scripted materials, brochure
Van Peperstraten 2010	Multifaceted empowerment strategy involving PtDA, DC, and reimbursement if fourth cycle	DC by trained in vitro fertilisation nurse	vitro fertilisation nurse	After PtDA, one 32‐minute DC session in person at an in vitro fertilisation clinic with follow‐up telephone call before oocytes pick up	n/r	training, PtDA, reimbursement offer
Vodermaier 2009	Decision board coaching	DC with PtDA on invitation to participate in decision making, options, verify understanding, motivated to ask questions in MD consult, discuss with family	research psych‐ologists	Once for 20 minutes in person at a gynaecology  department of a university hospital before surgeon consult within a few hours	n/r	decision board PtDA, information brochure

^{DC=decision coaching
EBinfo=evidence‐based information
MD=physician
n/r=not reported
PtDA=patient decision aid
SCOPED=Situation, Choices, Objectives, People, Evaluation, Decisions
SDM=Shared Decision Making}

*Cluster‐RCT

Twenty studies reported the coaching intervention comprised of a single decision‐coaching session, duration median was 31 minutes with a range from 8 to 120 minutes (Table 8; Table 9). Six studies reported multiple decision‐coaching sessions (2 to 4 sessions) including three where the coach was present for the consultation, duration 20 to 90 minutes for each session, and two studies with one or two sessions. Of 28 studies, only six reported the decision‐coaching intervention was tailored to individual patient/participant needs (Green 2001; Hacking 2013; Lepore 2012; Matloff 2006; McBride 2016; Sheridan 2012). Twenty‐five studies reported strategies to improve or maintain the intervention fidelity.

5. Characteristics of decision coaching (all studies facilitated progress in decision making and clarified the decision) (N = 28 studies).

Author (year)	Theoretical framework	Training	Assess/discuss decision making needs	Provide/discuss information	Clarify values	Encourage communication	Screen for implementation needs
Adam 2019	IPDAS for PtDA	yes checklist	n/r	yes (same as PtDA)	yes (same as PtDA)	n/r	n/a
Aoki 2019	IPDAS	yes	n/r	PtDAs	yes	yes	n/r
Berger‐Höger 2019*	6 steps of shared decision making (Kasper 2012)	Yes	n/r	PtDA	yes	yes	n/r
Bozic 2013	SCOPED	Yes	n/r	PtDA	yes, PtDA	yes	n/r
Brown 2019	ODSF	Yes	yes	PtDA	yes, worksheet	yes	n/r
Col 2007	n/r	Yes	n/r	yes	yes, PtDA	yes	n/r
Davison 1997	Empowerment Model (Conger 1988)	No	yes	PtDA	PtDA	yes	n/r
Deschamps 2004	ODSF for PtDA	No	n/r	yes	no	yes ‐ consult letter	n/r
Green 2001	n/r	Yes	yes	yes	yes	n/r	n/r
Hacking 2013	SCOPED	Yes	yes	identified questions/concerns about options for MD consult	yes	yes	yes, consultation plan
Hamann 2006*	n/r	Yes	n/r	PtDA	PtDA	yes	n/r
Hunter 2005	ODSF	No	n/r	yes	yes	n/r	n/r
Kearing 2016	ODSF	Yes	yes	PtDA	yes	n/r	n/r
Kennedy 2002	n/r	Yes	yes	PtDA	yes	yes	n/r
Langston 2010	n/r	Yes	yes	PtDA	yes, PtDA	yes	n/r
Lepore 2012	ODSF	Yes	yes	PtDA	yes	yes	yes, speaking to physician
Lerman 1997	Kessler 1979 genetic counselling, Janis  1977 decision making, Engel 1995 consumer behaviour	Yes	n/r	PtDA	yes	no	n/r
Matloff 2006	Health belief model	No	yes	yes	no	no	n/r
McBride 2016	SCOPED	Yes	yes	PtDA	yes, PtDA	yes‐ consultation plan	yes, consultation plan
Miller 2005a	Cognitive‐Social Health Information Processing (C‐SHIP) model	Yes	n/r	yes (PtDA)	PtDA	n/r	n/r
Mishel 2009	Uncertainty in Illness Theory	Yes	yes	evidence‐based booklet	n/r	yes	n/r
Myers 2005	Preventive Health Model	Yes	yes	PtDA	yes	yes	n/r
Myers 2011	Preventive Health Model	Yes	yes	PtDA	yes	no	n/r
Rahn 2018	Shared Decision Making (Elwyn 2001 )	Yes	yes	evidence‐based Wiki	yes	yes	yes
Shepherd 2019	SCOPED	Yes	n/r	no	no	yes	n/r
Sheridan 2012	n/r	Yes	yes	yes	yes	yes	yes
Van Peperstraten 2010	IPDAS	Yes	n/r	PtDA	yes	n/r	yes
Vodermaier 2009	n/r	n/r	yes	PtDA	yes	yes	n/r

Note: All studies reported “facilitate progress in decision making” and “focus on specific decision”.

^{C‐SHIP=Cognitive‐Social Health Information Processing
IPDAS=International Patient Decision Aid Standards
n/r=not reported
ODSF=Ottawa Decision Support Framework
PtDA=patient decision aid
SCOPED=Situation, Choices, Objectives, People, Evaluation, Decisions
SDM=Shared Decision Making}
 

* Cluster‐RCT

Most studies cited theoretical frameworks as a foundation for the decision‐coaching intervention. Frameworks which included the Ottawa Decision Support Framework (n = 5), SCOPED (Situation, Choices, Objectives, People, Evaluation, Decisions) (Belkora 2015; Belkora 2021) (n = 4), IPDAS (International Patient Decision Aid Standards) (Elwyn 2006) (n = 3), other individual theories (n = 8), or a combination of theories (n = 1). Seven studies did not identify a theoretical framework.

Of the eight elements of decision coaching (training/protocol, facilitate progress in decision making, focus on specific decision, decision‐making needs assessment/discussion, provide information/discuss options, clarify values, encourage communication, screen for implementation), studies reported between four and eight elements. All studies reported 'facilitate progress in decision making' and 'focus on specific decision' (data not shown). The least reported (n = 3 studies) was 'screen for implementation needs' (see Table 9. Characteristics of decision coaching).

Excluded studies

We excluded 234 studies after close scrutiny of the relevant papers (see Characteristics of excluded studies). The reasons for exclusion were: not decision coaching, wrong publication type (e.g. conference abstract, thesis), decision coaching in both intervention arms, wrong study design, physician counselling, lifestyle choice, advanced care planning, not non‐directive, no actual decision, group counselling, hypothetical decision, genetic counselling not described, wrong patient population.

Risk of bias in included studies

All 28 studies had at least some risk of bias. We report the details for risk of bias in the Characteristics of included studies, Figure 2 and Figure 3.

2.

3.

Allocation

Risk of bias for random sequence generation was low in 17 studies (60.7%), and unclear in 11 studies (39.3%). For concealment of allocation, risk of bias was low in 15 studies, (53.6%), and unclear in 13 studies (46.4%) (see Figure 2).

Blinding

Risk of bias for performance bias was low in six studies (21.4%), unclear in 20 studies (71.4%), and high for two studies (7.1%). For the high‐risk studies, one did not describe blinding and reported self‐assessed outcomes (Brown 2019). The other study reported that envelopes were opened by a research team member after baseline measures (McBride 2016), and group allocation was communicated to both participants and the clinical team. Risk of bias for outcome assessment was low for 18 studies (64.3%), unclear for nine studies (32.1%), and high for one study (3.6%) as outcome‐assessor blinding was not explicitly described (Berger‐Höger 2019) (see Figure 2).

Incomplete outcome data

Risk of bias for attrition was low in 14 studies (50%), and unclear for 13 studies (46.4%). There was high risk of bias for one study (3.6%) due to missing outcome data with imbalance between groups for outcomes measured post‐first consultation (12% decision coaching versus 28% usual care), and an overall high rate of attrition over time (44.5%) (Shepherd 2019) (see Figure 2).

Selective reporting

Risk of bias for reporting bias was low for eight studies (28.6%) and unclear for 20 studies (71.4%).

Other potential sources of bias

Risk of bias for other potential sources of bias was low for 22 studies (78.6%), unclear for four (14.3%), and high for two studies (7.1%). We rated two studies as having high risk of bias due to instrument limitations, overlap of personnel between intervention and control groups, disproportionate randomisation, and loss to follow‐up (Aoki 2019); and clustering not accounted for in the analysis of data (Hamann 2006). For the two cluster‐RCTs, risk of bias for selective recruitment of cluster participants was low in one study (Berger‐Höger 2019), and unclear in one study (Hamann 2006).

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4

Summary of findings 1. Comparison #1: Decision coaching versus usual care ‐ Summary of findings.

Outcomes	Relative effect (95% CI)	Number of participants (studies)	Quality of evidence (GRADE)	Comments
Preparation for decision‐making scale (Bennett 2010) (higher scores = feeling more prepared). Measured: Post‐consultation (Shepherd 2019)	The mean score for usual care was 74.0 (SD 20.1). The mean score was higher (better) for the intervention group: 90.2 (SD 12.6) (P < 0.0001).	110 (1 study)	⊕⊖⊖⊖[1] Very low	We are uncertain if decision coaching improves preparation for decision making, compared with usual care.
Decision self‐confidence: Decision self‐efficacy scale (O'Connor 1995): range 0 (no self‐efficacy) to 100 (perfect self‐efficacy). Measured: Post‐consultation (Hacking 2013; Shepherd 2019) (Analysis 1.1)	Mean scores ranged from 83 to 86 with usual care. The mean meta‐analysis score in the intervention groups was 5.2 points higher (better) than usual care (95% CI: 1.7 to 8.6).	201 (2 studies)	⊕⊖⊖⊖[2]  Very low	We are uncertain if decision coaching improves decision self‐confidence, compared to usual care.
Feeling uniformed	‐	‐	‐	Not measured
Unclear values	‐	‐	‐	Not measured
Feeling unsupported	‐	‐	‐	Not measured
Knowledge tests: Standardised on score from 0 (no knowledge) to 100 (perfect knowledge) Measured using different knowledge tests: post‐intervention (Green 2001) or one month (Matloff 2006) post‐intervention (Analysis 1.2)	Mean scores ranged from 63.3 to 74.0 with usual care. The mean meta‐analysis score in the intervention groups was 13.0 points higher (better) than usual care (95% CI: 6.2 to 19.8).	97 (2 studies)	⊕⊖⊖⊖[3] Very Low	We are uncertain if decision coaching improves knowledge, compared with usual care.
Adverse effects: Decision Regret Scale (Brehaut 2003): Standardised on score from 0 to 100 (none to worst regret) measured 3 (Shepherd 2019) and 6 months (Hacking 2013) post‐consultation	The mean scores for usual care groups were 19.0 (SD 22.9) (Shepherd 2019) and 17.1 (SD 16.0) (Hacking 2013). The mean scores were lower (better) in intervention group: 9.3 (SD 12.3) (P = 0.039) (Shepherd 2019) and 10.8 (SD 13.7) (P = 0.036) (Hacking 2013).	170 (2 studies)	⊕⊖⊖⊖[4]  Very low	We are uncertain if decision coaching improves (reduces) decision regret compared with usual care.
Hospital Anxiety and Depression (HAD) Scale (Zigmond 1983) Measured 3 months post‐consultation (Shepherd 2019) and 6 months (Hacking 2013) post‐intervention. Matloff 2006 measured worry about cancer 6 months post‐intervention (Diefenbach 1999) and therefore not included in the summary of findings for this outcome.	There was no difference between groups. For one study, data were not reported (Hacking 2013).	~175 (2 studies)	⊕⊖⊖⊖[5]  Very low	We are uncertain if decision coaching has an effect on anxiety compared with usual care.

[1] Evidence was downgraded by ‐1 level for methodological limitations (high risk of attrition bias), and ‐2 for imprecision as results were based on one small study.

[2] Evidence was downgraded ‐2 for methodological limitations (items rated as high risk of attrition bias in one study, unclear sequence generation in one study, mixed unclear and low risk of bias across other domains in the studies); ‐1 for imprecision as results were based on small studies and  overall sample size.

[3] Evidence was downgraded ‐2 for methodological limitations (items rated as unclear risk of bias including sequence generation for one study, and allocation concealment for both studies, with mixed unclear and low risk of bias across other domains); ‐1 for imprecision as results were based on small studies and overall sample size.  

[4] Evidence was downgraded ‐2 for methodological limitations (items rated as high risk of attrition bias, unclear sequence generation in one study, mixed unclear and low risk of bias across other domains in the studies); ‐1 for imprecision as results were based on small studies and overall sample size.

[5] Evidence was downgraded ‐2 for methodological limitations (items rated as high risk of attrition bias in one study, unclear sequence generation in two studies, allocation concealment in one study, mixed unclear and low risk of bias across other domains); ‐1 for imprecision as results were based on small studies and overall sample size.

Summary of findings 2. Comparison #2: Decision coaching versus evidence‐based information ‐ Summary of findings.

Outcomes	Relative effect (95% CI)	Number of participants (studies)	Quality of evidence (GRADE)	Comments
Preparation for decision‐making	‐	‐	‐	Not measured
Decision self‐confidence: Empowerment scale (McAllister 2011): Standardised on score from 0 to 100 (none to highest efficacy) measured 1 week post‐intervention (Adam 2019)	The mean score showed no difference between groups.	106 (1 study)	⊕⊖⊖⊖[1]  Very low	We are uncertain if decision coaching improves decision self‐confidence compared with evidence‐based information.
Feeling uninformed subscale of Decisional conflict Scale (O'Connor 1995b) measured post‐consultation (Deschamps 2004)	The mean score showed no difference between groups.	91 (1 study)	⊕⊖⊖⊖[2]  Very low	We are uncertain if decision coaching improves (reduces) feeling uninformed compared with evidence‐based information.
Unclear values	‐	‐	‐	Not measured
Feeling unsupported	‐	‐	‐	Not measured
Knowledge tests: Standardised on score from 0 to 100 (none to perfect) knowledge measured within one week post‐interventions (Analysis 2.1)	The mean scores across evidence‐based information groups were 64.2 to 96.0. The mean meta‐analysis score was no different between groups (SMD ‐0.23 95% CI: ‐0.50  to 0.04).	406 (3 studies)	⊕⊕⊖⊖[3] Low	Decision coaching may have little or no difference in improving knowledge compared with evidence‐based information.
Adverse effects: State and Trait Anxiety Inventory (Spielberger 1983): Standardised score from 0 to 100 (none to worst anxiety) measured post‐ intervention (Hunter 2005)	The mean score showed no difference between groups.	242 (1 study)	⊕⊕⊖⊖[4]  Low	Decision coaching may have little or no effect on anxiety compared with evidence‐based information.

 [1] Evidence was downgraded ‐2 for methodological limitations (6 of 7 items rated as unclear risk of bias including sequence generation and allocation concealment); ‐2 for imprecision as results were based on one small study.

[2] Evidence was downgraded ‐2 for methodological limitations (4 of 7 items rated as unclear risk of bias including sequence generation and allocation concealment); ‐2 for imprecision as results were based on one small study.

 [3] Evidence was downgraded ‐2 for methodological limitations (items rated as unclear risk of bias including sequence generation and allocation concealment for 2 of 3 studies).

 [4] Evidence was downgraded ‐2 for imprecision as results were based on one, small study.

Summary of findings 3. Comparison #3: Decision coaching plus evidence‐based information versus usual care ‐ Summary of findings.

Outcomes	Relative effect (95% CI)	Number of participants (studies)	Quality of evidence (GRADE)	Comments
Preparation for decision‐making scale (Bennett 2010): standardised scored from 0 to 100 (none to highest prepared) measured one month post‐intervention (Brown 2019)	The mean score showed no difference between groups.	37 (1 study)	⊕⊖⊖⊖[1]  Very low	We are uncertain if decision coaching plus evidence‐based information improves preparation for decision‐making compared with usual care.
Decision self‐confidence: Decision self‐efficacy scale (O'Connor 1995): range 0 to 100 (none to perfect self‐efficacy) measured 12 weeks post‐intervention (McBride 2016)	The mean score showed no difference between groups.	48 (1 study)	⊕⊖⊖⊖[2]  Very low	We are uncertain if decision coaching plus evidence‐based information improves decision self‐confidence compared with usual care.
Feeling uninformed subscale of Decisional Conflict Scale (O'Connor 1995b): standardised scored 0‐100 (lower score indicates feeling more informed) measured post‐intervention and post‐consultation (Analysis 3.1). A fourth study (Col 2007) measured change in feeling uninformed at 2 weeks post‐consultation and therefore not included in the summary of findings for this outcome. They reported a greater reduction (16.5 out of 100) for the decision coaching plus evidence‐based information compared with usual care (5.8 out of 100).	The mean scores across usual care groups were from 6.4 to 29.3. The mean difference was 5.83 lower (better) for the intervention group (95% CI: ‐8.9 to ‐2.8).	212 (3 studies)	⊕⊖⊖⊖[3] Very low	We are uncertain if decision coaching plus evidence‐based information improves (reduces) feeling uninformed compared with usual care.
Unclear values subscale of Decisional Conflict Scale (O'Connor 1995b):  standardised scored 0‐100 (lower scores indicate feeling clearer values) measured post‐intervention or post‐consultation (Analysis 3.2). A fourth study (Col 2007) measured change in unclear values at 2 weeks post‐consultation and therefore not included in the summary of findings for this outcome. They reported a greater reduction for the decision coaching plus evidence‐based information (7.3 out of 100) compared with usual care (‐0.25 out of 100).	The mean meta‐analysis score showed no difference between groups (SMD 0.02; 95% CI: ‐7.1 to 7.2).	212 (3 studies)	⊕⊖⊖⊖[4]  Very low	We are uncertain if decision coaching plus evidence‐based information  improves (reduces) unclear values compared with usual care.
Feeling unsupported subscale of Decisional Conflict Scale (O'Connor 1995b) standardised scored 0‐100 (lower score indicates feeling more supported) measured post‐intervention or post‐consultation (Analysis 3.3). A fourth study (Col 2007) measured change in feeling unsupported at 2 weeks post‐consultation and therefore not included in the summary of findings for this outcome. They reported a greater reduction for the decision coaching plus evidence‐based information (11.8 out of 100) compared with usual care (3.5 out of 100).	The mean meta‐analysis score showed no difference between groups (SMD ‐1.6; 95% CI: ‐5.4 to 2.2).	212 (3 studies)	⊕⊖⊖⊖[5] Very low	We are uncertain if decision coaching plus evidence‐based information  improves (reduces) feeling unsupported compared with usual care.
Knowledge tests: Standardised on score from 0 to 100 (none to perfect knowledge), using various knowledge tests. Measured within one month post‐intervention or post‐consultation (Analysis 3.4). For a further five studies not included in the meta‐analysis, two studies reported no differences, measured post‐intervention (Miller 2005a) and two weeks post‐consultation (Col 2007). Three studies (Bozic 2013; Hamann 2006; Sheridan 2012) reported the proportion of patients who scored higher on a knowledge test was better for decision coaching plus evidence‐based information compared with usual care (58.3% versus 33.3%; P = 0.01) (Bozic 2013); (15.0 versus 10.9; P = 0.01) (Hamann 2006); (47% versus 13%) (Sheridan 2012). Measures were post‐intervention (Bozic 2013), and post‐consultation (Sheridan 2012, Hamann 2006).	The mean scores across usual care groups was from 27.5 to 54.7. The mean meta‐analysis score in the intervention groups was 9.3 SD higher (better) (6.6 to 12.1 higher) than usual care.	1073 (5 studies)	⊕⊕⊖⊖[6] Low	Decision coaching plus evidence‐based information may improve knowledge compared with usual care.
Adverse effects:   Decision Regret Scale (Brehaut 2003) (Brown 2019; McBride 2016) Other DecisionRegret (Clark 1997), (Mishel 2009). Measured at up to 3 months post‐intervention	The mean scores showed no difference between groups (Brown 2019; McBride 2016). The mean score for other decision regret was 4.2 (SD 2.0) intervention group versus 3.9 (1.8) usual care group (P < 0.05) (Mishel 2009).	250 (3 studies)	⊕⊖⊖⊖[7]  Very Low	We are uncertain if decision coaching plus evidence‐based information improves (reduces) decision regret compared with usual care.
State and Trait Anxiety Inventory (Spielberger 1983) (Davison 1997; Kennedy 2002; Van Peperstraten 2010). Measured post‐intervention (Van Peperstraten 2010), 6 weeks post‐consultation (Davison 1997), and 12 months post‐consultation (Kennedy 2002).  Hospital Anxiety and Depression Scale (Zigmond 1983) (Lepore 2012). Measured: post‐intervention	The mean scores showed no difference between groups.	1017 (4 studies)	⊕⊖⊖⊖[8]  Very low	We are uncertain if decision coaching plus evidence‐based information has an effect on anxiety compared with usual care.

 [1] Evidence was downgraded ‐2 for methodological limitations (items rated as unclear risk of bias for allocation concealment, high risk of bias for blinding); ‐2 for imprecision as results were based on one small study.  

[2] Evidence was downgraded ‐1 for methodological limitations (items rated as high risk of bias for blinding, 2 of 7 items unclear risk of bias); ‐2 for imprecision as results were based on one small study.

[3] Evidence was downgraded ‐2 for methodological limitations (items rated as high risk of bias for blinding in two studies, unclear sequence generation in one study, unclear allocation concealment in one study); ‐1 for imprecision (small studies).

[4] Evidence was downgraded ‐2 for methodological limitations (items rated as high risk of bias for blinding in two studies, unclear sequence generation in one study, allocation concealment in one study); ‐1 for imprecision (small studies).

[5] Evidence was downgraded ‐2 for methodological limitations (items rated as high risk of bias for blinding in two studies, unclear sequence generation in one study, allocation concealment in one study); ‐1 for imprecision (small studies).

[6] Evidence was downgraded ‐1 for methodological limitations (items rated as unclear sequence generation in two studies, allocation concealment in four studies), and ‐1 due to inconsistency (heterogeneity; 2 of 5 studies that did not contribute to pooled results reported no effect). 

[7] Evidence was downgraded ‐1 for methodological limitations (items rated as unclear allocation concealment in one study), ‐1 for inconsistency (1 of 3 studies showed an effect in favour of the intervention), and ‐2 for imprecision as results were based on small studies and overall sample size.

[8] Evidence was downgraded ‐1 for methodological limitations (items rated as unclear risk of bias for allocation concealment in two studies), and ‐2 for imprecision (missing data for two studies, small sample for remaining two studies).

Summary of findings 4. Comparison #4: Decision coaching plus evidence‐based information versus evidence‐based information ‐ Summary of findings.

Outcomes	Relative effect (95% CI)	Number of participants (studies)	Quality of evidence (GRADE)	Comments
Preparation for decision‐making	‐	‐	‐	Not measured
Decision self‐confidence	‐	‐	‐	Not measured
Feeling uninformed subscale of Decisional Conflict Scale (O'Connor 1995b) standardised scored 0 to 100 (lower score indicate feeling more informed) change in score from baseline measured 2 weeks post‐consultation (Col 2007)	The mean score from baseline to post‐intervention showed no difference between groups.	67 (1 study)	⊕⊖⊖⊖[1] Very low	We are uncertain if decision coaching plus evidence‐based information improves (reduces) feeling uninformed compared with evidence‐based information.
Unclear values subscale of Decisional Conflict Scale (O'Connor 1995b) standardised scored 0 to 100 (lower scores indicate feeling more clear values) change in score from baseline measured post 2 weeks post‐consultation (Col 2007)	The mean score from baseline to post‐intervention showed no difference between groups.	67 (1 study)	⊕⊖⊖⊖[2]  Very low	We are uncertain if decision coaching plus evidence‐based information improves (reduces) unclear values compared with evidence‐based information.
Feeling unsupported subscale of Decisional Conflict Scale (O'Connor 1995b) standardised scored 0 to 100 (lower score indicates feeling more supported) change in score from baseline measured 2 weeks post‐consultation (Col 2007)	The mean score from baseline to post‐intervention showed no difference between groups.	67 (1 study)	⊕⊖⊖⊖[3]  Very low	We are uncertain if decision coaching plus evidence‐based information improves (reduces) feeling unsupported compared with evidence‐based information.
Knowledge tests: Standardised on score from 0 to 100 (none to perfect knowledge) measured within a month post‐intervention (Analysis 4.1). A fourth study (Col 2007)  measured change from baseline at two weeks post‐consultation and therefore not included in the summary of findings for this outcome. They reported no difference between groups. A fifth study, measured at two weeks post‐intervention (Kearing 2016), did not provide data on standard deviations and therefore not included in the summary of findings for this outcome. They reported no difference in knowledge for decision coaching plus evidence‐based information compared with evidence‐based information (65% versus 58%).	The mean scores across evidence‐based information groups were 44.0 to 70.4. The mean meta‐analysis score was no different between groups (SMD 0.18; CI: ‐0.20  to 0.56).	573 (3 studies)	⊕⊖⊖⊖[4] Very low	We are uncertain if decision coaching plus evidence‐based information improves knowledge compared with evidence‐based information.
Adverse effects: State and Trait Anxiety Inventory (Spielberger 1983): (Kennedy 2002). Hospital Anxiety and Depression Scale (Zigmond 1983) (Rahn 2018). Measured 2 weeks post‐intervention to 12 months post‐consultation	The mean score showed no difference between groups.	340 (2 studies)	⊕⊖⊖⊖[5]  Very low	We are uncertain if decision coaching plus evidence‐based information has an effect on anxiety compared with evidence‐based information.

[1] Evidence was downgraded ‐1 for methodological limitations (3 of 7 items rated as unclear risk of bias that include blinding, selective reporting, and other bias (recruitment prematurely ended due to publication of findings)); ‐2 for imprecision as results were based on a single small study.

[2] Evidence was downgraded ‐1 for methodological limitations (3 of 7 items rated as unclear risk of bias that include blinding, selective reporting, and other bias (recruitment prematurely ended due to publication of findings)); ‐2 for imprecision as results were based on a single small study.

[3] Evidence was downgraded ‐1 for methodological limitations (3 of 7 items rated as unclear risk of bias that include blinding, selective reporting, and other bias (recruitment prematurely ended due to publication of findings)); ‐2 for imprecision as results were based on a single small study.

[4] Evidence was downgraded ‐2 for methodological concerns (items for unclear risk of bias for allocation concealment for four studies and sequence generation for three studies); ‐1 for inconsistency (due to heterogeneity).

[5] Evidence was downgraded ‐2 for methodological concerns (items rated as unclear risk of bias including allocation concealment and sequence generation for one study); ‐1 for imprecision as results were based on small studies and overall sample size.

See Table 1, Table 2, Table 3, and Table 4. We reported on timing of the intervention in relation to post‐intervention (exposure to the intervention), or post‐consultation (post‐exposure to the intervention followed by consultation with healthcare provider).

Patients or public as research partners

None of the studies reported on a role for patients as research partners according to the GRIPP2 criteria: patients on research team, patient advisory team, training for patients, aim is patient and public involvement (PPI), methods to support PPI, results about PPI, discuss PPI, reflection on PPI (Table 10).

6. Patients as Research Partners: Guidance for Reporting Involvement of Patients and Public (GRIPP2) (N = 28 studies).

Author (year)	Patients on research or advisory team	Training for patients	Aim PPI	Methods PPI	Results PPI	Discuss PPI	Reflection PPI
Adam 2019	no	n/r	no	n/a	n/a	n/a	n/a
Aoki 2019	no	n/r	no	n/a	n/a	n/a	n/a
Berger‐Höger 2019*	no	n/r	no	n/a	n/a	n/a	n/a
Bozic 2013	no	n/r	no	n/a	n/a	n/a	n/a
Brown 2019	no	n/r	no	n/a	n/a	n/a	n/a
Col 2007	no	n/r	no	n/a	n/a	n/a	n/a
Davison 1997	no	n/r	no	n/a	n/a	n/a	n/a
Deschamps 2004	no	n/r	no	n/a	n/a	n/a	n/a
Green 2001	no	n/r	no	n/a	n/a	n/a	n/a
Hacking 2013	no	n/r	no	n/a	n/a	n/a	n/a
Hamann 2006*	no	n/r	no	n/a	n/a	n/a	n/a
Hunter 2005	no	n/r	no	n/a	n/a	n/a	n/a
Kearing 2016	no	n/r	no	n/a	n/a	n/a	n/a
Kennedy 2002	no	n/r	no	n/a	n/a	n/a	n/a
Langston 2010	no	n/r	no	n/a	n/a	n/a	n/a
Lepore 2012	no	n/r	no	n/a	n/a	n/a	n/a
Lerman 1997	no	n/r	no	n/a	n/a	n/a	n/a
Matloff 2006	no	n/r	no	n/a	n/a	n/a	n/a
McBride 2016	no	n/r	no	n/a	n/a	n/a	n/a
Miller 2005a	no	n/r	no	n/a	n/a	n/a	n/a
Mishel 2009	no	n/r	no	n/a	n/a	n/a	n/a
Myers 2005	no	n/r	no	n/a	n/a	n/a	n/a
Myers 2011	no	n/r	no	n/a	n/a	n/a	n/a
Rahn 2018	no	n/r	no	n/a	n/a	n/a	n/a
Shepherd 2019	no	n/r	no	n/a	n/a	n/a	n/a
Sheridan 2012	no	n/r	no	n/a	n/a	n/a	n/a
Van Peperstraten 2010	no	n/r	no	n/a	n/a	n/a	n/a
Vodermaier 2009	no	n/r	no	n/a	n/a	n/a	n/a

n/a =not applicable
n/r =not reported
PPI=Patient and Public Involvement

* Cluster‐RCT

1. Comparison #1: Decision coaching versus usual care

Four studies evaluated decision coaching versus usual care (Green 2001; Hacking 2013; Matloff 2006; Shepherd 2019) (see Table 1 and Table 11). 

7. Comparison #1: Decision coaching versus usual care (n = 4 studies) for outcomes not reported in meta‐analysis.

Study (year)	Outcome and Instrument	Timing	N coaching	Mean (SD) coaching	N usual care	Mean (SD) usual care	Notes
	1.1 PREPARATION FOR ACTIVE PARTICIPATION IN MAKING A HEALTH DECISION
Shepherd 2019	Preparation for decision making: 4 of 10 items scored 0 to 100. Preparation for Decision Making Scale ©Graham 1995 (Zigmond 1983)	Post‐1st consultation	60	90.21 (12.62)	50	74.0 (20.1)	p<0.0001
Hacking 2013	Decision self‐confidence; DSE scale ©O'Connor 1995	6 months post‐consultation	50	89.0 (9.91)	40	82.7 (16.6)	p=0.032
	1.2 RESOLVED DECISIONAL NEEDS
	1.3 ADVERSE EFFECTS
Shepherd 2019	Decision regret: 5 items scale of 1 to 5; scored 0 to 100; Decision Regret Scale ©O'Connor 1996 (Brehaut 2003)	3 months post‐ consultation	37	9.32 (12.26)	31	19.03 (22.86)	P=0.039
Hacking 2013	Decision regret; Decision Regret Scale ©O'Connor 1996 (Brehaut 2003)	6 months post‐consultation	58	10.8 (13.7)	44	17.1 (16.0)	p=0.036
Shepherd 2019	Anxiety; 7 items on 0‐3 scale; scored 0 to 21; Hospital Anxiety and Depression Scale (HADS) (Zigmond 1983)	3 months post‐ consultation	39	5.67 (4.13)	34	6.00 (4.46)	no difference
Hacking 2013	Anxiety; HADS (Zigmond 1983)	6 months post‐consultation	n/r	n/r	n/r	n/r	no difference
Matloff 2006	Worry about cancer; 1 item (1 not at all to 4 almost all the time); modified from Diefenbach 1999	6 months post‐intervention					No results provided for comparison
	1.4 SECONDARY OUTCOMES none reported

Abbreviations:
^{DSE=decision self‐efficacy
HADS=Hospital Anxiety and Depression Scale
n/r=not reported
SD=standard deviation}

 

Note: values are mean (SD) unless otherwise stated. 

*Values reported in the text were standardised out of 100

1.1 Preparation for active participation in making a health decision (patient‐reported)

1.1.1 Preparation for decision making

One study reported the mean Preparation for Decision Making score measured post‐consultation was higher in the decision coaching compared with usual care group (90.2 versus 74 out of 100; P < 0.0001) (Shepherd 2019). We are uncertain if decision coaching improves preparation for decision making compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (high risk of attrition bias), and ‐2 for imprecision as results were based on one small study.

1.1.2 Decision self‐confidence 

Meta‐analysis of two studies reported decision self‐confidence (using the decision self‐efficacy measure) was higher in the decision‐coaching group compared with usual care (SMD 5.16, 95% CI: 1.74, 8.58; 2 studies, 201 participants) (Analysis 1.1) (Hacking 2013; Shepherd 2019). Measures were post‐consultation. At six months’ post‐decision coaching, decision self‐confidence remained higher in the decision‐coaching group (89.0 versus 82.7 out of 100; P = 0.032) (Hacking 2013). We are uncertain if decision coaching improves decision self‐confidence compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as high risk of attrition bias in one study, unclear sequence generation in one study, mixed unclear and low risk of bias across other domains in the studies), and ‐1 for imprecision as results were based on small studies and overall sample size.

1.1. Analysis.

Comparison 1: Comparison #1: Decision coaching versus usual care, Outcome 1: Decision self‐confidence

1.1.3 Patient involvement in decision making

No studies reported this outcome.

1.1.4 Stage of decision making

No studies reported this outcome.

1.2 Resolution of modifiable decisional needs (patient‐reported)

1.2.1 Feeling uninformed

No studies reported this outcome.

1.2.2 Unclear values

No studies reported this outcome.

1.2.3 Feeling unsupported

No studies reported this outcome.

1.2.4 Knowledge

Meta‐analysis of two studies showed knowledge was higher in the decision‐coaching group compared with usual care (SMD 12.98, 95%CI: 6.21, 19.76; 2 studies, 97 participants) (Analysis 1.2) (Green 2001; Matloff 2006). Measures were reported immediately (Green 2001), and at one month post‐intervention (Matloff 2006). We are uncertain if decision coaching improves knowledge compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as unclear risk of bias including sequence generation for one study, and allocation concealment for both studies, with mixed unclear and low risk of bias across other domains), and ‐1 for imprecision as results were based on small studies and overall sample size.  

1.2. Analysis.

Comparison 1: Comparison #1: Decision coaching versus usual care, Outcome 2: Knowledge

1.3 Adverse effects (patient, observer‐reported)

1.3.1 Decision regret

Two studies reported the mean decision‐regret score was lower, indicating less decision regret, in the decision‐coaching group measured at three months post‐consultation compared with usual care (9.3 versus 19.0 out of 100; P = 0.039) (Shepherd 2019), and six months post‐consultation (10.8 versus 17.1 out of 100; P = 0.036) (Hacking 2013). We are uncertain if decision coaching improves (reduces) decision regret compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as high risk of attrition bias in one study, unclear sequence generation in one study, mixed unclear and low risk of bias across other domains in the studies), and ‐1 for imprecision as results were based on small studies and overall sample size.

1.3.2 Anxiety

One study measured anxiety at three months post‐consultation (Shepherd 2019), and two studies measured anxiety at six months post‐intervention (Hacking 2013; Matloff 2006). There was no difference in anxiety in the decision‐coaching group compared with usual care. For one study, data were not reported (Hacking 2013). We are uncertain if the decision has an effect on anxiety compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as high risk of attrition bias in one study, unclear sequence generation in two studies, allocation concealment in one study, mixed unclear and low risk of bias across other domains), and ‐1 for imprecision as results were based on small studies and overall sample size.

1.4 Secondary outcomes

1.4.1 Satisfaction with decision making

No studies reported this outcome.

1.4.2 Decision quality

No studies reported this outcome.

1.4.3 Health systems resources (length of consultation, cost to health system)

No studies reported this outcome.

1.4.4 Quality of life

No studies reported this outcome.

1.4.5 Quality of decision coaching (observer‐reported)

No studies reported this outcome.

 

2. Comparison #2: Decision coaching versus evidence‐based information

Four studies evaluated decision coaching versus evidence‐based information (Adam 2019; Deschamps 2004; Green 2001; Hunter 2005) (see Table 2 and Table 12). 

8. Comparison #2: Decision coaching versus evidence‐based information (n = 4 studies) for outcomes not reported in meta‐analysis.

Study (year)	Outcome and Instrument	Timing	N coaching	Mean (SD) coaching	N EBinfo	Mean (SD) EBinfo	Notes
	2.1 PREPARATION FOR ACTIVE PARTICIPATION IN MAKING A HEALTH DECISION
Adam 2019	Decision self‐confidence: Empowerment (range 24 low to 168 high); GCOS‐24 (McAllister 2011)	1 week post‐intervention	54	67.7 (11.6)*	52	62.4 (11.3)*	no difference
Deschamps 2004	Patient involvement in decision making: not MD controlled (n = proportion that did not indicate decision was mostly controlled by the MD); Control preferences scale (Degner 1997)	Post‐consultation	43	n = 41 (95.3%)	48	n = 47 (97.9%)
	2.2 RESOLVED DECISIONAL NEEDS
Deschamps 2004	Feeling uninformed: 16 items, unclear which scale, unable to standardise; Decisional Conflict Scale (DCS) uninformed subscale ©O'Connor 1995b	Post‐consultation	43	1.8 (SD n/r)	48	1.7 (SD n/r)	no difference
	2.3 ADVERSE EFFECTS
Hunter 2005	Anxiety: State anxiety (women): 20 items, range 20‐80; State and trait anxiety inventory (STAI) (Spielberger 1983)	Post‐intervention	126	47.98 (10.14)	116	45.50 (9.69)	no difference
	2.4 SECONDARY OUTCOMES
Deschamps 2004	Satisfaction with intervention: 10 items, 0‐4 scale (not at all helpful) to 4 (helped a great deal), total possible score of 40	Post‐consultation	42	27 (5.5)	48	28 (6.1)	no significant differences

Abbreviations:
^{DCS=Decisional Conflict Scale
EBinfo=evidenced‐based information
GCOS‐24=Genetic Counselling Outcome Scale 24‐item
MD=physician
n/r=not reported
SD=standard deviation
STAI=State and trait anxiety inventory}

Note: values are mean (SD) unless otherwise stated.

*Values reported in the text were standardised out of 100.

2.1 Preparation for active participation in making a health decision (patient‐reported)

2.1.1 Preparation for decision making

No studies reported this outcome.

2.1.2 Decision self‐confidence

One study reported decision self‐confidence (measured at one week post‐intervention with an empowerment measure) was no different for decision coaching compared with evidence‐based information (Adam 2019). We are uncertain if decision coaching improves decision self‐confidence compared with evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological limitations (six of seven items rated as unclear risk of bias including sequence generation and allocation concealment), and ‐2 for imprecision as results were based on one small study.

2.1.3 Patient involvement in decision making

One study measured patient perceived involvement with a healthcare provider post‐consultation and reported there was no difference for decision coaching compared with evidence‐based information (Deschamps 2004). We are uncertain if decision coaching increases patient involvement in decision making compared with evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological limitations (four of seven items rated as unclear risk of bias including sequence generation and allocation concealment), and ‐2 for imprecision as results were based on one small study.

2.1.4 Stage of decision making

No studies reported this outcome.

 

2.2 Resolution of modifiable decisional needs (patient‐reported)

2.2.1 Feeling uninformed

One study measured feeling informed immediately post‐consultation and reported the feeling uninformed subscale score was no different with decision coaching compared with evidence‐based information (Deschamps 2004). We are uncertain if decision coaching improves (reduces) feeling uninformed compared with evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological limitations (four of seven items rated as unclear risk of bias including sequence generation and allocation concealment), and ‐2 for imprecision as results were based on one small study.

2.2.2 Unclear values

No studies reported this outcome.

2.2.3 Feeling unsupported

No studies reported this outcome.

2.2.4 Knowledge

Meta‐analysis of three studies reported that knowledge was no different between decision coaching and evidence‐based information groups (SMD ‐0.23, 95% CI: ‐0.50 to 0.04; 3 studies, 406 participants) (Analysis 2.1) (Adam 2019; Green 2001; Hunter 2005). Measures were reported immediately (Green 2001; Hunter 2005), and at one week post‐intervention (Adam 2019). Decision coaching may have little or no difference on knowledge compared with evidence‐based information due to low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as unclear risk of bias including sequence generation and allocation concealment for two of three studies).

2.1. Analysis.

Comparison 2: Comparison #2: Decision coaching versus evidence‐based information, Outcome 1: Knowledge

 

2.3 Adverse effects (patient, observer reported)

2.3.1 Decision regret

No studies reported this outcome.

2.3.2 Anxiety

One study reported the mean anxiety score measured post‐intervention was no different with decision coaching compared with evidence‐based information (Hunter 2005). Decision coaching may have little or no effect on anxiety compared with evidence‐based information due to low certainty of evidence; downgraded ‐2 for imprecision as results were based on one small study.

 

2.4 Secondary outcomes

2.4.1 Satisfaction with decision making

One study reported the mean post‐consultation satisfaction score was no different with decision coaching compared with evidence‐based information (Deschamps 2004). We are uncertain if decision coaching improves satisfaction with decision making compared with evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological limitations (four of seven items rated as unclear risk of bias including sequence generation and allocation concealment), and ‐2 for imprecision as results were based on one small study.

2.4.2 Decision quality

No studies reported this outcome.

2.4.3 Health systems resources (e.g. length of consultation, cost to health system)

No studies reported this outcome.

2.4.4 Quality of life

No studies reported this outcome.

2.4.5 Quality of decision coaching (observer‐reported)

No studies reported this outcome.

 

3. Comparison #3: Decision coaching plus evidence‐based information versus usual care

Seventeen studies evaluated decision coaching plus evidence‐based information versus usual care (Aoki 2019; Berger‐Höger 2019; Bozic 2013; Brown 2019; Col 2007; Davison 1997; Hamann 2006; Kennedy 2002; Langston 2010; Lepore 2012; Lerman 1997; McBride 2016; Miller 2005a; Mishel 2009; Sheridan 2012; Van Peperstraten 2010; Vodermaier 2009) (see Table 3 and Table 13).

9. Comparison #3: Decision coaching plus evidence‐based information versus usual care (n = 17 studies) for outcomes not reported in meta‐analysis.

Study (year)	Outcome and Instrument	Timing	N coaching + EBinfo	Mean (SD) coaching + EBinfo	N usual care	Mean (SD) usual care	Notes
	3.1 PREPARATION FOR ACTIVE PARTICIPATION IN MAKING A HEALTH DECISION
Brown 2019	Preparation for decision making: 10 items scored 0 to 100. Preparation for Decision Making Scale ©Graham 1995	1 month post‐ intervention	16	83.44 (13.26)	21	77.74 (22.3)	no difference
McBride 2016	Decision self‐confidence: 11 items; scored 0 to 100; Decision Self‐Efficacy (DSE) scale ©O'Connor 1995	12 weeks post‐intervention	25	93.18 (7.1)	23	86.88 (16.69)	no difference (high baseline self‐efficacy)
Aoki 2019	Patient involvement in decision making: 10 items, 5‐point scale; Combined Outcome Measure for Risk Communication and Treatment Decision‐making Effectiveness COMRADE (Edwards 2003)	Post‐ intervention	32	88.0 median; 9 IQR	53	76.0 median; 7 IQR	P<0.001
Hamann 2006*	Patient‐perceived involvement in decision making (10 items, 5‐point scale, total 100); COMRADE (Edwards 2003)	Post‐ intervention Post‐ consultation	49 49	79.5 (18.6) 76.8 (20.9)	58 58	69.7 (20.0) 73.5 (19.3)	p=0.03 p=0.18 (after MD consult)
Berger‐Höger 2019*	Patient involvement in decision making: SDM behaviours: range 0 (fully disagree) to 4 (fully agree); Multifocal APProach to the sharing‘ IN Shared Decision‐Making (MAPPIN‐Q) (Kasper 2012)	Post‐ consultation	36	3.87 [95% CI 3.78 to 3.96] patient‐reported; 3.42 [95% CI 3.09 to 3.74] MD‐reported	28	3.82 [95% CI 3.68 to 3.96] patient reported; 3.44 [95% CI 3.04 to 3.83] MD reported
Davison 1997	Observed role in decision making	Post‐consultation					significantly more assumed active role than  control group
Sheridan 2012	Patient involvement in decision making: Reported SDM (n = proportion reporting shared decisions)	Post‐consultation	38	n = 28 (74%)	51	n=39 (76%)	no difference
Van Peperstraten 2010	Impact on decision‐making process	Post‐intervention					no difference between groups
Vodermaier 2009	Patients’ perception of decision making – 5‐point scale (n = proportion who reported shared decision making or patient made the decision versus physician‐made decision)	1 week post‐consultation	53	n = 39 (73.6%)	54	n=38 (70.4%)	no difference
Davison 1997	Preferred level of involvement in decision making (n = proportion with assumed role as active or collaborative versus passive); Control preferences scale (Degner 1997)	6 weeks post‐consultation	30	n = 27 (90%)	30	n=20 (66.7%)	p<0.001
Van Peperstraten 2010	Patient involvement in decision making: Empowerment (n = proportion of couples who wanted to decide for themselves on the number of embryos transferred); Self‐Efficacy Scale (Luszczynska 2005); 5‐point scale dichotomised as fully empowered (wanting to make the decision with the doctor only as adviser) and not fully empowered (wanting to make the decision with the doctor as a decision maker)	Post‐ intervention	127	n = 116 (91%)	135	n=99 (73%)	p<0.001
Sheridan 2012	Patient involvement in decision making: participated at preferred level (n = proportion reporting participation at preferred level)	Post‐consultation	38	n = 27 (71%)	51	n=39 (76%)
	3.2 RESOLVED DECISIONAL NEEDS
Col 2007	Uninformed: 1‐5; DCS uninformed subscale ©O'Connor 1995b	Baseline and 2 weeks post‐consultation	36	change of 16.5 (20.75)†	32	change of 5.75 (11.25)†	greater reduction compared to usual care
Col 2007	Unclear values: 1‐5; DCS unclear values subscale ©O'Connor 1995b	Baseline and 2 weeks post‐consultation	36	change of 7.25 (24.25)†	32	change of ‐0.25 (20.5)†	greater reduction compared to usual care
Col 2007	Unsupport: 1‐5; DCS unsupport subscale ©O'Connor 1995b	Baseline and 2‐week post‐consultation	36	change of 11.75 (18.0)†	32	Change of 3.5 (14.5)†	greater reduction compared to usual care
Bozic 2013	Knowledge: 19 items (n = proportion who scored > 50%); Decision quality instrument (Sepucha 2011)	Post‐1st consultation	60	n = 35  (58.3%)	60	n=20 (33.3%)	P=0.01
Col 2007	Knowledge:  7‐item multiple choice, change from baseline	Baseline and 2 weeks post‐consultation	36	+1.17 (1.65)/7	31	0.41 (1.21)/7	no difference
Hamann 2006*	Knowledge: 7‐item multiple choice (possible range of scores n/r)	Post‐consultation	49	15.0 (4.4)	58	10.9 (5.4)	p=0.01
Sheridan 2012	Knowledge: 4‐item true/false, n =  % men having 100% knowledge	Post‐intervention	58	n = 27 (47%)	70	n=9 (13%)
Miller 2005a	Knowledge: 17‐item true/false	Post‐intervention	nr		nr		no difference
	3.3 ADVERSE EFFECTS
Brown 2019	Decision regret: 5‐item scale of 1 to 5; scored 0 to 100; Decision Regret Scale ©O'Connor 1996 (Brehaut 2003)	1 and 3 months post‐intervention					no difference‐authors say likely measured too soon
McBride 2016	Decision regret: 0‐100; ©O'Connor 1996 (Brehaut 2003)	12 weeks post‐intervention	25	40.4 (10.3)	22	38.4 (17)	no difference
Mishel 2009	Decision Regret: 3 items only; Quality of life scale (Clark 1997)	3 months post‐intervention	93	3.93 (1.82)	74	4.17 (1.96)	p<0.05
Davison 1997	Anxiety: State anxiety, 20 items, range 20‐80; State and Trait Anxiety Inventory (STAI) (Spielberger 1983)	6 weeks post‐consultation	30	35.5 (down 9.0)	30	34.5 (down 2.5)	no difference
Lepore 2012	Anxiety: State anxiety, 7 items on a scale of 0‐3; scored 0‐21:  ≤ 7 = normal,  ≥ 11 = abnormal; Hospital Anxiety and Depression Scale (HADS) (Zigmond 1983)	Post‐intervention	215	2.02 (SE.147)	216	2.16 (SE.146)	no difference
Van Peperstraten 2010	Anxiety: 20‐80; STAI (Spielberger 1983; Van der Bij 2003)	Post‐intervention	n/r	36.4 [95% CI 34.7‐38.2]	n/r	34.7 [95% CI 33.3‐36.1]	no difference
Kennedy 2002	Anxiety; STAI – short form (Marteau 1992)	12 months post‐consultation	143	n/r	121	n/r	No difference
	3.4 SECONDARY OUTCOMES
Miller 2005a	Satisfaction with decision‐making process: Five items (5‐point Likert scales); instrument n/r	2 weeks post‐intervention		84.25 (21)†		87.75 (18.75)†	no differences between groups
Vodermaier 2009	Satisfaction with decision‐making process: Yes/No (n = proportion satisfied with the process); One item from scale by Man‐Son‐Hing 1999	1 week post‐intervention	47	n = 42 (89.4%)	54	n=50 (92.6%)	no difference
Col 2007	Satisfaction with decision‐making process: range 0‐100; instrument n/r	2 weeks post‐consultation	36	63.5% (15.90)	32	56.4% (19.2)	no difference
Kennedy 2002	Satisfaction with decision‐making process (n = proportion of participants); One item from scale by Single question: “How would you rate the opportunities you have been given to become involved in making decisions about your treatment”	2 years post‐consultation	199	excellent: n = 75 (37.7%) good: n = 69 (34.7%) fair: n = 31 (15.6%) poor: n = 24 (12.1%)	183	excellent n=44 (24.0%)  good n=78 (42.6%) fair n=46 (25.1%) poor n=15 (8.2%)	OR 1.49; 95% CI 1.11‐2.01; p=0.008 controls; no diff PtDA only
Berger‐Höger 2019*	Decision quality: Informed choice (%); Adequate risk knowledge and a positive attitude towards the chosen option were classified as making informed choice (Marteau 2001)	Post‐consultation	36	47.66 [95% CI 12.64 to 82.68]	28	0 [95% CI 0.00 to 0.00]	p = 0.016
Berger‐Höger 2019*	Length of consultation: duration of whole decision making process; minutes	Post‐consultation	36	58.1 (13.4)	28	24.3 (6.3)
Bozic 2013	Length of consultation: entire patient visit; minutes	Post‐consultation	61	53.5 (23.8)	62	51 (21.1)	p=0.38 no difference
Bozic 2013	Length of consultation: Surgeon visit; minutes	Post‐consultation	59	20.9 (6.8)	60	21 (7.2)	p=0.91 no difference
Hamann 2006*	Length of consultation: Psychiatrist visit; minutes	Post‐consultation	n/r	64	n/r	60	no difference
Vodermaier 2009	Length of consultation: MD consult; minutes	Post‐consultation	53	about  mean 15	54	about mean 15	no difference; time reported by # 10 min, 15 min, 25 min, 35 min
Van Peperstraten 2010	Costs	Post‐consultation					saved €169.75 per couple in intervention group
Kennedy 2002	Costs; mean costs	Post‐consultation		$1566		$2,751	mean difference $1184 95% CI $684‐2110 control
Brown 2019	Quality of life: Health‐related quality of life; SF‐36v2 (Ware 2007)	1 month, 3 months post‐intervention	15	n/r	21	n/r	No differences
Kennedy 2002	Quality of life: Role physical dimension; SF‐36 (Brazier 1992)	Baseline, 2 years post‐consultation	176	n/r	157	n/r	Favours coaching+PtDA versus usual care p=0.04
Kennedy 2002	Quality of life: SF‐36 (Brazier 1992) physical function social function role emotional mental health energy or vitality pain general health	Baseline, 2 years post‐consultation	176	n/r	157	n/r	No difference
Kennedy 2002	Quality of life: EuroQol‐5 dimensions (EQ‐5D) Health utility index (0‐1); 1 = the value of full health; EQ‐5D (Brooks 1996)	Baseline, 2 years post‐consultation	300	0.835	298	0.797	No difference
McBride 2016	Quality of life: Health‐related quality of life; EQ‐5D (Brooks 1996)	Baseline, 12 weeks post‐intervention	26	69.42 (21.28)	23	64.13 (23.92)	No difference
Mishel 2009	Quality of life: Perception of overall quality of life; Single item 1 to 10 scale	Baseline, 1 and, 3 months post‐intervention	93	7.24 (0.20) 7.61 (1.79) 7.57 (1.73)	74	7.18 (0.26) 7.33 (2.23) 7.72 (1.72)	No differences

Abbreviations:
^{CI=confidence interval
COMRADE=Combined Outcome Measure for Risk Communication and Treatment Decision‐making Effectiveness
DCS=Decisional conflict scale
DSE=Decision self‐efficacy
EBinfo=evidence‐based information
EQ‐5D=EuroQol‐5 
HADS=Hospital Anxiety and Depression Scale 
IQR=Interquartile range
MAPPIN‐Q=Multifocal APProach to the sharing‘ IN
MD=physician
n/r=not reported} 

^{PtDA=patient decision aid
SD=standard deviation
SDM=shared decision making
SE=standard error
SF‐36(v2)=36‐Item Short Form Health Survey (version 2)
STAI=State and Trait Anxiety Inventory}
 

Note: values are mean (SD) unless otherwise stated.

*Cluster‐RCT

^† Values reported in the text were standardised out of 100^†.

3.1 Preparation for active participation in making a health decision (patient‐reported)

3.1.1 Preparation for decision making

One study measured preparation for decision making and/or related elements at one month post‐intervention (Brown 2019). The mean preparation for the decision‐making score was no different with decision coaching plus evidence‐based information compared with usual care. We are uncertain if decision coaching plus evidence‐based information improves preparation for decision making compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as unclear risk of bias for allocation concealment, high risk of bias for blinding), and ‐2 for imprecision as results were based on one small study. 

3.1.2 Decision self‐confidence

One study reported that decision self‐confidence (using decision self‐efficacy measure) measured at 12 weeks post‐intervention was no different between the decision coaching plus evidence‐based information compared with usual care groups (McBride 2016). We are uncertain if decision coaching plus evidence‐based information improves decision self‐confidence compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (items rated as high risk of bias for blinding, two of seven items unclear risk of bias), and ‐2 for imprecision as results were based on one small study.

3.1.3 Patient involvement in decision making

Seven studies measured patient ‘perceived involvement’ in decision making. Two studies measured post‐intervention (Aoki 2019), and post‐consultation (Hamann 2006), using the same validated scale (Combined Outcome Measure for Risk Communication and Treatment Decision‐making Effectiveness (COMRADE) (Edwards 2003)). These studies reported a higher score for the decision coaching plus evidence‐based information group compared with usual care: 88.0 median (9 IQR) versus 76.0 median (7 IQR) (P < 0.001) (Aoki 2019); 79.5 versus 69.7% (P = 0.03) (Hamann 2006). One additional study, measuring post‐consultation and using another validated scale, reported no difference between groups (Berger‐Höger 2019) (Table 13). 

The remaining four studies reported results from observed (Davison 1997), or unvalidated measures of perceived involvement in decision making (Sheridan 2012; Van Peperstraten 2010; Vodermaier 2009). Davison 1997 reported that more men perceived themselves as taking an active role in decision making for the decision coaching plus evidence‐based information group. Three studies reported no differences between groups (Sheridan 2012; Van Peperstraten 2010; Vodermaier 2009). Measures were reported immediately post‐intervention (Van Peperstraten 2010), post‐consultation (Sheridan 2012), and at one (Vodermaier 2009), and six weeks post‐consultation (Davison 1997).

One study reported the mean preferred level of involvement was higher in the decision coaching plus evidence‐based information group compared with usual care (90% versus 66.7%; P < 0.001) (Davison 1997). Van Peperstraten 2010 reported that the proportion of patients indicating they wanted to decide for themselves was higher in the decision coaching plus evidence‐based information group (91% versus 73%; P < 0.001). Another study reported that the numbers of participants reporting they had participated at preferred level of involvement was lower for the decision coaching plus evidence‐based information (71%) compared to usual care (76%) (Sheridan 2012). 

We are uncertain if decision coaching plus evidence‐based information has any effect on patients’ perceived involvement, preferred level of involvement or participation at preferred level of involvement in decision making compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (items rated as high risk of ‘other’ bias in two studies and for blinding in one study, unclear sequence generation in two studies, allocation concealment in two studies), and ‐2 for inconsistency (differences in methodological approaches and effects of interventions across studies).

3.1.4 Stage of decision making

No studies reported on this outcome.

 

3.2 Resolution of modifiable decisional needs (patient‐reported)

3.2.1 Feeling uninformed

Meta‐analysis of three studies showed reduced feeling uninformed (Decisional Conflict subscale) in the decision coaching plus evidence‐based information group compared with usual care (SMD ‐5.83; 95% CI: ‐8.90 to ‐2.76; 3 studies, 212 participants) (see Analysis 3.1) (Berger‐Höger 2019; Brown 2019; Vodermaier 2009). Measures were post‐intervention (Brown 2019), one week post‐intervention (Vodermaier 2009), and post‐consultation (Berger‐Höger 2019). A fourth study, measured at two weeks post‐consultation, reported a greater reduction (16.5 out of 100) in feeling uninformed for the decision coaching plus evidence‐based information compared with usual care (5.8 out of 100) (Col 2007). We are uncertain if decision coaching plus evidence‐based information improves (reduces) feeling uninformed compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as high risk of bias for blinding in two studies, unclear sequence generation in one study, unclear allocation concealment in one study), and ‐1 for imprecision (small studies).

3.1. Analysis.

Comparison 3: Comparison #3: Decision coaching plus evidence‐based information versus usual care, Outcome 1: Uninformed subscale ‐ Decisional conflict

3.2.2 Unclear values

Meta‐analysis of three studies showed the unclear values (Decisional Conflict subscale) was no different for the decision coaching plus evidence‐based information compared with usual care (SMD 0.02, 95% CI ‐7.10 to 7.15; 3 studies, 212 participants) (see Analysis 3.2) (Berger‐Höger 2019; Brown 2019; Vodermaier 2009). Measures were post‐consultation (Berger‐Höger 2019), at one week (Vodermaier 2009), and one month post‐intervention (Brown 2019). A fourth study, measured at two weeks post‐consultation, reported a greater reduction for the decision coaching plus evidence‐based information group (7.3 out of 100) compared with usual care (‐0.25 out of 100) (Col 2007). We are uncertain if decision coaching plus evidence‐based information improves (reduces) unclear values compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as high risk of bias for blinding in two studies, unclear sequence generation in one study, allocation concealment in one study) and ‐1 for imprecision (small studies).

3.2. Analysis.

Comparison 3: Comparison #3: Decision coaching plus evidence‐based information versus usual care, Outcome 2: Unclear values ‐ Decisional conflict

3.2.3 Feeling unsupported

Meta‐analysis of three studies showed that feeling unsupported (Decisional Conflict subscale) was no different for decision coaching plus evidence‐based information compared with usual care (SMD ‐1.59, 95% CI ‐5.40 to 2.22; 3 studies, 212 participants) (see Analysis 3.3) (Berger‐Höger 2019; Brown 2019; Vodermaier 2009). Measures were post‐consultation (Berger‐Höger 2019), at one week (Vodermaier 2009), and one month post‐intervention (Brown 2019). A fourth study, measured at two weeks post‐consultation, reported a greater reduction for the decision coaching plus evidence‐based information group (11.75 out of 100) compared with usual care (3.5 out of 100) (Col 2007). We are uncertain if decision coaching plus evidence‐based information improves (reduces) feeling unsupported compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological limitations (items rated as high risk of bias for blinding in two studies, unclear sequence generation in one study, allocation concealment in one study), and‐1 for imprecision (small studies).

3.3. Analysis.

Comparison 3: Comparison #3: Decision coaching plus evidence‐based information versus usual care, Outcome 3: Feeling Unsupported ‐ Decisional conflict

3.2.4 Knowledge

Ten studies reported on knowledge using a range of measures. Meta‐analysis of five studies showed increased knowledge for decision coaching plus evidence‐based information compared with usual care (SMD 9.3, 95% 6.6 to 12.1; 5 studies, 1073 participants) (see Analysis 3.4) (Berger‐Höger 2019; Brown 2019; Lepore 2012; Lerman 1997; Van Peperstraten 2010). Measures were post‐intervention (Lepore 2012; Van Peperstraten 2010), post‐consultation (Berger‐Höger 2019), and one month post‐intervention (Brown 2019; Lerman 1997). Of these studies, one study (Lepore 2012) was an outlier, but there were no obvious reasons for this from the published report. For the remaining five studies not included in meta‐analysis, two studies reported no differences (Col 2007; Miller 2005a), measured post‐intervention (Miller 2005a) and two weeks post‐consultation (Col 2007). Three studies reported a higher proportion of patients who scored better on a knowledge test with decision coaching plus evidence‐based information compared with usual care (58.3% versus 33.3%; P = 0.01) (Bozic 2013); (15.0 versus 10.9; P = 0.01) (Hamann 2006) (47% versus 13%) (Sheridan 2012). Measures were post‐intervention (Sheridan 2012), and post‐consultation (Bozic 2013; Hamann 2006). Decision coaching plus evidence‐based information may improve knowledge compared with usual care due to low certainty of evidence; downgraded ‐1 for methodological limitations (items rated as unclear sequence generation in two studies, allocation concealment in four studies), and ‐1 due to inconsistency (heterogeneity; two of five studies that did not contribute to pooled results reported no effect).  

3.4. Analysis.

Comparison 3: Comparison #3: Decision coaching plus evidence‐based information versus usual care, Outcome 4: Knowledge

 

3.3. Adverse effects

3.3.1 Decision regret

Two studies reported no difference in the mean decision regret score for decision coaching plus evidence‐based information compared with usual care (Brown 2019; McBride 2016). Brown 2019 reported measures at one and three months, and McBride 2016 at 12 weeks post‐intervention. Another study (Mishel 2009), at three months post‐intervention, reported reduced decision regret for decision coaching plus evidence‐based information compared with usual care (3.9 versus 4.2; P < 0.05). We are uncertain if decision coaching plus evidence‐based information improves (reduces) decision regret compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (items rated as unclear allocation concealment in one study), ‐1 for inconsistency (one of three studies showed an effect in favour of the intervention), and ‐2 for imprecision as results were based on small studies and overall sample size. 

3.3.2 Anxiety

Four studies reported no difference in the mean anxiety scores for decision coaching plus evidence‐based information compared with usual care (Davison 1997; Kennedy 2002; Lepore 2012; Van Peperstraten 2010). Measures were post‐intervention (Lepore 2012; Van Peperstraten 2010), at six weeks (Davison 1997), and twelve months post‐consultation (Kennedy 2002). We are uncertain if decision coaching plus evidence‐based information has an effect on anxiety compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (items rated as unclear risk of bias for allocation concealment in two studies), and ‐2 for imprecision (missing data for two studies, small sample for remaining two studies).

 

3.4 Secondary outcomes

3.4.1 Patient‐reported satisfaction with decision coaching

Four studies reported no difference in satisfaction with the decision‐making process for the decision coaching plus evidence‐based information compared with usual care (Col 2007; Kennedy 2002; Miller 2005a; Vodermaier 2009). Measures were at one week (Vodermaier 2009), and two weeks post‐intervention (Miller 2005a); and two weeks (Col 2007), and two years post‐consultation (Kennedy 2002). We are uncertain if decision coaching plus evidence‐based information improves satisfaction compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (items rated at low and unclear risk of bias across all studies, plus unclear risk of bias for allocation concealment in one study), ‐1 for inconsistency (due to different instruments, timing), and ‐1 for imprecision (small studies).

3.4.2 Decision quality

One study measured post‐consultation reported that the number of patients making an informed choice was higher with decision coaching plus evidence‐based information (47.6%) compared with usual care (0%) (Berger‐Höger 2019). For quality decisions, participants require adequate risk knowledge and a positive attitude towards the chosen option. We are uncertain if decision coaching plus evidence‐based information improves decision quality compared with usual care due to very low certainty of evidence; downgraded ‐1 for methodological limitations (items rated as high risk of bias for blinding (detection bias)), and ‐2 for imprecision as results were based on a single small study.

3.4.3 Health systems resources (length of consultation, costs to health system)

Four studies reported on length of consultation for decision coaching plus evidence‐based information compared with usual care. For duration of the whole decision‐making visit, one study reported longer time ((58.1; SD 13.4 versus 24.3; SD 6.3) minutes) (Berger‐Höger 2019), and one study reported no difference between groups ((53.5; SD 23.8 versus 51.0; SD 21.1) minutes; P = 0.38) (Bozic 2013). For the healthcare provider visit, three studies reported no difference in time ((20.9; SD 6.8 versus 21.0; SD 7.2) minutes; P = 0.91) (Bozic 2013), (60.4 versus 60.0 minutes) (Hamann 2006), (15.0 versus 15.0 minutes) (Vodermaier 2009).

Two studies reported on costs to the health system for decision coaching plus evidence‐based information compared with usual care. One study reported limited data on overall savings per couple in the intervention group (saved €169.75) (Van Peperstraten 2010), and the other reporting mean difference between groups (mean difference $1184; 95% CI $684 to 2110) (Kennedy 2002). The timing of measures for length of consultation and costs to the health system were reported post‐consultation. We are uncertain if decision coaching plus evidence‐based information improves length of consultations or costs to the health system, compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological concerns (mix of high risk of bias for blinding and 'other' bias, unclear risk of bias for sequence generation for two studies and allocation concealment for one study), and ‐1 for imprecision (small studies).

3.4.4 Quality of life

Three studies reported no difference in quality of life for the decision coaching plus evidence‐based information group compared with usual care (Brown 2019; McBride 2016; Mishel 2009). Another study reported improved scores for the physical dimension of the SF‐36 tool in favour of decision coaching plus evidence‐based information (P = 0.04) (Kennedy 2002). But for more generic quality of life measures (SF‐36 overall, Euro‐QoL), there were no differences between groups. Measures were at twelve weeks (McBride 2016); one and three months (Mishel 2009); one and three months post‐intervention (Brown 2019); and at two years post‐consultation (Kennedy 2002). We are uncertain if decision coaching plus evidence‐based information has an effect on quality of life compared with usual care due to very low certainty of evidence; downgraded ‐2 for methodological concerns (items for unclear risk of bias for allocation concealment for two studies and sequence generation for one study), ‐1 for inconsistency (due to different instruments, timing), and ‐1 for imprecision (small numbers of participants).

3.4.5 Quality of decision coaching (observer‐reported)

No studies reported this outcome.

 

4. Comparison #4: Decision coaching plus evidence‐based information versus evidence‐based information

Seven studies evaluated decision coaching plus evidence‐based information compared with evidence‐based information (Col 2007; Kearing 2016; Kennedy 2002; Lerman 1997; Myers 2005; Myers 2011; Rahn 2018) (see Table 4 and Table 14).

10. Comparison #4: Decision coaching plus evidence‐based information versus evidence‐based information (n = 7 studies) for outcomes not reported in meta‐analysis.

Study (year)	Outcome and Instrument	Timing	N coaching + EBinfo	Mean (SD) coaching + EBinfo	N EBinfo	Mean (SD) EBinfo	Notes
	4.1 PREPARATION FOR ACTIVE PARTICIPATION IN MAKING A HEALTH DECISION
Rahn 2018	Patient involvement in decision making: SDM, scale range from 0 to 4 with 4 indicating the highest levels of shared decision‐making; MAPPIN’SDM (Kasper 2012)	n/r	33	87.5 (15.0)*	26	80.0 (10.0)*
Kearing 2016	Stage of decision making: % that already made choice; Single question: ‘How far along are you with this decision?’	2 weeks post‐intervention	82	74%	86	52%	p=0.003
	4.2 RESOLVED DECISIONAL NEEDS
Col 2007	Feeling uninformed: 3 items on a scale of 1‐5, scored 0‐100; Decisional Conflict Scale (DCS) uninformed subscale ©O'Connor 1995b	Baseline and 2‐week post‐consultation	36	Change of 16.5 (20.75)*	31	Change of 23.0 (23.0)*	no difference
Col 2007	Unclear values: 3 items on a scale of 1‐5, scored 0 to 100; DCS unclear values subscale ©O'Connor 1995b	Baseline and 2‐week post‐consultation	36	change of 7.25 (24.25)*	31	change of 18.25 (21.5)*	no difference
Col 2007	Feeling unsupported: 3 items on a scale of 1‐5, scored 0 to 100; DCS unsupport subscale ©O'Connor 1995b	Baseline and 2‐week post‐consultation	36	change of 11.75 (18.0)*	31	Change of 13.75 (14.5)*	no difference
Col 2007	Knowledge: 7 items multiple choice, change from baseline; instrument n/r	Baseline and 2 weeks post‐consultation	36	+1.17 (1.65)/7	31	+1.94 (1.71)/7	no difference
Kearing 2016	Knowledge: 5 multiple choice (unable to standardise); instrument n/r	2 weeks post‐ intervention	82	65% (SD n/r)	86	58% (SD n/r)	P<0.001 from baseline; no diff btw groups
	4.3 ADVERSE EFFECTS
Kennedy 2002	Anxiety; the State and Trait Anxiety Inventory (STAI) – short form (Marteau 1992)	12 months post‐consultation	143	n/r	124	n/r	No difference
Rahn 2018	Anxiety; Hospital Anxiety and Depression Scale (HADS) (Zigmond 1983)	Baseline, 2 weeks, 6 months	38	n/r	35	n/r	anxiety scores were inconspicuous
	4.4 SECONDARY OUTCOMES
Col 2007	Satisfaction with decision‐making process: range 0‐100; instrument n/r	2 weeks post‐consultation	36	63.5% (15.90)	31	63.7% (13.10)	no difference
Kennedy 2002	Satisfaction with decision‐making process (n = proportion of participants); Single question: “How would you rate the opportunities you have been given to become involved in making decisions about your treatment”	2 years post‐consultation	199	excellent n=75 (37.7%) good n=69 (34.7%) fair n=31 (15.6%) poor n=24 (12.1%)	186	excellent n=51 (27.4%) good n=85 (45.7%) fair n=33 (17.7%) poor n=17 (9.1%)	no difference
Kennedy 2002	Agreement between preference and treatment (n = proportion who stated a preference post‐consultation and underwent treatment)	12 months post‐consultation	186	n=118 (63.40%)	170	n=100 (58.80%)	p=0.372
Rahn 2018	Decision quality: Informed choice (%); Multi‐dimensional Measure of Informed Choice (MMIC) (Marteau 2001)	6 months post‐consultation	31	n=15 (48%)	20	n=6 (30%)	p=0.193
Kennedy 2002	Costs; mean costs	Post‐consultation		$1,566		$2,026	mean difference $461 95% CI $236‐696 PtDA only
Kennedy 2002	Quality of life: Role physical dimension; SF‐36 (Brazier 1992)	Baseline, 2 years post‐consultation	176	n/r	164	n/r	No difference
Kennedy 2002	Quality of life: SF‐36 (Brazier 1992) physical function social function role emotional mental health energy or vitality pain general health	Baseline, 2 years post‐consultation	176	n/r	164	n/r	No difference
Kennedy 2002	Quality of life: EuroQol‐5 dimensions (EQ‐5D) Health utility index (0‐1); 1 = the value of full health; EQ‐5D (Brooks 1996)	Baseline, 2 years post‐consultation	300	0.835	296	0.839	No difference
Rahn 2018	Quality of life: Disease‐specific quality of life	Baseline, post‐intervention, post‐consultation, 2 wks, 3, 6 months					High quality of life and stable

Abbreviations:
^{DCS=Decisional conflict scale
EQ‐5D=EuroQol‐5 
HADS=Hospital Anxiety and Depression Scale 
MAPPIN=Multifocal APProach to the sharing‘ IN
MMIC=Multi‐dimensional Measure of Informed Choice
n/r=not reported
SD=standard deviation}

^{SDM=shared decision making
SF‐36=36‐Item Short Form Health Survey 
STAI=State and Trait Anxiety Inventory}

Note: values are mean (SD) unless otherwise stated.

*Values reported in the text were standardised out of 100.

                                                                                                                                    

4.1 Preparation for active participation in making a health decision (patient‐reported)

4.1.1 Preparation for decision making

No studies reported this outcome.

4.1.2 Decision self‐confidence

No studies reported this outcome.

4.1.3 Patient involvement in decision making

In one study that did not report timing of the measure (Rahn 2018), patient‐reported perceived involvement was higher with decision coaching plus evidence‐based information compared with evidence‐based information (87.5% versus 80%). We are uncertain if decision coaching plus evidence‐based information increases levels of patient perceived involvement in decision making compared with evidence‐based information only due to very low certainty of evidence; downgraded ‐2 for methodological concerns (four of seven items rated as unclear risk of bias including allocation concealment and sequence generation), and ‐2 for imprecision as results were based on one small study and sample size.

4.1.4 Stage of decision making

One study, measured at two weeks post‐intervention, reported a higher proportion who indicated their stage of decision making as already having made a choice in the decision coaching plus evidence‐based information compared with evidence‐based information (74% versus 52%; P = 0.003) (Kearing 2016). We are uncertain if decision coaching plus evidence‐based information improves stage of decision making compared with evidence‐based information only due to very low certainty of evidence; downgraded ‐1 for methodological limitations (five of seven items rated as unclear risk of bias including allocation concealment), and ‐2 for imprecision as results were based on one small study and sample size.

 

4.2 Resolution of modifiable decisional needs (patient‐reported). 

4.2.1 Feeling uninformed

One study reported that the mean change in feeling uninformed (Decisional Conflict subscale) measured two weeks post‐consultation was no different between decision coaching plus evidence‐based information compared with evidence‐based information only (Col 2007). We are uncertain if decision coaching plus evidence‐based information improves levels of feeling uninformed compared with evidence‐based information due to very low certainty of evidence; downgraded ‐1 for methodological limitations (three of seven items rated as unclear risk of bias that include blinding, selective reporting, and other bias (recruitment prematurely ended due to publication of findings)), and ‐2 for imprecision as results were based on one small study.

4.2.2 Unclear values

One study reported that the mean change in unclear values (Decisional Conflict subscale) measured two weeks post‐consultation was no different between decision coaching plus evidence‐based information compared with evidence‐based information only (Col 2007). We are uncertain if decision coaching plus evidence‐based information improves (reduces) unclear values compared with evidence‐based information due to very low certainty of evidence; downgraded ‐1 for methodological limitations (three of seven items rated as unclear risk of bias that include blinding, selective reporting, and other bias (recruitment prematurely ended due to publication of findings)), and ‐2 for imprecision as results were based on one small study.

4.2.3 Feeling unsupported

One study reported that the mean change in feeling unsupported (Decisional Conflict subscale) measured two weeks post‐consultation was no different between decision coaching plus evidence‐based information compared with evidence‐based information only (Col 2007). We are uncertain if decision coaching plus evidence‐based information improves (reduces) feeling unsupported compared with evidence‐based information due to very low certainty of evidence; downgraded ‐1 for methodological limitations (three of seven items rated as unclear risk of bias that include blinding, selective reporting, and other bias (recruitment prematurely ended due to publication of findings)), and ‐2 for imprecision as results were based on one small study.

4.2.4 Knowledge

Five studies reported results using a range of measures (Col 2007; Kearing 2016; Lerman 1997; Myers 2011; Rahn 2018). Meta‐analysis of three studies showed knowledge was no different between decision coaching plus evidence‐based information compared to evidence‐based information groups (SMD 0.18, 95% CI ‐0.20 to 0.56; 3 studies, 573 participants) (see Analysis 4.1) (Lerman 1997; Myers 2011; Rahn 2018). Measures were at one week (Myers 2011), two weeks (Rahn 2018), and one month post‐intervention (Lerman 1997). Another study measured change in knowledge at two weeks post‐consultation and reported no difference between groups (Col 2007). A fifth study, measured at two weeks post‐intervention (Kearing 2016), reported no difference in knowledge for decision coaching plus evidence‐based information compared with evidence‐based information (65% versus 58%); however, they did not provide data on standard deviations and, therefore, were not included in the meta‐analysis. We are uncertain if decision coaching plus evidence‐based information improves knowledge compared with evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological concerns (items for unclear risk of bias for allocation concealment for four studies and sequence generation for three studies), and ‐1 for inconsistency (due to heterogeneity).

4.1. Analysis.

Comparison 4: Comparison #4: Decision coaching plus evidence‐based information versus evidence‐based information, Outcome 1: Knowledge

 

4.3 Adverse effects (patient, observer‐reported)

4.3.1 Decision regret

No studies reported this outcome.

4.3.2 Anxiety

Two studies reported that anxiety scores were no different between decision coaching plus evidence‐based information compared with evidence‐based information (Kennedy 2002; Rahn 2018). Measures were at two weeks and six months post‐intervention (Rahn 2018), and at 12 months post‐consultation (Kennedy 2002). We are uncertain if decision coaching plus evidence‐based information has an effect on anxiety compared with evidence‐based information due to low certainty of evidence; downgraded ‐2 for methodological concerns (items rated as unclear risk of bias including allocation concealment and sequence generation for one study), and ‐1 for imprecision as results were based on small studies and overall sample size.

 

4.4 Secondary outcomes

4.4.1 Patient‐reported satisfaction with decision coaching

Two studies reported no difference in the satisfaction with decision coaching scores between decision coaching plus evidence‐based information compared with evidence‐based information (Col 2007; Kennedy 2002). Measures were at two weeks (Col 2007), and two years post‐consultation (Kennedy 2002). We are uncertain if decision coaching plus evidence‐based information affects satisfaction compared to evidence‐based information due to very low certainty of evidence; downgraded ‐2 for inconsistency (due to different instruments, timing), and ‐1 for imprecision as results were based on small studies.

4.4.2 Decision quality

One study (Rahn 2018), measured at six months post‐intervention, reported that the proportion of patients who made an informed choice was no different for decision coaching plus evidence‐based information compared to evidence‐based information only. We are uncertain if decision coaching plus evidence‐based information has an effect on decision quality compared to evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological concerns (four of seven items rated as unclear risk of bias including allocation concealment and sequence generation), and ‐2 for imprecision as results were based on one small study.

4.4.3 Health systems resources (length of consultation, cost to health system)

One study (Kennedy 2002), measured at post‐consultation, reported lower costs to the health system (mean difference $461) for decision coaching plus evidence‐based information compared to evidence‐based information. We are uncertain if decision coaching plus evidence‐based information has an effect on costs to the health system compared to evidence‐based information due to very low certainty of evidence; downgraded ‐2 for inconsistency (due to no reporting of instrument, timing), and ‐2 for imprecision as results were based on one small study.

4.4.4 Quality of life

One study (Kennedy 2002), measured at two years post‐consultation, reported no difference in quality of life physical dimensions or general quality of life for decision coaching plus evidence‐based information compared with evidence‐based information only. One study measured disease‐specific quality of life post‐intervention and at two weeks and six6 months post‐consultation but did not report any data (Rahn 2018). We are uncertain if decision coaching plus evidence‐based information has an effect on quality of life compared to evidence‐based information due to very low certainty of evidence; downgraded ‐2 for methodological limitations (several items rated at unclear risk of bias including allocation concealment and sequence generation in one study), ‐1 for inconsistency (due to limited reporting on outcomes, timing), and ‐2 for imprecision as results were based on small studies.

4.4.5 Quality of decision coaching (observer reported)

No studies reported this outcome.

Discussion

Summary of main results

We identified 28 included studies using a range of decision‐coaching interventions alone or in combination with evidence‐based information (e.g. a patient decision aid). All intervention‐comparator groups reported some studies that measured the outcome of ‘knowledge’. There was variable reporting on the other six primary outcomes (e.g. preparation for decision making, decision self‐confidence, feeling uninformed, unclear values, feeling unsupported, and adverse effects (decision regret, anxiety)). 

For decision coaching compared to usual care (n = 4 studies), we are uncertain of the effects on primary outcomes due to either these outcomes not being measured or due to very low certainty of evidence (i.e. preparation for decision making, decision self‐confidence, knowledge, decision regret, anxiety). 

For decision coaching compared to evidence‐based information only (n = 4 studies), there is low certainty of evidence that participants exposed to decision coaching may have little or no changes in knowledge or anxiety. The other primary outcomes were either not measured by included studies, or we are uncertain of the effects due to very low certainty of evidence (i.e. decision self‐confidence, feeling uninformed).

For decision coaching plus evidence‐based information compared to usual care (n = 17 studies), there is low certainty‐evidence that participants may have improved knowledge with the intervention. The other primary outcomes were either not measured by included studies, or we are uncertain of the effects due to very low certainty of evidence (i.e. preparation for decision making, decision self‐confidence, feeling uninformed, unclear values, feeling unsupported, decision regret, anxiety).

For decision coaching plus evidence‐based information compared to evidence‐based information only (n = 7 studies), we are uncertain of the effects on primary outcomes due to either the outcomes not being measured or due to very low certainty‐evidence (i.e. feeling uninformed, unclear values, feeling unsupported, knowledge, anxiety).

Only knowledge and adverse effects (e.g. decision regret or anxiety) were measured consistently enough to consider some conclusions but, due to low certainty‐evidence, no strong conclusions were possible. Our results must be interpreted with caution, due to the small number of eligible studies within comparisons and generally small sample sizes, and the variation in the design and quality of these studies.

Overall completeness and applicability of evidence

Our international and interdisciplinary team is composed of knowledge users (e.g. patient partners, healthcare providers, educators) and researchers with experience evaluating and/or teaching decision coaching. Our team provided guidance and support throughout the review processes. The patient partner on our executive team participated on project governance and study decision making. We feel quite confident that the review is complete and knowledge users on our research team describe the evidence as applicable.

Our review was set up to be inclusive and to report on studies from a range of settings with a broad range of participants (e.g. patients, decision coaches), so that we could best identify evidence on decision coaching as an intervention. Given the extensiveness of our search strategy for electronic databases and handsearching, our team feels confident that we were able to identify the relevant published studies. We are also aware that there is the possibility that a relevant study was missed or not identified for inclusion in the review. There were challenges when the intervention appeared to meet the definition of decision coaching but there were too few details provided on the intervention in the published study. Hence, we added a third round of screening for included studies that focused on the decision‐coaching intervention. If there were not enough details in the published paper, we contacted the authors requesting additional information to determine whether studies were about decision coaching and to confirm if the study should have been included (see Characteristics of excluded studies). 

Our synthesis of the evidence showed a high degree of heterogeneity which is to be expected with the design and testing of a complex intervention such as decision coaching. Furthermore, heterogeneity was greater when there were different evaluation measures, such as knowledge tests, that had to be specific to the healthcare decision.

The comparison with the greatest number of studies was decision coaching with evidence‐based information compared with usual care. In these studies, the evidence‐based information component was most often a patient decision aid. Previous research has shown that, compared to usual care, people who have used a patient decision aid have improved knowledge, feel more informed, feel clearer about their values, and participate more in decision making (Stacey 2017b). As we can only report on the outcomes when exposed to the complete intervention, it was difficult for us to draw conclusions about decision coaching used alone as an intervention, rather than in conjunction with a decision aid. The other decision‐coaching intervention‐comparator groups had too few studies with small sample sizes to determine the effectiveness of decision coaching. It is unclear if decision coaching always needs to be paired with evidence‐based information or if decision coaching alone is an effective intervention.

Our findings about the decision coaching intervention reflect the challenges of determining the effective and ineffective components of complex interventions (Danko 2018). For example, there are a range of names and descriptions for decision coaching and the steps described for different decision‐coaching interventions are multifaceted. The reported contextual descriptions (who, how, where, when, how much) for the decision coaching interventions were limited and showed variability between studies (Table 8). As a result, it is not yet possible to determine the dose and intensity of decision coaching required to achieve positive outcomes. 

Decision coaching interventions have been evaluated with men and women having a range of ages (18 to 85 years) and for 13 different types of decisions. Interestingly, none of the included studies evaluated the use of decision coaching with children, or evaluated decision coaching provided by peer‐support workers. In addition, most of the studies overlooked whether decisions were congruent with values, and did not consider involvement in decision making. 

The included studies in our review were conducted in higher‐income countries with no studies set in lower‐or middle‐ income countries. Evaluation of decision coaching in higher‐income countries was not surprising given the costs associated with human resources required to deliver decision coaching. Interestingly, only one of the 28 studies conducted a cost‐analysis and determined that, for the group who received decision coaching plus the evidence‐based information, it was less costly than the two comparators of evidence‐based information only or usual care (Kennedy 2002). While one other study also reported on costs (Van Peperstraten 2010), a full cost‐analysis was not conducted. Further research is required to determine costs and their associate impacts within health systems (Scalia 2020; Trenaman 2014), and to determine the impacts of decision coaching from patient, organisational and societal perspectives.

Quality of the evidence

The GRADE appraisal of the certainty of evidence indicates low or very low certainty for all the outcomes. Common reasons for downgrading included small number of studies, small sample sizes, and inconsistencies in measurement.

Many of the individual studies, appraised using the Cochrane Risk of Bias tool, had items that commonly were rated as having ‘low’ or ‘unclear’ risk of bias (Figure 2; Figure 3). Six studies had items judged to be at ‘high’ risk of bias due to not blinding participants and personnel (McBride 2016; Brown 2019), not blinding the person conducting the outcome assessment (Berger‐Höger 2019), incomplete outcome data (Shepherd 2019), and other (e.g. only self‐reports used, disproportionate numbers randomised to each arm and loss to follow‐up, clustering was not accounted for in analysis) (Aoki 2019; Hamann 2006).

Potential biases in the review process

We identified no potential biases in the review process. According to AMSTAR II (Shea 2017), we met 16 of 16 review quality criteria. In our review, we placed boundaries on how decision coaching was defined as a concept (i.e. healthcare providers trained or using a protocol et cetera) because it is poorly defined in the literature. We also excluded group decision coaching as we could not distinguish the effects of peer support from the effects of decision coaching. These decisions may be made differently by other review teams and might influence what is included in other reviews of decision‐coaching interventions.

Agreements and disagreements with other studies or reviews

Our review is focussed on decision coaching as an intervention to prepare patients for participating in making healthcare decisions. Our findings for decision coaching with evidence‐based information compared to usual care (e.g. improved knowledge) are consistent with previous reviews evaluating decision coaching within trials evaluating patient decision aids (Rahn 2021; Stacey 2012); likely because 14 of the 17 included studies used a patient decision aid as the source of evidence‐based information. These findings are also consistent with the systematic review of 105 RCTs evaluating patient decision aids for preparing patients for making healthcare decisions (Stacey 2017b). The previous review on coaching (Alders 2017) indicated that all but one of the interventions used additional material such as question prompt lists, audiotape of the consultation, or an educational DVD booklet; however, the review did not identify the DVD booklet as also supporting patients’ preparation for communicating with their physician (Mishel 2009). This same study was included in our review and this DVD booklet met the description of a patient decision aid that is known to independently improve patients’ outcomes. 

Our review showed decision coaching with evidence‐based information may improve knowledge compared to usual care. When decision coaching was compared to evidence‐based information, there was little or no difference in knowledge. Previous reviews of decision coaching used with patient decision aids also reported improved knowledge (Rahn 2021; Stacey 2012; Stacey 2017b) and knowledge of options, benefits and harms is a key element required when patients are making healthcare decisions (Joseph‐Williams 2014; Makoul 2006).

The fact that decision coaching was frequently combined with evidence‐based information confirms that decision coaching can be used as part of a multifaceted intervention for some healthcare decisions. These multifaceted decision‐coaching interventions also provided patients with the information they could review on their own at home. Previous reviews of interventions to enhance SDM identified the need for interventions targeting the patient (e.g. patient decision aids) and interventions targeting the healthcare professional (e.g. training) (Légaré 2018). In our review, the interventions targeting healthcare professionals included training and/or protocols for decision coaching.

Decision coaching with or without evidence‐based information did not appear to worsen any adverse effects. In fact, anxiety measured using the State‐Trait inventory (Spielberger 1983) or the Hospital Anxiety and Depression Scale (Zigmond 1983) showed that there may be little or no difference between any of the four comparison groups using decision‐coaching interventions alone or in combination with evidence‐based information, although there is considerable uncertainty associated with these results. These findings are consistent with anxiety measured in studies of patient decision aids that also indicated no worsening of anxiety (Bekker 2003; Stacey 2017b). For decision coaching with or without evidence‐based information, we are uncertain if there is an effect on decision regret (very low certainty‐evidence of no difference). Our findings on regret are consistent with the other recent reviews (Rahn 2021; Stacey 2017b).

Authors' conclusions

Implications for practice.

Given the limited findings from our review, it is difficult to discuss implications for practice. For those providing non‐directive decision coaching to patients facing healthcare decisions, it is unlikely to cause adverse effects and may result in some benefits. At a minimum, healthcare providers can use decision coaching with evidence‐based information such as patient decision aids as supported in our review and consistent with strong evidence in the Cochrane review of patient decision aids (Stacey 2017b). The intent of our study was to determine the effectiveness of decision coaching with healthcare providers, and to be inclusive of healthcare professionals and health support workers  who may be situated in lower‐ or middle‐income countries (e.g. peer support workers, lay health workers) (Jull 2019; WHO 2010). However, the restriction to inclusion of RCTs only meant that we screened out other study designs that included health support workers (e.g. Simmons 2017). We will use the information to inform plans for future updates of our decision‐coaching review, to ensure that such gaps are addressed with adjustments to the selection criteria. Further research is required to determine the contributions of decision coaching; that is, what works for whom, and in what circumstances.

Implications for research.

Future studies need to have adequate sample sizes and more consistent reporting of the coaching intervention to properly evaluate the effectiveness of decision coaching as an intervention to improve preparation for decision making and resolving decisional needs without causing adverse effects for patients facing healthcare decisions for themselves or a family member. It would be beneficial if subsequent studies used more consistent outcome measures that could facilitate meta‐analysis. Our review highlights the need for further studies to evaluate the effectiveness of decision coaching. Further research needs to be done, to determine if decision coaching has a role as a stand‐alone intervention or may be more useful as part of the implementation of evidence‐based information such as patient decision aids. 

Given the complexity of the decision‐coaching intervention, future studies should include better descriptions of the intervention (e.g. logic models) and process measures to aid in understanding the function of decision coaching as an intervention (Thomas 2020). As previously mentioned, further research is required to determine costs and related impacts within healthcare systems (Scalia 2020; Trenaman 2014).

Another area identified in our review is the lack of patient partner engagement in the conduct of studies about decision coaching. Most of the included studies in our review (Table 5) were conducted in countries that are advancing patient‐oriented research including USA, Canada, and the UK; however, the studies in our review did not report on engagement with patients as partners (Staniszewska 2017). Finally, given the relational aspect of decision coaching, it was interesting to see none of the decision‐coaching sessions had been recorded and analysed to measure the decision‐coaching process. Future research that reports on the decision‐coaching encounters will benefit the field of patient‐oriented healthcare and support the development of decision coaching as a healthcare intervention.

What's new

Date	Event	Description
30 July 2019	Amended	Updated affiliations

History

Protocol first published: Issue 7, 2019

Notes

This paper is based on standard text and guidance provided by Cochrane Consumers and Communication (CCCG 2021).

Appendices

Appendix 1. CENTRAL search strategy

ID             Search

#1             MeSH descriptor: [Decision Making] this term only

#2             MeSH descriptor: [Decision Support Techniques] this term only

#3             (decision* or decid* or choice or choose or prefer*):ti

#4             ((informed or aid*) NEAR (choice* or decision*)):ti

#5             #1 OR #2 OR #3 OR #4

#6             MeSH descriptor: [Directive Counseling] this term only

#7             MeSH descriptor: [Counseling] this term only

#8             MeSH descriptor: [Health Education] this term only

#9             MeSH descriptor: [Patient Education as Topic] this term only

#10          MeSH descriptor: [Patient Participation] this term only

#11          MeSH descriptor: [Physician‐Patient Relations] this term only

#12          MeSH descriptor: [Referral and Consultation] this term only

#13          (assess* or coach* or guidance or counsel* or prepar*):ti

#14          #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13

#15          #5 AND #14

Appendix 2. MEDLINE search strategy

Ovid MEDLINE® In‐Process & Other Non‐Indexed Citations and Ovid MEDLINE® 1946 to Present:

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐

Jull_Medline_final
1. *Decision Making/ 2. decision support techniques/
3. (decision* or decid* or choice or choose or prefer*).ti.
4. (informed adj (choice* or decision*)).tw.
5. or/1‐4
6. Directive Counseling/ or *Counseling/
7. *Health Education/
8. Patient Education as Topic/
9. *Patient Participation/
10. *Physician‐Patient Relations/
11. "Referral and Consultation"/
12. (assess* or coach* or guidance or counsel* or prepar*).ti.
13. or/6‐12
14. and/5,13
15. randomized controlled trial.pt.
16. controlled clinical trial.pt.
17. randomized.ab.
18. placebo.ab.
19. drug therapy.fs.
20. randomly.ab.
21. trial.ab.
22. groups.ab.
23. or/15‐22
24. exp animals/ not humans.sh.
25. 23 not 24
26. and/14,25

Appendix 3. Embase search strategy

1. *decision making/        

2. exp decision support system/ 

3. (decision* or decid* or choice or choose or prefer*).ti.                  

4. ((aid* or informed) adj (choice* or decision*)).tw.          

5. or/1‐4                

6. directive counseling/ or *counseling/ 

7. *health education/      

8. patient education/      

9. *Patient Participation/               

10. *doctor patient relation/       

11. patient referral/          

12. "empowerment"/      

13. (assess* or coach* or guidance or counsel* or prepar* or knowledg* or power* or empower*).ti.

14. or/6‐13          

15. and/5,14       

16. randomized controlled trial/                 

17. controlled clinical trial/          

18. single blind procedure/ or double blind procedure/   

19. crossover procedure/                

20. random*.tw.                

21. placebo*.tw.                

22. ((singl* or doubl*) adj (blind* or mask*)).tw. 

23. (crossover or cross over or factorial* or latin square).tw.           

24. (assign* or allocat* or volunteer*).tw.               

25. or/16‐24       

26. 15 and 25     

Appendix 4. PsycInfo search strategy

1. random*.ti,ab,hw,id.                  

2. intervention.ti,ab,hw,id.          

3. trial.ti,ab,hw,id.           

4. placebo*.ti,ab,hw,id.                  

5. groups.ab.       

6. ((singl* or doubl* or trebl* or tripl*) and (blind* or mask*)).ti,ab,hw,id.                

7. (cross over or crossover).ti,ab,hw,id.   

8. latin square.ti,ab,hw,id.            

9. (assign* or allocat* or volunteer*).ti,ab,hw,id.                 

10. (control or controlled).ti,ab,hw,id.   

11. treatment effectiveness evaluation/  

12. mental health program evaluation/   

13. exp experimental design/       

14. "2100".md. 

15. or/1‐14          

16. animal.po.   

17. 15 not 16      

18. *Decision Making/    

19. decision support systems/     

20. (decision* or decid* or choice or choose or prefer*).ti.               

21. ((aid* or informed) adj (choice* or decision*)).tw.       

22. or/18‐21       

23. *Counseling/                

24. *Health Education/ or *"Educational Programs"/         

25. client education/       

26. exp Client Participation/        

27. Therapeutic Processes/            

28. Professional Referral/ or Professional Consultation/   

29. exp interpersonal control/    

30. empowerment/          

31. (assess* or coach* or guidance or counsel* or prepar* or knowledg* or power* or empower*).ti.

32. or/23‐31       

33. and/22,32    

34. 17 and 33     

Appendix 5. CINAHL search strategy

#	Query
S25	S5 AND S14 AND S24
S24	S15 or S16 or S17 or S18 or S19 or S20 or S21 or S22 or S23
S23	TI (singl* or doubl* or tripl* or trebl*) and TI (blind* or mask*)
S22	AB (singl* or doubl* or tripl* or trebl*) and AB (blind* or mask*)
S21	AB (random* or trial or placebo*) or TI (random* or trial or placebo*)
S20	MH Quantitative Studies
S19	MH Placebos
S18	MH Random Assignment
S17	MH Clinical Trials+
S16	PT Clinical Trial
S15	"randomi?ed controlled trial" or PT randomized controlled trial
S14	S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13
S13	TI (assess* or coach* or guidance or counsel* or prepar*)
S12	(MM "Referral and Consultation")
S11	(MM "Physician‐Patient Relations")
S10	(MM "Consumer Participation")
S9	MM patient education
S8	MM health education
S7	(MH "Counseling")
S6	"Directive Counseling"
S5	S1 OR S2 OR S3 OR S4
S4	TX ((informed OR aid*) (choice* or decision*))
S3	TI (decision* or decid* or choice or choose or prefer*)
S2	MM decision support techniques
S1	MM Decision Making

Appendix 6. Proquest search strategy

S7	S2 AND S5 AND noft(random*) Nursing & Allied Health Database
S6	S2 AND S5 Nursing & Allied Health Database
S5	ti((assess* OR coach* OR guidance OR counsel* OR prepar*)) OR mainsubject((health OR patient) education) OR mainsubject(patient participation) OR mainsubject(counseling) OR mainsubject(referral) OR mainsubject(consultation model) Nursing & Allied Health Database
S2	mainsubject.Exact("decision making" OR "decision making, computer‐assisted") OR noft(((informed OR aid*) adj (choice* OR decision*))) OR ti((decision* OR decid* OR choice OR choose OR prefer*)) Nursing & Allied Health Database

Appendix 7. Web of Science search strategy

STUDY author	TITLE
Adam	Assessing an Interactive Online Tool to Support Parents' Genomic Testing Decisions.
Rahn	Nurse‐led immunotreatment DEcision Coaching In people with Multiple Sclerosis
Berger‐Hoger	Informed shared decision‐making supported by decision coaches for women with ductal carcinoma in
Myers	Preparing African‐American men in community primary care practices to decide whether or not to have prostate cancer screening.
Hunter	A randomized trial comparing alternative approaches to prenatal diagnosis counseling in advanced maternal age patient
Kennedy	Effects of decision aids for menorrhagia on treatment choices, health outcomes
Green	Education about genetic testing for breast cancer susceptibility
Rothert	An educational intervention as decision support for menopausal women.
Lerman	Controlled trial of pretest education approaches to enhance informed decisio

Data and analyses

Comparison 1. Comparison #1: Decision coaching versus usual care.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Decision self‐confidence	2	201	Mean Difference (IV, Random, 95% CI)	5.16 [1.74, 8.58]
1.2 Knowledge	2	97	Mean Difference (IV, Random, 95% CI)	12.98 [6.21, 19.76]

Comparison 2. Comparison #2: Decision coaching versus evidence‐based information.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Knowledge	3	406	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.50, 0.04]

Comparison 3. Comparison #3: Decision coaching plus evidence‐based information versus usual care.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Uninformed subscale ‐ Decisional conflict	3	212	Mean Difference (IV, Random, 95% CI)	‐5.83 [‐8.90, ‐2.76]
3.2 Unclear values ‐ Decisional conflict	3	212	Mean Difference (IV, Random, 95% CI)	0.02 [‐7.10, 7.15]
3.3 Feeling Unsupported ‐ Decisional conflict	3	212	Mean Difference (IV, Random, 95% CI)	‐1.59 [‐5.40, 2.22]
3.4 Knowledge	5	1073	Std. Mean Difference (IV, Random, 95% CI)	9.33 [6.55, 12.10]

Comparison 4. Comparison #4: Decision coaching plus evidence‐based information versus evidence‐based information.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Knowledge	3	573	Std. Mean Difference (IV, Random, 95% CI)	0.18 [‐0.20, 0.56]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Adam 2019.

Study characteristics
Methods	RCT
Participants	54 (coaching) versus 52 (decision aid) parents of children with early‐onset (age ≤ 5 years) epilepsy of unknown cause considering genome‐wide sequencing
Interventions	Intervention: standard pre‐test genetic counselling Comparator: DECIDE (an interactive online educational tool and decision aid)
Outcomes	Primary: Knowledge  Secondary: Empowerment
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Parents were randomized as a couple to receive one of the two interventions"
Allocation concealment (selection bias)	Unclear risk	"Parents were randomized as a couple to receive one of the two interventions"
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No information
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	No information
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	106 of 121 completed, but unclear how many in which group
Selective reporting (reporting bias)	Unclear risk	No protocol/registration identified
Other bias	Low risk	Appeared to be free of other sources of bias

Aoki 2019.

Study characteristics
Methods	RCT
Participants	35 (coaching + decision aid) versus 53 (usual care) undergraduate and postgraduate students aged 20 years and older who visited the outpatient services for first‐time diagnosis of major depressive episode including depressive phase of bipolar disorder
Interventions	Intervention: 7‐day shared decision‐making programme with option presentation consultation, external deliberation with a decision‐aid booklet, decision coaching by a nurse, and decision‐making consultation Comparator: usual care with decision about treatment made in the initial consultation
Outcomes	Primary: patient‐perceived involvement in medical decisions (COMRADE) Secondary: satisfaction, consultation duration, sharing information with others, looking up information on options/treatments,  persistence with treatment, severity of depressive symptoms, medication adherence
Notes	No health consumers were described as being on the team; no funding received; research ethics board approval obtained
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Participants were randomly assigned to one of two arms, following the restricted randomization and minimization method of item 8 in CONSORT 2010 (Moher 2012)" COMMENT:   *Minimisation may be implemented without a random element, and this is considered to be equivalent to being random.
Allocation concealment (selection bias)	Low risk	The randomisation was conducted by a research assistant not directly involved in the study.
Blinding of participants and personnel (performance bias) ‐ Study Level	Low risk	Clinicians and nurses were not blinded because of the design of the study. Low risk because objective measures used
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	A research assistant blinded to group allocation collected data at baseline, after the decision‐making consultation, and at each visit during the 6‐month trial period.
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	Reasons for attrition clearly described (see Figure 1). Missing outcome data balanced across groups: Intervention 20/35 (57%), control 32/53 (60%). However, it was unclear how the high rate of missing data influenced the results.
Selective reporting (reporting bias)	Low risk	Registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN000009239) before the commencement of data collection. Outcomes reported were consistent with the study protocol.
Other bias	High risk	The numbers randomised to each arm were disproportionate (35 intervention; 53 control) and there was an extremely high rate of loss to follow‐up before the 6‐month follow‐up: intervention 20/35 (57%) and control 32/53 (60%).

Berger‐Höger 2019.

Study characteristics
Methods	Cluster‐RCT
Participants	37 (coaching + decision aid) versus 30 (usual care) German women, aged 18 years or older, with primary histologically confirmed ductal carcinoma in situ facing primary treatment decisions
Interventions	Intervention: Nurse‐led coaching of shared decision making. Patients were provided with the decision aid, at least one nurse‐led decision coaching session, and a final shared decision making physician encounter. Comparator: Standard care
Outcomes	Primary: Extent of informed shared decision making Secondary: patients and healthcare professionals perspectives of shared decision making, informed choice (knowledge, attitude, uptake), decisional conflict, duration of coaching sessions and physician encounters
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The statistician (BH) provided a computer‐generated allocation sequence. During study progress, allocation might have become predictable. Thus, we used a random permuted block design with block sizes of 4, 6 or 8 to randomize clusters."
Allocation concealment (selection bias)	Low risk	"The allocation was concealed. An independent external person prepared sealed opaque envelopes. After baseline assessment of the respective cluster and its professionals, two researchers (BBH, KL) opened the sealed opaque envelope and revealed the center’s allocation on site. Patients were recruited by the participating physicians (electronic supplementary material S2) and kept unaware of their allocation status. After the final physician encounter, they were asked to guess whether they had received standard care or the new counselling approach."
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	"[P]erson prepared sealed opaque envelopes"... "After baseline assessment of the respective cluster and its professionals, two researchers (BBH, KL) opened the sealed opaque envelope and revealed the center’s allocation on site"... "Patients were recruited by the participating physicians (electronic supplementary material S2) and kept unaware of their allocation status. After the final physician encounter, they were asked to guess whether they had received standard care or the new counselling approach." Participants were blinded so low risk of bias for that item. Unclear if personnel blinded
Blinding of outcome assessment (detection bias) ‐ Study Level	High risk	"The primary outcome was the extent of informed shared decision‐making assessed by the observer‐based instrument of the validated inventory Multifocal APProach to the sharing‘ IN Shared Decision‐Making (MAPPIN’SDM). It assesses the mutual shared decision‐making‐behavior of health professionals and patients based on video‐recordings." "Due to the structural inequality between intervention and control group, video raters could not be blinded."
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	For eight patients, missing values were imputed (5 patients with missing values in 1, 2 or 3 items, 3 patients with missing values in all 11 items).
Selective reporting (reporting bias)	Low risk	The study protocol was available and all of the study’s prespecified (primary and secondary) outcomes that were of interest in the review have been reported in the prespecified way.
Other bias	Low risk	Cluster analysis on an individual level was planned; however, there were unanticipated low cluster sizes that resulted in unstable intracluster correlation coefficient estimations. As a result, cluster analysis was used as this is more robust, given the limitations of their recruitment/clusters.  Selective recruitment of cluster participants: low risk of bias

Bozic 2013.

Study characteristics
Methods	RCT
Participants	95 (coaching + decision aid) versus 103 (usual care) patients with a primary diagnosis of osteoarthritis of the hip or knee seeing an orthopaedic surgeon for the first time for this problem, and no history of a lower‐extremity joint arthroplasty and considered medically appropriate for total joint arthroplasty
Interventions	Intervention: Shared decision making intervention. decision aid (DVD, booklet), communication aid, question listing with coach, taped consultation, surgeon's dictated note Comparator: Usual care. Map, directions, one‐page info on signs and symptoms, diagnosis and treatment options for hip and knee osteoarthritis
Outcomes	Primary: Informed decision (knowledge) Secondary: Treatment choice (uptake), patient and provider satisfaction with the process of using shared decision making tools,  length of consultation time
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“The randomization was blocked with use of random permuted blocks in groups of four, six, or eight to help ensure that the groups were balanced” (p. 1634).
Allocation concealment (selection bias)	Low risk	“Patients were randomized to either the intervention group or the control group with use of the sealed envelop method” (p. 1634).
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	“[S]urgeons were not blinded to the intervention” (p. 1635). Knowing the allocation could be due to greater investment in decision‐making. Insufficient information to make a judgement
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Outcomes were objectivelymeasured and not subject to interpretation.
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	62% (123/198) retention rate, therefore, high attrition rate ‐ however the attrition was balanced between groups.
Selective reporting (reporting bias)	Low risk	Protocol available
Other bias	Low risk	Appeared to be free of other sources of bias

Brown 2019.

Study characteristics
Methods	Pragmatic RCT
Participants	19 (coaching + decision aid) versus 22 (usual care) adults ≥ 70 years of age with advanced chronic kidney disease attending hospital‐based nephrology services considering renal replacement therapy
Interventions	Intervention: OPTIONS PtDA (booklet, audio, worksheet); coaching renal nurse trained in delivering the OPTIONS intervention 1 month later, 3 months later if no decision yet Comparator: Standard care
Outcomes	Primary: Decision regret, decisional conflict Secondary: Knowledge, quality of life, participants’ and nurses’ perceptions of the usefulness of the PtDA (preparation for decision making)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	...through a computer‐generated program using block randomisation
Allocation concealment (selection bias)	Unclear risk	Randomisation occurred once the eligibility of the participant was confirmed, consent provided and baseline data collected. Allocation of the participant to either intervention or standard care occurred through a computer‐generated program using block randomisation. Nurse not blinded
Blinding of participants and personnel (performance bias) ‐ Study Level	High risk	No blinding and self‐assessed outcomes
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	To minimise bias, the outcome research assistant and the lead researcher were blinded to group allocation.
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Few dropouts. All participants included
Selective reporting (reporting bias)	Low risk	Primary outcomes as reported in study registration (ACTRN 12614001090606)
Other bias	Low risk	Appeared to be free of other sources of bias

Col 2007.

Study characteristics
Methods	RCT
Participants	50 (coaching + decision aid) versus 45 (decision aid) versus 50 (usual care) perimenopausal and postmenopausal women ages 45 to 60 years considering menopausal treatment
Interventions	Intervention: Coaching plus decision aid Comparison: Decision aid only Control: usual care (pamphlet about hormone therapy and menopause)
Outcomes	Primary: decisional conflict Secondary: patient satisfaction with the decision making process, satisfaction with the decision made,  knowledge about hormone therapy
Notes	Funded by Robert Wood Johnson Foundation and Agency for Healthcare Research and Quality; research ethics board approval at all sites; no health consumers on the research team
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central allocation: "Participants were centrally and securely randomized at the time of enrolment into 1 of 3 arms using sealed envelopes."
Allocation concealment (selection bias)	Low risk	Central allocation: "Participants were centrally and securely randomized at the time of enrolment into 1 of 3 arms using sealed envelopes."
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No mention of blinding
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	No mention of blinding, however, the outcomes were objectively measured.
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	See Figure 3 for numbers lost to follow‐up. Missing data balanced across groups
Selective reporting (reporting bias)	Unclear risk	No study protocol available; not enough information to permit judgement
Other bias	Unclear risk	Recruitment was prematurely ended at 145 participants of the planned sample size of 240 in August 2002 when the Women's Health Initiative findings were published.

Davison 1997.

Study characteristics
Methods	RCT
Participants	30 (coaching + decision aid) versus 30 (usual care) men newly diagnosed with prostate cancer facing treatment decisions
Interventions	Intervention: Empowerment intervention (asked to think of information they need to assist with choosing treatment, reviewed list of questions to give them an idea of types of questions they may wish to ask, showed men where to find answers in written information package, gave final list of questions to men, encouraged to read info and use list of questions, encouraged to participate in decision‐making) Comparator: Usual care
Outcomes	Patients' preferences for control over treatment decision making ‐ preferred roles and assumed roles (preparation for decision making), anxiety and depression (adverse effects) Note: Outcomes not classified as primary or secondary
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“The group to which subjects were assigned was predetermined by a block randomization procedure. This ensured there were an equal number of subjects in both groups for each physician.” (p. 5, Data collection)
Allocation concealment (selection bias)	Unclear risk	Not mentioned; group assignment predetermined by block randomisation procedure (p. 5)
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No blinding; study did not report on how the results could be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	Unclear blinding and whether outcomes could be affected by unblinded assessor
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	No flow diagram; p. 12 explains why certain men did not listen to audiotape. All men approached by study investigator agreed to participate; only 1 man refused to complete the second set of questionnaires.
Selective reporting (reporting bias)	Unclear risk	Protocol not mentioned
Other bias	Low risk	Appeared to be free of other sources of bias

Deschamps 2004.

Study characteristics
Methods	RCT
Participants	67 (coaching) versus 61 (decision aid) peri‐ and postmenopausal females considering hormone replacement therapy
Interventions	Intervention: Pharmacist consultation (reviewed risks and benefits of hormone replacement therapy based on prescribing guidelines of Society of Obstetricians and Gynaecologists of Canada, charts and graphs to visually show population data, agreed on plan, noted in consult letter to the patient’s attending physician) Comparator: Decision aid
Outcomes	Perception of being informed, decisional conflict, satisfaction with the intervention, patient's role in decision making (preparation for decision making), uptake of the decision,  satisfaction with decision, adherence Note: Outcomes not classified as primary or secondary
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information provided
Allocation concealment (selection bias)	Unclear risk	No information provided
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Unclear blinding
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	P. 4 ‐ flow diagram; p. 3 reasons for attrition mentioned. Baseline characteristics included
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Low risk	Appeared to be free of other potential biases

Green 2001.

Study characteristics
Methods	RCT
Participants	29 (coaching) versus 29 (decision aid) versus 14 (usual care) women aged 18 years and older with a first‐degree relative diagnosed with breast cancer considering genetic testing for breast cancer susceptibility
Interventions	Intervention: Genetic counselling Comparator: Decision aid (education by an interactive computer) Control: Usual care
Outcomes	Knowledge, uptake Note: Outcomes not classified as primary or secondary
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“[B]lock randomization schedule to one of three groups in a 2:2:1 ratio” (p. 2)
Allocation concealment (selection bias)	Unclear risk	No information provided
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	“[G]enetic counsellor blinded to randomization until just prior to the session” (p. 2); unclear if participants were blinded
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured and not subjective for interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	“Values do not always add up to the number of participants due to missing data”; reasons not mentioned (p. 4). “Participants’ baseline knowledge was reflected in the control group’s answers”; participants balanced in study groups
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Low risk	Appeared to be free of other sources of bias

Hacking 2013.

Study characteristics
Methods	RCT
Participants	63 (coaching) versus 60 (usual care) patients who had just received a diagnosis of localised or early stage primary prostate cancer who had a decision to make regarding cancer management
Interventions	Intervention: Decision navigation. Intervention group patients met with their navigator by telephone or in person prior to their specialist treatment consultation. The aim of this meeting was to assist patients in identifying and framing key questions and concerns regarding cancer management options to generate a personal consultation plan for the appointment. The navigator produced a draft consultation plan and, after incorporating patient edits, gave copies to the physician in advance. Comparator: Usual care
Outcomes	Decision self‐efficacy; decision regret; decisional conflict; anxiety; depression
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"[R]andomised by computer" (the sequence generation process was not described)
Allocation concealment (selection bias)	Low risk	“The allocation sequence was concealed from the researcher enrolling participants. Patient details were entered into the web tool which allocated participants to one of two parallel groups, to receive decision navigation (intervention group) or usual care (control group). Patients’ allocation was revealed after consent and before baseline measures were completed.”
Blinding of participants and personnel (performance bias) ‐ Study Level	Low risk	“It was not possible to blind patients or physicians to patient group allocation, given that both patient and physician were actively involved in the intervention, and although patient report measures were collected by an independent researcher, again it was not possible to conceal the patient’s group”. Low because objective measures used
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	“Analysis was conducted by a researcher (JKA) blind to allocation and study procedures”. Low ‐ objective measures used
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	Loss to follow‐up was 3/63 (4.8%) in the intervention group and 7/60 (11.7%) in the control group.
Selective reporting (reporting bias)	Unclear risk	"The RCT protocol was approved by the National Health Service Lothian and the National Health Service Lothian ethics committee (Ref:08/F1102/45)." No access to the protocol provided
Other bias	Low risk	Appeared to be free of other potential sources of bias

Hamann 2006.

Study characteristics
Methods	RCT
Participants	54 (coaching + decision aid) versus 59 (usual care) men and women aged 18–65 years who had an International Classification of Diseases, Tenth Revision diagnosis of schizophrenia or schizophreniform disorder considering treatment options
Interventions	Intervention: Shared decision making to inform patients about their treatment options and prepare them for a planning talk with their physicians Comparator: Usual care
Outcomes	Severity of illness, doctor–patient relationship, rating patients' performance in the planning talk, physician satisfaction with treatment results, consultation time, uptake of decisions, patient's attitudes towards treatment, knowledge, patients' perceived involvement in medical decisions, satisfaction with care Note: Outcomes not classified as primary or secondary
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“[O]ne member of each pair being randomly assigned to the control or to the interventional condition” (p. 266). Sequence generation method was not stated
Allocation concealment (selection bias)	Unclear risk	No mention of allocation concealment
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No information provided
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	No information provided
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Reasons for attrition mentioned
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	High risk	Clustering was not accounted for in the analysis (high risk of bias). Selective recruitment of cluster participants: the participants in the study went to where there was an empty bed but with no explicit system in place (unclear risk of bias).

Hunter 2005.

Study characteristics
Methods	RCT
Participants	126 (coaching) versus 116 (decision aid) [110 group coaching participants excluded from this review] pregnant women aged ≥ 35 years with  gestational age ≤ 18 weeks considering prenatal diagnosis testing
Interventions	Intervention: Individual decision coaching (information about risks and limitations, laboratory results, and conditions being tested for, patient's values and beliefs about prenatal diagnosis, the birth of an abnormal child, and pregnancy termination) Comparator: Decision aid
Outcomes	Primary: knowledge, anxiety, decisional conflict Secondary: satisfaction with the intervention, uptake, pregnancy outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	p. 4 ‐ randomised in blocks of 30, 10 to each intervention group
Allocation concealment (selection bias)	Low risk	p. 4 ‐ “the allocations were provided in opaque envelopes”.
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Unclear blinding
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured and not subjective to interpretation.
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	Outcomes reported on fewer participants but no rationale provided for missing data
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Low risk	Appeared to be free of other potential biases

Kearing 2016.

Study characteristics
Methods	RCT
Participants	98 (coaching + decision aid) versus 101 (decision aid) patients with lumbar spinal stenosis considering treatment options
Interventions	Intervention: coaching plus a patient decision aid (video) Control: patient decision aid (video)
Outcomes	Primary: knowledge, decisional conflict Secondary: uptake of treatment choice, satisfaction with treatment, stage of decision making
Notes	No involvement of consumers on the research team; funding from the Informed Medical Decisions Foundation and the National Institute for Arthritis, Musculoskeletal and Skin Diseases
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"[P]articipants were randomly assigned using a computer generated permuted block randomization sequence".
Allocation concealment (selection bias)	Unclear risk	No further information
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Not reported
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	Not reported
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Balanced between groups
Selective reporting (reporting bias)	Unclear risk	No protocol/registration identified
Other bias	Unclear risk	Cross‐over study and unclear when outcomes were assessed

Kennedy 2002.

Study characteristics
Methods	RCT
Participants	300 (coaching + decision aid) versus 296 (decision aid) versus 298 (usual care) women referred from primary to secondary care with uncomplicated menorrhagia facing treatment options
Interventions	Intervention: Structured interview group plus PtDA (video, booklet)  Comparator: PtDA (video, booklet)  Control: A standard practice control group with no intervention
Outcomes	Primary: General health status, change in treatment preferences, uptake Secondary: Agreement between women’s preferences and treatments uptake, anxiety (short‐term follow‐up), health system costs, severity of condition, satisfaction, quality of life
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Allocation sequence was generated by computer and stratified by consultant and the age at which the woman left full‐time education (p. 3).
Allocation concealment (selection bias)	Low risk	“Secure randomization ensured by using a central telephone randomization system” (p. 3)
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Possibility of contamination bias; clinicians could have applied the experience gained from consultations with the intervention groups in their consultations with the control group (p. 6).
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear if blinding used but most outcomes were objectively measured and not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Table 1 and Figure 1 flow diagram (p. 4‐5)
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Low risk	Appeared to be free from other sources of bias

Langston 2010.

Study characteristics
Methods	RCT
Participants	114 (coaching + decision aid) versus 108 (usual care) women aged ≥ 18 years seeking a first trimester procedure for a spontaneous or induced abortion considering treatment to avoid pregnancy
Interventions	Intervention: Structured contraceptive counselling and PtDA (contraceptive flipchart and samples) Comparator: Usual care
Outcomes	Primary: Uptake (proportion of participants choosing a very effective contraceptive method) Secondary: Timing of initiation (immediate or delayed), adherence (continuation of the chosen option)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Using a random‐number table, we determined the sequence for 1:1 allocation constrained by blocks of 10” (p. 363, Methods ‐ study procedures)
Allocation concealment (selection bias)	Low risk	“Randomization assignments were sealed inside numbered, opaque envelopes” (p. 363, Methods ‐ study procedures)
Blinding of participants and personnel (performance bias) ‐ Study Level	Low risk	“No blinding of participants or coordinators was feasible due to the nature of the intervention. Physician‐providers did not know the participant’s allocation group, did not discuss the study with patients, and were asked not to change their counselling” (p. 363, Methods ‐ study procedures)
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured and not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	For “method initiation on the day of the procedure”, it was only said that the “[p]articipants in the intervention group were not more likely to initiate the requested method immediately compared to those in the usual care group”; possible that the results contradicted the hypothesis and were excluded for this reason
Selective reporting (reporting bias)	Unclear risk	No mention of study protocol; not enough information to permit judgement
Other bias	Low risk	Appeared to be free of other potential biases

Lepore 2012.

Study characteristics
Methods	RCT
Participants	244 (coaching + decision aid) versus 246 (usual care) men aged 45 to 70 years old of black African descent for prostate cancer screening
Interventions	Intervention: decision‐support intervention with PtDA (consisting of tailored telephone education about prostate cancer testing) Comparator: education on fruit and vegetable consumption
Outcomes	Primary: knowledge, decisional conflict, verified physician visit to discuss testing Secondary: uptake, anxiety
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“The principal investigator used a computer‐generated randomization schedule to randomize the participant.” (p. 322)
Allocation concealment (selection bias)	Unclear risk	“The principal investigator used a computer‐generated randomization schedule to randomize the participant and emailed the randomization assignment to the  interventionist.” (p. 322)
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Interventionists were not blind to condition. We can assume that patients were blinded as the study design was a telephone call for both intervention and control groups (p. 322).
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	“Data collectors were blind to condition but the interventionists were not” (p. 322).
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Did not appear to be missing any outcome data
Selective reporting (reporting bias)	Low risk	Appeared to have reported on all prespecified outcomes (protocol)
Other bias	Low risk	Appeared to be free of other potential sources of bias

Lerman 1997.

Study characteristics
Methods	RCT
Participants	122 (coaching + decision aid) versus 114 (decision aid) versus 164 (waiting‐list control) women at low to moderate risk who have a family history of breast and/or ovarian cancer considering BRCA1 testing
Interventions	Intervention: Counselling plus education Comparator: Education Control: waiting‐list control
Outcomes	Primary: knowledge, perceived risk Secondary: uptake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information provided
Allocation concealment (selection bias)	Unclear risk	No information provided
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Unclear blinding
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured and not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	Of 440 women, 400 completed 1‐month follow‐up interviews; no reasons provided; baseline data/characteristics included (p. 2)
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Low risk	Appeared to be free of other potential biases

Matloff 2006.

Study characteristics
Methods	RCT
Participants	32 (coaching) versus 32 (usual care) women with one first‐degree relative with breast cancer considering genetic testing
Interventions	Intervention: Genetic counselling and personalised risk assessment Comparator: Usual care
Outcomes	Knowledge, perceived risk, worry (about breast cancer, heart disease, and osteoporosis),  decisional conflict, uptake, satisfaction with the counseling intervention (intervention group only)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No mention of blinding
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	No mention of blinding
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Numbers lost to follow‐up provided in Figure 1. Missing data balanced across groups
Selective reporting (reporting bias)	Unclear risk	No mention of study protocol; not enough information to permit judgement
Other bias	Low risk	Appeared to be free of other sources of bias

McBride 2016.

Study characteristics
Methods	Pilot RCT
Participants	30 (coaching + decision aid) versus 26 (usual care) patients with a diabetic foot ulcer considering treatment options
Interventions	Intervention: Decision coaching (decision navigation) plus PtDA Comparator: Usual care
Outcomes	Primary outcomes: decisional self‐efficacy, adherence Secondary outcomes: severity of the condition, health‐related quality of life, decision conflict and decision regret.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Group allocation sequence was determined by an external researcher through the use of a computerised generated random number table (www.randomizer.org).
Allocation concealment (selection bias)	Low risk	To ensure allocation concealment from the research team and participants, opaque, sealed envelopes were employed which contained a note of group allocation. Envelopes were opened by a member of the research team after baseline measures had been taken.
Blinding of participants and personnel (performance bias) ‐ Study Level	High risk	Envelopes were opened by a member of the research team after baseline measures had been taken. Group allocation was then communicated to participants and the foot team. Because of the interactive nature of the intervention, it was not possible for group allocation to be blinded.
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	There was no mention that outcomes assessors were blinded. Outcomes were measured using self‐report (clinician and patient perspectives).
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Numbers lost to follow‐up provided in Figure 1. Missing data balanced across groups. Intention‐to‐treat analysis was also conducted and reported separately, whereby baseline (last) scores were carried forward and entered in place of missing data for repeated measures.
Selective reporting (reporting bias)	Unclear risk	No information about study registration or protocol publication given
Other bias	Low risk	Appeared to be free of other sources of bias

Miller 2005a.

Study characteristics
Methods	RCT
Participants	279 randomised to coaching plus PtDA versus usual care (numbers per group not reported) women over the age of 18 years and who expressed concerns about their risks for breast or ovarian cancer or requested information about risk assessment services or genetic testing during a self‐initiated call to the Cancer Information Service
Interventions	Intervention: Educational intervention on risk of inherited cancer and benefits/limitations of genetic testing Comparator: Usual care
Outcomes	Uptake, knowledge, perceived risk, actual risk, satisfaction with service, coping self‐efficacy Note: outcomes not classified as primary or secondary
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Low risk ("[R]andomized by the CATI system" (p. 4) after self‐initiated telephone contact)
Allocation concealment (selection bias)	Low risk	"[C]omputerized assisted telephone interview system (CATI)" (p. 4)
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Blinding was not addressed.
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured and not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	Reasons stated for initial dropout of study participants (p. 8). Patients contacted offered reasons for dropping out. Study protocol allowed patients to be reached up to 13 times at follow‐up; but some still not able to be reached
Selective reporting (reporting bias)	Unclear risk	No indication that the trial was registered in a central trials registry
Other bias	Low risk	Appeared to be free of other sources of bias

Mishel 2009.

Study characteristics
Methods	RCT
Participants	93 (coaching + written information + video) + 74 (control) men with prostate cancer staging (T1a, b, c or T2a, b), Gleason score less than 10, prostate‐specific antigen level less than 20, at least 10 days before the treatment consultation appointment. There was a second intervention group with 89 + a support person receiving the same intervention (results not reported in this review).
Interventions	Intervention: booklet on prostate cancer "Treatment Choices for Early Stage Prostate Cancer: Patients' Questions ‐ Doctors' Answers" from members of the Radiation Oncology Research Unit in Canada (before it was changed to be a patient decision aid), digital video disc presenting communication strategies, 4 decision‐coaching telephone calls by trained nurse   Control: handout on staying healthy during treatment
Outcomes	Primary: uncertainty management: problem‐solving subscale, patient‐provider communication subscale, prostate cancer knowledge scale Secondary: patients' perception of self‐communication, types of information used and its helpfulness, decisional regret, mood disturbances, quality of life
Notes	No patients were partners on the research team; trial registration not reported; funded by the NIH NINR; research ethics approval obtained; no reported conflicts of interest
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"The design for the study was a 3 x 2 randomized block, repeated measures design with three levels of the intervention: (treatment direct, treatment supplemented and control) and two levels of ethnicity (Caucasian and African‐American). Patients were randomly assigned to one of the treatment groups or to the control groups." No further information provided
Allocation concealment (selection bias)	Unclear risk	No information provided
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No information provided
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	No information provided
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	Numbers analysed or lost to follow‐up after baseline not reported.  No information provided, no flow chart
Selective reporting (reporting bias)	Unclear risk	No information about study registration or protocol publication given
Other bias	Low risk	Appeared to be free from other potential bias

Myers 2005.

Study characteristics
Methods	RCT
Participants	121 (coaching + decision aid) versus 121 (decision aid) African‐American men aged 40‐69 considering prostate cancer screening
Interventions	Intervention: Enhanced intervention group. Screening decision education session plus informational booklet Comparator: Standard intervention group (informational booklet)
Outcomes	Uptake of option (complete or partial screening)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information provided
Allocation concealment (selection bias)	Unclear risk	No information provided
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Unclear blinding
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes were objectively measured and not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	P. 4 ‐ flow diagram; reasons for not receiving intervention or completing endpoint chart mentioned. Baseline characteristics included
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Unclear risk	Baseline characteristics similar except the enhanced intervention group consisted of more men who were educated beyond high school and who were married

Myers 2011.

Study characteristics
Methods	RCT
Participants	156 (coaching plus decision aid) versus 157 (decision aid) males 50–69 years of age, had no history of prostate cancer or benign prostatic hyperplasia, and did not have a prostate‐specific antigen test in the previous 11 months considering prostate cancer screening
Interventions	Intervention: Enhanced intervention. Nurse‐led decision counselling session plus brochure on prostate cancer screening Comparator: Brochure on prostate cancer screening
Outcomes	Primary: knowledge, decisional conflict Secondary: completeness of informed decision‐making, uptake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how sequence was generated
Allocation concealment (selection bias)	Low risk	P. 241 (section 2.1): “Using a system of sealed envelopes”
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	No mention of blinding of either personnel or participants; all patient charts had a generic note placed in them by the nurse educator to prompt the physician to discuss prostate cancer screening.
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes assessed were not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	P. 243 (section 3.3): did not account for why 24 audio‐recordings were excluded
Selective reporting (reporting bias)	Unclear risk	No mention of study protocol; not enough information to permit judgement
Other bias	Low risk	Appeared to be free of other potential biases

Rahn 2018.

Study characteristics
Methods	Pilot RCT
Participants	38 (coaching + evidence‐based information) versus 35 (usual care) participants with multiple sclerosis aged 18 years or older with suspected or relapsing‐remitting multiple sclerosis facing immunotreatment decisions on first‐line drugs
Interventions	Intervention: Coaching plus evidence‐based information (up to three structured nurse‐led decision coaching sessions, access to an evidence‐based online information platform DEcision Coaching In people with Multiple Sclerosis (DECIMS‐Wiki)) Comparator: Usual care (access to the DECIMS‐Wiki, but received otherwise care‐as‐usual)
Outcomes	Primary: informed choice Secondary: decisional conflict, perceived involvement in shared decision making, values matched choice, coping self‐efficacy, trust in physicians and decision coaches, anxiety, depression, quality of life
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No comments
Allocation concealment (selection bias)	Unclear risk	We agreed it was hard to know how this was maintained: unclear
Blinding of participants and personnel (performance bias) ‐ Study Level	Low risk	Reviewers discussed comments and agreed on low risk of bias.
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	No comments
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	While the authors were transparent about missing data, we agreed that there was concern about how this significant attrition impacted the results in terms of bias. Given the transparency, however, we agreed on unclear risk of bias.
Selective reporting (reporting bias)	Low risk	No comments
Other bias	Low risk	Appeared to be free of other sources of bias

Shepherd 2019.

Study characteristics
Methods	RCT
Participants	68 (coaching) versus 69 (usual care) patients diagnosed with Stage II, III, and IV colorectal cancer considering oncology treatment (chemotherapy) following surgery
Interventions	Intervention: Consultation planning, recording and summarising Comparator: Usual care
Outcomes	Primary: Decision self‐efficacy, decisional conflict, decision regret, preparation for decision making Secondary: anxiety, depression
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Author SS used an online automated randomisation service to determine group allocation.
Allocation concealment (selection bias)	Low risk	Author SS used an online automated randomisation service to determine group allocation. Upon notification of group allocation, the intervention participants took part in a telephone “consultation planning” appointment before their clinic appointment. Usual care participants were informed, and subsequent contact was limited to answering questions about and delivery of questionnaires. Two 'navigators' delivered the intervention; they were research psychologists not involved in data collection or randomisation.
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Upon notification of group allocation, the intervention participants took part in a telephone “consultation planning” appointment before their clinic appointment. Usual care participants were informed, and subsequent contact was limited to answering questions about and delivery of questionnaires. One author conducted recruitment and (possibly) notified the participants (not clear that the author was part of the team directly involved in delivering the intervention). Participants would probably be unaware of assignment. Other team members delivered the intervention and would be aware of who was receiving the intervention.
Blinding of outcome assessment (detection bias) ‐ Study Level	Unclear risk	Not reported. Only those receiving the intervention were asked to complete the DSE. Unclear if or how blinding was managed
Incomplete outcome data (attrition bias) ‐ Study level	High risk	Reason for missing outcome data likely to be related to true outcome with imbalance in numbers across intervention groups for outcomes measured post‐first consultation (12% decision coaching versus 28% usual care), and an overall high rate of attrition over time (44.5%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Low risk	Appeared to be free of other sources of bias

Sheridan 2012.

Study characteristics
Methods	RCT
Participants	60 (coaching + video + information) versus 70 (control) men aged 40–80 years old with no prior history of prostate cancer considering prostate cancer screening
Interventions	Intervention: Coaching plus video based decision aid plus additional information on two more clearly beneficial men’s health screening services (cardiovascular disease screening and colon cancer screening) Comparator: Video on seat belt safety
Outcomes	Primary: perception that prostate cancer screening requires a decision, knowledge, patient role in decision making, participated at preferred role Secondary: uptake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation used computer‐generated random numbers that were sealed in opaque envelopes.
Allocation concealment (selection bias)	Low risk	Randomisation used computer‐generated random numbers that were sealed in opaque envelopes.
Blinding of participants and personnel (performance bias) ‐ Study Level	Low risk	"Men then proceeded to their visit with their physician (who was notified only about patients’ participation, but not group assignment)." Although described as single‐masked in the study registration, due to the attention control intervention in the control groups, both participants and staff could be considered blinded.
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Self‐reported outcome measures, therefore, low risk considering masking of participants.
Incomplete outcome data (attrition bias) ‐ Study level	Low risk	See Figure 1 for reporting on missing data (only 2 participants in intervention group due to false inclusion)
Selective reporting (reporting bias)	Low risk	Applied outcome measures as described in the study registration (clinicaltrials.gov/ct2/show/NCT00630188?term=00630188&draw=2&rank=1)
Other bias	Unclear risk	The small size of our study resulted in differential distribution of confounders among study groups.

Van Peperstraten 2010.

Study characteristics
Methods	RCT
Participants	152 (coaching + decision aid) versus 156 (usual care) couples on the waiting list (women < 40 years old) for a first in vitro fertilisation cycle ever or a first cycle after previous successful in vitro fertilisation deciding how many embryos should be transferred
Interventions	Intervention: multifaceted empowerment strategy Comparator: standard in vitro fertilisation care
Outcomes	Primary: uptake Secondary: empowerment, decision evaluation (satisfaction/uncertainty, informed choice, and decision control), knowledge, anxiety, depression, costs of implementing the intervention
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated list (p. 2, Methods section)
Allocation concealment (selection bias)	Low risk	Central allocation (p. 2, Methods section)
Blinding of participants and personnel (performance bias) ‐ Study Level	Low risk	“Because of the nature of the intervention it was not possible to blind the participants or in vitro fertilisation doctors to the allocation. Participation in our trial did not change the normal in vitro routine.” (p. 2, Methods section)
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Unclear blinding but outcomes assessed were not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	There were categories in each column of table 1 (p. 3) where the denominators did not match the number of people in the group and no reason was given to explain why this would be or if this affected the study.
Selective reporting (reporting bias)	Low risk	Outcomes same as those registered with ClinicalTrials.gov
Other bias	Low risk	The study appeared to be free of other sources of bias

Vodermaier 2009.

Study characteristics
Methods	RCT
Participants	74 (coaching + decision aid) versus 78 (usual care) patients aged 18 and 75 years with a histologically proven breast cancer or strong suspicion through diagnostic procedures considering treatment decisions
Interventions	Intervention: Decision board coaching (an informational and decisional intervention and a brochure in addition to standard care) Comparator: Standard care
Outcomes	Primary: decisional conflict Secondary: uptake, length of consultation, time point of decision making, patient perception of role in treatment decision making, perceived involvement in care, satisfaction
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“Randomisation after the patient gave written informed consent”; “Random assignment was performed by means of numbered cards in envelopes”; “stratified by age group” (p. 2)
Allocation concealment (selection bias)	Low risk	“[N]umbered cards in envelopes” (p. 2)
Blinding of participants and personnel (performance bias) ‐ Study Level	Unclear risk	Not blinded ‐ unclear if this would introduce bias to outcome assessed
Blinding of outcome assessment (detection bias) ‐ Study Level	Low risk	Not blinded but outcomes were objectively measured and not subjective to interpretation
Incomplete outcome data (attrition bias) ‐ Study level	Unclear risk	Flow diagram, p. 5; baseline characteristics not included
Selective reporting (reporting bias)	Unclear risk	No information provided
Other bias	Low risk	Appeared to be free of other potential biases

^{BRCA1: Breast cancer type 1 susceptibility protein
COMRADE: Combined Outcome Measure for Risk Communication and Treatment Decision‐making Effectiveness
DECIMS: DEcision Coaching In people with Multiple
Sclerosis
DSE: Decision Self‐Efficacy
MAPPIN'SDM: Multifocal APProach to the sharing‘ IN Shared Decision‐Making
PtDA: Patient decision aid}
 

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Aboumatar 2013	Not decision coaching
Adanikin 2013	Physician counselling
Adarkwah 2017	Conference abstract
Adekpedjou 2020	Not decision coaching
Akman 2010	Directive counselling
Aktan‐Collan 2000	Wrong study design
Alegria 2018	Not decision coaching
Allen 2018	Wrong intervention
AlSagheir 2020	Not decision coaching
Andaroon 2020	Not decision coaching
Arimori 2006	Decision coaching in both groups
Athens 2017	Wrong study design
Augstein 2007	Not decision coaching
Bailey 2021	Not decision coaching
Barnieh 2011	Not decision coaching
Barry 1997	Not decision coaching
Basu 2015	Conference abstract
Becker 2009	Not decision coaching
Belkora 2012	Decision coaching in both groups
Belkora 2015	Not an RCT
Benoit 2020	Conference abstract
Bergeron 2019	Not decision coaching
Berry 2013	Not decision coaching
Berry 2018	Not decision coaching
Biesecker 2018	Not decision coaching
Bjorklund 2012	Not decision coaching
Boulware 2018	Not decision coaching
Bowen 2002	Decision coaching in both groups
Bowen 2017	Not decision coaching
Briggs 2004	Advanced care planning
Brown 1999	No specific decision ‐ just prep for consult
Buhse 2015	Lifestyle choice
Buhse 2017	Conference abstract
Buhse 2018	Decision coaching delivered as a group
Calzone 2005	Decision coaching in both groups
Carlson 2021	Not decision coaching
Causarano 2015	Decision coaching in both groups
Cervantes 2007	Wrong study design
Cha 2018	Conference abstract
Chambers 2015	Counselling couples not decision prep
Chan 2011	Not decision coaching
Chan 2018	Advanced care planning
Chaney 2011	Conference abstract
Charles 2006	Decision coaching in both groups
Chhatriwalla 2019	Wrong study design
Choi 2014	Review article
Clark 2003	Physician counselling
Clarke‐Pounder 2015	Not decision coaching
Cohan 2009	Not decision coaching
Cohen 2004	Physician counselling
Cooper 2014	Advanced care planning
Cooper 2018	Not decision coaching
Costanza 2011	Wrong study design
Coulter 2007	Wrong study design
Cuypers 2018	Not decision coaching
Cuypers 2019	Not decision coaching
Damir 2011	Conference abstract
Davison 1997a	Not decision coaching
Davison 2003	Not an RCT
Davison 2007	Decision coaching in both groups
Deen 2012	Not decision coaching
Dehlendorf 2019	Not decision coaching ‐ rather SDM with provider
Deinzer 2006	Wrong study design
Deinzer 2009	Physician counselling
Di 2017	Not decision coaching
Dobke 2008	Not decision coaching
Dobscha 2006	Not decision coaching
Duncan 2010	Wrong study design
Duru 2017	Conference abstract
Dwamena 2012	Wrong patient population
Edbrooke‐Childs 2019	Not decision coaching
Edwards 2004	Physician counselling
El‐Jawahri 2015	Not decision coaching
Epstein 2018	Not decision coaching
Fan 2018	Genetic counselling not described
Farmer 2012	Not decision coaching
Ferguson 2017	Wrong study design
Finucane 1988	Not decision coaching
Fischer 2006	Not an RCT
Fischl 2010	Not decision coaching
Fitzpatrick 2016	Not decision coaching
Fooladi 2018	Advanced care planning
Foy 2019	Letter to the editor
Frijling 2003	Not decision coaching
Frosch 2001	Not decision coaching
Fujimoto 2017	Not decision coaching
Gagne 2017	Decision coaching in both groups
Gammon 2018	Not decision coaching
Garbers 2012	Not decision coaching
Ghasemi 2018	Not decision coaching
Giordano 2012	Physician counselling
Green 1997	Not an RCT. The authors described their experience with developing an interactive CD‐ROM on gene testing and breast cancer, and responded to anticipated criticisms of this technology.
Green 2004	Decision coaching in both groups
Green 2005	Decision coaching in both groups
Green 2015	Not decision coaching; advanced care planning
Green 2018	Not decision coaching; advanced care planning
Green 2020	Not decision coaching; advanced care planning
Greenfield 1985	Not decision coaching
Greenfield 1988	Not for a specific decision – it was in preparation for any decision
Guillen 2019	Decision coaching in both groups
Hamann 2011	Not decision coaching
Hamann 2013	Not decision coaching
Hamann 2020	Patients were provided with group training in SDM
Harmsen 2013	Not decision coaching
Harter 2016	Not decision coaching
Heisler 2014	Decision coaching in both groups
Helden 2012	Conference abstract
Helmes 2006	Not decision coaching
Henteleff 2006	Not decision coaching
Hess 2011a	Conference abstract
Hess 2011b	Conference abstract
Hippman 2016	Not decision coaching
Holt 2009	Decision coaching in both groups
Hooker 2011	Wrong intervention
Hoseini Haji 2020	Directive counselling
Hsu 2003	Lifestyle choice
Hu 2021	Not decision coaching
Hulin 2017	Decision coaching in both groups
Jacobsen 2012	Participation in a trial decision
Jalil 2020	Not decision coaching
Jayadevappa 2019	Not decision coaching
Kaplan 1989	Not for a specific decision – it was in preparation for any decision
Katz 2005	Not an RCT
Keyserling 2014	Lifestyle choice
Kim 2005	Not decision coaching; not an RCT
Kim 2007	Not an RCT
Kinney 2014	Not decision coaching
Kinney 2016	Not decision coaching
Kirk 2016	Conference abstract
Knerr 2017	Not decision coaching
Kosaka 2017	Not decision coaching
Labrias 2020	Conference abstract
Larsson 2017	Not decision coaching
Lauffenburger 2019	Not decision coaching
Lazcano 2000	Not decision coaching
Le 2015	Not decision coaching
LeBlanc 2015	Not decision coaching
Legare 2003	Not decision coaching
Levi 2017	Not decision coaching; Advanced care planning; not an RCT
Lewis 2018	The aim was to educate people about their results. Counselling (not described, delivered by providers ‐ also not clear on whether they were trained) did not aim to prepare people to make a decision with a health care provider.
Liao 2016	Conference abstract
Liberatore 2003	Not an RCT
Lurie 2011	Compared 2 decision aids, no counselling
Manchanda 2016	Decision coaching in  both groups
Manne 2010	Not decision coaching
Martin 1999	Thesis
Maslin 1998a	Not decision coaching
Maslin 1998b	Not decision coaching
Massey 2016	Not decision coaching
McIlvennan 2018	Not decision coaching
Metz 2019	Not decision coaching
Miller 2005b	Decision coaching in both groups
Minneci 2019	Not decision coaching
Mino 2020	Conference abstract
Mohamed 2018	Conference abstract
Moudi 2020	Not an RCT; quasi‐experimental study design
Murray 2004	Hypothetical decision
Myers 2012	Conference abstract
Myers 2019	It was about increasing colorectal screening behaviours and main outcome was adherence. It was not about a coach providing non‐directive support for a specific decision.
NCT00262899	Decision coaching in both groups
Nichol 1992	Not an RCT
Nosratabadi 2018	Decision coaching in both groups
O'Cathain 2002	Participation in a trial decision
Oduncu 2002	Not an RCT
Omer 2007	Lifestyle choice
Parapiboon 2020	NOT non‐directive
Park 2005	Not decision coaching
Paz 2020	Not decision coaching
Pearson 2020	Directive counselling
Pendley 2018	Conference abstract
Penson 2011	Commentary
Peralta 2019	Conference abstract
Peralta 2020	Not decision coaching
Perestelo‐Perez 2017	Not decision coaching
Pieterse 2005	Not an RCT
Rahn 2014	Conference abstract
Rahn 2016	Conference abstract
Rahn 2019	Conference abstract
Raphaelis 2018	Not decision coaching
Redfern 2009	Lifestyle choice
Reinhardt 2014	Hypothetical decision
Rimer 2001	Not decision coaching
Rimer 2002	Directive counselling
Rothert 1997	Decision coaching delivered as a group
Roussi 2010	Both groups received genetic counselling
Schroy 2012	Not decision coaching
Schubart 2019	Advanced care planning
Schwartz 2014	Decision coaching in both groups
Schwartz 2018	Not decision coaching
Sepucha 2002	Not focussed on a specific decision
Sepuchra 2000	Physician counselling
Shelin 2019	Decision coaching in both groups
Sheridan 2011	Automated support
Sheridan 2014	Automated support
Simoni 2015	Lifestyle choice
Simons 2017	Conference abstract
Skinner 2002	Not decision coaching: tailored print materials versus non‐tailored print materials
Street 2010	Not decision coaching
SUPPORT 1996	Decision coaching in both groups
Sweet 2014	No description of the genetic counselling intervention, training of health care provider; seemed to be an education‐based intervention that provided directive information on conditions that people were identified as being at risk of
Swoboda 2016	Lifestyle choice
Swoboda 2017	Thesis
Talen 2011	Not decision coaching
Tea 2018	Decision coaching in both groups
Thomas 2013	Not decision coaching
Thomson 2007	Physician counselling
Tingey 2015	Not decision coaching
Toohill 2014	Not decision coaching
Van Peperstraten 2011	Conference abstract
Van Roosmalen 2004	Not decision coaching
Van Tol‐Geerdink 2013	Wrong intervention
Veroff 2013	Decision coaching in both groups
Vitinius 2019	Conference abstract
Vitinius 2020	Conference abstract
Voepel‐Lewis 2020	Not decision coaching
Volk 1999	Educational videotape versus no intervention
Wakefield 2008	Decision coaching in both groups
Walczak 2017	Advanced care planning
Wang 2005	Not decision coaching
Waterman 2015	Not decision coaching
Waterman 2017	Conference abstract
Waterman 2018a	Conference abstract
Waterman 2018b	Not decision coaching
Waterman 2019	Conference abstract
Wensaas 2014	Conference abstract
Wenzel 2019	Conference abstract
Westermann 2013	Directive counselling
Wevers 2013	Conference abstract
Wevers 2014	Not decision coaching
Wevers 2017	Not decision coaching
Wilkie 2020	Not decision coaching
Wolf 1996	Not decision coaching
Wolf 2000	Not decision coaching
Yee 2014	Decision coaching in both groups

^{CD‐ROM: Compact disc read‐only memory
RCT: Randomized control trial
SDM: Shared decision making}

Characteristics of studies awaiting classification [ordered by study ID]

Hadizadeh‐Talasaz 2021.

Methods	RCT
Participants	Pregnant women with a previous caesarean section
Interventions	Intervention: 90‐min counselling session based on the three talk model of shared decision making Control: Usual care
Outcomes	Decisional conflict, decision regret
Notes

Perfors 2021.

Methods	RCT
Participants	Newly diagnosed patients with curable cancer (breast, lung, colorectal, gynaecologic and melanoma)
Interventions	Intervention: Gastric Residual Volume in Feeding Advancement in Preterm Infants intervention group were offered structured follow‐up guidance from primary care consisting of two components: (1) a Time Out consultation with the physician and (2) structured follow‐up during cancer treatment by a primary care oncology nurse and the physician Control: Usual care
Outcomes	Primary: Perceived level of shared decision making Secondary: Received information, perceived self‐efficacy
Notes

Wagner 2021.

Methods	Hybrid Cluster RCT
Participants	Six HIV clinics Clients were eligible if they met the following criteria: (1) in a serodiscordant relationship (partner’s HIV‐negative status confirmed by rapid HIV test prior to enrolment), (2) of reproductive age (men age 15‐60 years; women age 15‐45), (3) considering childbearing with their partner (determined via triage screening item), (4) not currently pregnant (determined by a pregnancy test prior to enrolment), and (5) reports having disclosed HIV status to partner.
Interventions	Intervention groups: "Our Choice" using either a high intensity Safer conception counselling or low intensity Safer conception counselling approach (differentiated by amount of training and supervision) Control: Usual care
Outcomes	Primary: Use of appropriate reproductive method to achieve stated reproductive goal (among whole sample), Accurate use of safer conception methods (among those trying to conceive), use of modern contraceptives (among those not trying to conceive) Secondary: Any use of safer conception methods (among those trying to conceive), use of any method to prevent pregnancy (among those not trying to conceive), pregnancy status, partner seroconversion
Notes	See also: Additional file 2: Supplemental Table 1. Our Choice safer conception counselling topics by session for a more detailed description of the coaching intervention.

Wang 2021.

Methods	RCT
Participants	Males aged 40‐80 with a prostate specific antigen index higher than 4.0 ng/mL
Interventions	Intervention: biopsy decision‐support intervention based on the Ottawa Decision Support Framework tailored to individual patient’s needs using counselling, decision aid, and coaching Control: written health education flyer
Outcomes	Knowledge, decision self‐efficacy, decisional conflict,  prostate biopsy decision
Notes

^{HIV=human immunodeficiency virus}
 

Characteristics of ongoing studies [ordered by study ID]

Chambers 2008.

Study name	ProsCan for Men: Randomised controlled trial of a decision‐support intervention for men with localised prostate cancer
Methods	RCT
Participants	700 men newly diagnosed with prostate cancer
Interventions	A tele‐based nurse delivered 5‐session decision‐support/psychosocial intervention vs usual care
Outcomes	Cancer threat appraisal; decision‐related distress and bother from treatment side effects; involvement in decision making; satisfaction with healthcare; heathcare utilisation; use of healthcare resources; and a return to previous activities
Starting date	Not yet assessed
Contact information	Suzanne K Chambers, Griffith University
Notes	Trial #: ACTRN012607000233426

DRKS00015527.

Study name	Evaluation of a decision‐coaching programme for structured decision support in preference sensitive decisions of risk‐adapted prevention for BRCA1/2 gene mutation carriers (EDCP‐BRCA)
Methods	RCT
Participants	400 women newly diagnosed with a BRCA 1/2 gene mutation
Interventions	Decision coaching versus standard care
Outcomes	Congruence between the preferred and actual role in the decision‐making process; satisfaction with the actual role, decisional conflict, knowledge and attitude towards prevention strategies, stage of decision making, symptoms of anxiety and depression, coping self‐efficacy, subjective impact of the test result and self‐concept of BRCA1/2 carriers
Starting date	2019/11/01
Contact information	Ms.  Anna  Isselhard , Institut für Gesundheitsökonomie und klinische Epidemiologie, Universitätsklinikum Köln, Germany
Notes	Trial #: DRKS00015527

IRCT2017100231117N5.

Study name	Comparison of effect of crisis counselling on anxiety, decision‐making style and strategies in women with unplanned pregnancy in intervention (receive crisis counselling) with control (do not receive crisis counselling) groups
Methods	RCT
Participants	70 women aged 15‐49 with unplanned pregnancy
Interventions	Crisis counselling versus control
Outcomes	Anxiety; decision‐making style and strategies about unplanned pregnancy
Starting date	2017‐11‐01
Contact information	Soghra Khani, Mazandaran University of Medical Sciences, Iran
Notes	Trial #: IRCT2017100231117N5

IRCT20200105046014N1.

Study name	The effect of individual counseling on the intention to choose
Methods	RCT
Participants	Aged 18‐35 years old Single pregnancy
Interventions	The 'intervention group' received three Individual of Fogg model counselling sessions for 45‐60 minutes and once every two weeks, based on a structure developed by the researcher and approved by a professor of psychology. Intervention 2: Control group: The 'control group' received routine counselling about vaginal birth after caesarean by midwives at health centres.
Outcomes	Primary: Decision conflict, decision self‐efficacy, fear of delivery, Fogg Model Structures, self‐efficacy of delivery, the intent to choose vaginal birth after caesarean
Starting date
Contact information
Notes	Trial #: IRCT20200105046014N1 Unable to access trial record at this time

Karra 2021.

Study name	User‐centered counseling and male involvement in contraceptive decision making
Methods	RCT
Participants	Married women aged 18‐35
Interventions	1. A control group (T0), in which a woman received a private counselling session on the full range of 13 contraceptive methods following a standard counselling process. 2. An intervention group (T1), in which a woman was encouraged (but not compelled) to invite her husband to a joint counselling session. A woman (and her husband if she chose to invite him) then received a private counselling session on a full range of 13 contraceptive methods. 3. An intervention group (T2), in which a woman received a private counselling session on 5 targeted contraceptive methods based on her baseline preferences for family planning. 4. An intervention group (T3), in which a woman was encouraged to invite her husband to a joint counselling session. A woman (and her husband if she chose to invite him) was then counselled on 5 targeted contraceptive methods based on her baseline preferences for family planning.
Outcomes	Primary: Attitude or knowledge of family planning, contraceptive use, contraceptive preferences, pregnancy and fertility outcomes, use of local health facilities, husband’s preferences, women’s autonomy, empowerment, and decision making
Starting date	2019‐06‐14
Contact information	Bagrey Ngwira
Notes	Trial #: AEARCTR‐0004194

Makarov 2021.

Study name	Trial of community health worker‐led decision coaching
Methods	RCT
Participants	Black men attending a routine primary care appointment
Interventions	Intervention group: a decision aid along with decision coaching on prostate specific antigen screening from a community health worker Control Group: a decision aid along with community health worker interaction on dietary and lifestyle modification to serve as an attention control
Outcomes	Decision self‐efficacy Prostate Specific Antigen screening rates
Starting date	October 15, 2019
Contact information	Shannon Ciprut: shannon.ciprut@nyumc.org Natalia Martinez‐Lopez: Natalia.Martinez-lopez@nyulangone.org
Notes	Trial #: NCT03726320

NCT00685256.

Study name	Standard genetic counseling with or without a decision guide in improving communication between mothers undergoing BRCA1/2 testing and their minor‐age children
Methods	RCT
Participants	245 mothers at least 21 years old who self‐identified as primary caregivers to minor‐age children (ages 8‐21 years old) undergoing BRCA1/2 testing
Interventions	Genetic counselling with communication aid versus standard genetic counselling
Outcomes	Decision conflict;  parents' decision satisfaction with their communication decision;  decision quality; disclosure of genetic testing results by parents; child‐rearing alliance between parents; parent‐child communication; family functioning; children's stress and worry; parents' knowledge and awareness of the advantages and disadvantages of disclosing maternal BRCA1/2 test results to their children; parents' preferences and values regarding family communication of genetic test results; decision self‐efficacy; cognitive appraisals
Starting date	March 2008
Contact information	Kenneth Tercyak, Lombardi Comprehensive Cancer Center
Notes	Trial #: NCT00685256

NCT01263678.

Study name	Helping patients with spinal stenosis make a treatment decision: a randomized study assessing the benefits of health coaching (SST HCoach RCT)
Methods	RCT
Participants	199 patients over 18 years old with lumbar spinal stenosis
Interventions	Decision coaching versus usual care
Outcomes	Decisional conflict;  self‐efficacy; the number of treatment decision‐related clinical contacts; treatment follow‐through; decision regret
Starting date	November 2010
Contact information	Jon D Lurie, M.D., Dartmouth‐Hitchcock Medical Center
Notes	Trial #: NCT01263678

NCT03956004.

Study name	The impact of a technology‐enabled decision aid, for patients with hip and knee osteoarthritis on decision quality, level of shared decision‐making, patient satisfaction and magnitude of limitations
Methods	RCT
Participants	145 patients with a presumptive diagnosis of knee osteoarthritis aged between 45 and 89
Interventions	Full decision aid, consisting of education on osteoarthritis and treatment options, preferences and values elicitation, and personalised risk/benefit estimates based on patient's response to patient‐reported outcome measures versus decision aid only
Outcomes	Knee Osteoarthritis Decision Quality instrument: decision process (SDM) and concordance (received treatment that matched their stated preference); level of shared decision making; patient satisfaction with management of condition; measure of patient‐reported, knee‐related stiffness, pain, and function; decision to undergo total knee replacement surgery (yes/no); average time of patient visit and average time spent with provider
Starting date	March 28, 2019
Contact information	Kevin Bozic, Dell Medical School at The University of Texas at Austin
Notes	Trial #: NCT03956004

NCT04659005.

Study name	Nurse‐led decision counseling on hepatocellular carcinoma screening
Methods	RCT
Participants	Patients with hepatitis B virus infection (hepatitis B surface antigen positive), aged 18‐65 years old
Interventions	Intervention: Nurse‐led decision counselling Control: Usual care
Outcomes	Primary: Hepatocellular carcinoma  screening uptake rates Secondary: Knowledge, decisional conflict, preventive health model construct instrument
Starting date	January 1, 2021
Contact information	Caixia LI: caixiali@link.cuhk.edu.hk Carmen Chan: whchan@cuhk.edu.hk
Notes	Trial #: NCT04659005

Nelson 2020.

Study name	A web‐based intervention to reduce decision conflict regarding HIV pre‐exposure prophylaxis
Methods	RCT
Participants	Men and women aged 18 and older, who are racially categorised as black, and assessed by the referring health care provider as being a good candidate for starting HIV pre‐exposure prophylaxis
Interventions	Decision‐support intervention versus information only
Outcomes	Decisional conflict
Starting date	April 9, 2019
Contact information	LaRon E Nelson, PhD, RN: laron.nelson@yale.edu
Notes	Trial #: NCT03637244

Ockhuysen‐Vermey 2008.

Study name	Communicating breast cancer risks: a genetic counsellor's role in improving patient understanding to increase informed decision‐making
Methods	RCT
Participants	300 women with a family history of breast cancer
Interventions	The intervention: additional risk information is given to healthy women with a family history of breast cancer immediately after standard counselling with the clinical geneticist.
Outcomes	Adequate risk perception; cognitive evaluation (knowledge about hereditary breast cancer, informed decisions); psychological and affective evaluation ([cancer] anxiety, worry); evaluation of perceived benefits and helpfulness of the additional risk counselling; expected intention or actual uptake of methods of breast cancer detection and prevention
Starting date	15/07/2004
Contact information	Dr D R M Timmermans, VU Medisch Centrum, Afd Sociale Geneeskunde, Amsterdam, Netherlands
Notes	Trial #: ISRCTN14566836

Scalia 2019.

Study name	Implementation of uerine fibroid Option Grid patient decision aids across five organisational settings (UPFRONT)
Methods	Stepped wedge RCT
Participants	Females showing new or recurrent symptoms of uterine fibroids
Interventions	Uterine Fibroid Option Grid encounter patient decision aid (text only) Uterine Fibroid Option Grid encounter patient decision aid (text + pictures) Uterine Fibroid Option Grid encounter patient decision aid (online)
Outcomes	Primary: The number of eligible patients that can be identified who receive the uterine fibroid Option Grid patient decision aid Secondary: Measuring Organizational Readiness for patient Engagement, Attitudes toward Decision Aids fOr PatienTs, quality of shared decision‐making, fidelity assessment, NoMAD Normalization Process Theory survey, utility of Option Grid patient decision aids and clinician approach to implementation, measuring shared decision making, Uterine Fibroid Symptom and health‐related Quality Of Life Questionnaire symptom severity subscale, the COmprehensive Score for financial Toxicity, Chew health literacy measure, resource utilisation (ambulatory and hospital), treatment choice
Starting date	August 1, 2019
Contact information	Glyn Elwyn, MD,PhD: glynelwyn@gmail.com          Marie‐Anne Durand, PhD: Marie‐Anne.Durand@dartmouth.edu
Notes	Trial #: NCT03985449

Shirzad 2020.

Study name	Comparing the effect of "Motivational Interviewing" with "Information, Motivation and Behavioural skills" on choosing mode of delivery in pregnant women
Methods	RCT
Participants	Pregnant women in gestational age 24 to 32 weeks
Interventions	1:  face‐to‐face motivational interviewing  2: face‐to‐face Information, Motivation and Behavioural skills 3: Information, Motivation and Behavioural skills using mobile app
Outcomes	Behavioural intention, self‐efficacy, mode of delivery, usability of the mobile app
Starting date	2018‐11‐22
Contact information	Elham Shakibazadeh: shakib@zums.ac.ir
Notes	Trial #: IRCT20151208025431N7

TCTR20191226001.

Study name	The effect of empowerment education on decisional conflict and attitude towards fertility and childbearing of women with a history of failed pregnancy
Methods	RCT
Participants	Women with a history of failed pregnancy
Interventions	Empowerment education versus control
Outcomes	Decision conflict; attitude towards fertility and childbearing
Starting date	30 January 2020
Contact information	Masoumeh Kheirkhah, Iran University of Medical Sciences, Shahid Hemmat Highway, Tehran
Notes	Trial #: TCTR20191226001

TCTR20200320002.

Study name	Decisions of adolescent mothers toward using etonogestrel implant: comparison between the people that knowledge from Print Media in individual counseling and counseling from video‐based group counseling
Methods
Participants	Adolescent mothers
Interventions	Intervention: group counselling by education video, individual counselling by brochure
Outcomes	Primary outcome: Decision use of  etonorgestrel implant [time frame after counselling before discharge use of  implanon]  Secondary outcome: Continuous use of implant contraceptive [time frame 1 year follow‐up]
Starting date
Contact information
Notes	Trial #: TCTR20200320002 Unable to access trial record at this time

^{BRCA1: Breast cancer type 1 susceptibility protein
HIV: human immunodeficiency virus
RCT: randomized control trial
SDM: shared decision making}
 

Differences between protocol and review

1. New team authors (MC, JF, AD, JZ) and three who withdrew (AC, BH, WO)

2. The term “evidence‐based information (such as a decision aid)” replaced “another intervention type, such as a patient decision aid”.

3. Updated the search strategy in May 2020 according to Anne Parkhill, Information Specialist, and this search strategy was repeated in June 2021.

4. Updates to the selection criteria: Additional detail to the description of the decision coach from “trained” to “trained or using a protocol in decision coaching”. We also excluded articles that described healthcare providers who were not prepared for decision coaching with training or a protocol, or not described as having direct interests in providing decision coaching. We also excluded studies providing decision coaching to groups because of the potential influence of others in the group on decision making.  

We defined included comparisons (e.g. usual care or other) with the qualifier “only” to be explicit about pairing decision coaching with another intervention, and we are clear that “evidence‐based information” includes patient decision aids.

5. Defined “Types of outcome measures”: We took the following steps before including outcomes in the Results and Summary of findings tables: If the outcomes of interest were reported with only one measure (Decisional Conflict subscales) by the article, then no further action was needed. If reported with more than one measure (e.g. knowledge) we had planned that the outcome measures for each article would be listed and two review authors would independently make decisions about what is most relevant to patients; any differences were resolved by the involvement of a third author. For outcomes with more than one measure (e.g. knowledge) that could not be included in meta‐analysis, we reported results descriptively.

6. For the search strategy, we made the following changes:

• excluded databases of dissertations and conference proceedings, as these were unlikely to report on a full study that had met the peer‐reviewed requirements of our study;

• excluded studies evaluating decision coaching delivered to a group because it is difficult to determine the unique effect of decision coaching versus the group effect;

• excluded studies evaluating advanced care planning because values and characteristics of decisions can change as patients get closer to making the actual decision;

• excluded genetic counselling studies when the genetic counselling intervention was not described and/or was not the primary intervention being evaluated.

7. Updated 'Primary outcomes': Given that the main aim of decision coaching is to prepare patients to participate in decision making under ‘Preparation for active participation in making a health decision’ we included other indicators of preparation for active participation in making a health decision: decisional self‐confidence and patient involvement in decision making (patient, observer‐reported) including perceived involvement, preferred level of involvement, participated at preferred level of involvement. ‘Quality of decision coaching’ was removed as primary outcome, as it is not an outcome of decision coaching: decision coaching supports patients to participate in the process of decision making with a goal of improving the quality of the decision‐making process and as a consequence, the quality of the decision. It is now included under ‘Other outcomes related to the decision‐making process’. 

8. Details on our 'Data extraction and management': Duplicate data extraction was performed on key elements only, including description of decision coaching, outcomes, and risk of bias based on editorial feedback.

Contributions of authors

The work presented here is the result of shared interests of an international network that includes patient partners who are healthcare consumers, healthcare providers, educators, and researchers with expertise in SDM, various research methodologies, Cochrane Reviews, and knowledge translation (KT). All authors have discussed and agreed upon their roles in this review, as follows.

Study proposal: JJ, SKo, LB, AC, SD, IDG, BH, JK, SKi, FL, KL, AL, WO, TR, AR, CR, MS, DS

Study: JJ, SKo, LB, SD, IDG, JK, SKi, FL, KL, AL, TR, AR, CR, MS, DS, MC, JF, AD, JZ contributed to the decisions leading to the conduct, writing, and agreed on the final version of the report.

Patient partners on this review: MS, AL

Team leaders: JJ, SKo, DS, MS

Project coordinator: MC

International collaborative research group: CR, AD, LB, AC, SD, JF, IDG, BH, JK, SKi, FL, KL, WO, TR, AR, JZ

JJ is the guarantor for this review. JJ will be responsible for conducting the review update.

Sources of support

Internal sources

• No sources of support provided

External sources

• CIHR Project Grant, Canada

  CIHR Project Grant April 2019‐September 2020 (FRN# PJT‐162135)

• Canada Research Chair in Shared Decision Making and Knowledge Translation, Canada

  FL holds a Canada Research Chair in Shared Decision Making and Knowledge Translation

• University Research Chair in Knowledge Translation to Patients, Canada

  DS holds a University Research Chair in KT to Patients

• CIHR Foundation Grant, Canada

  IDG is a recipient of a CIHR Foundation Grant (FRN# 143237)

Declarations of interest

Janet Jull: none known

Sascha Köpke (SKo): Co‐authorship of one included study

Maureen Smith: receives honoraria from the Ontario Ministry of Health to attend meetings, has received travel scholarships to attend conferences as a consumer, and receives an honorarium for roles as a co‐investigator and a knowledge user on two Canadian Institutes of Health Research grants.

Meg Carley: none known

Jeanette Finderup: none known

Anne C Rahn: Co‐authorship of one included study

Laura Boland: none known

Sandra Dunn: none known

Andrew Dwyer: is an Assistant Professor of Nursing at Boston College whose research focusses on developing more person‐centered approaches to care. He receives funding from Boston College and the U.S. National Institutes of Health (U.S.A.) and receives funding to cover travel expenses for his faculty participation in a Swiss rare diseases summer school. Dr. Dwyer has no competing interests to declare.

Jürgen Kasper: Co‐authorship of one included study

Simone Maria Kienlin (SKi): none known

France Légaré: none known

Krystina B Lewis: none known

Anne Lyddiatt: none known

Claudia Rutherford: none known

Jungqiang Zhao: none known

Tamara Rader: none known

Ian D Graham: none known

Dawn Stacey: is a Professor in the School of Nursing at the University of Ottawa and Senior Scientist at the Ottawa Hospital Research Institute where she conducts funded studies to evaluate the effectiveness of patient decision aids and decision coaching. The institution where she is employed, the University of Ottawa, has received funding to support her research studies from national granting agencies and cancer programmes. She has received funding for consultation with the Washington State Health Care Authority for the development and implementation of criteria for certifying patient decision aids. Finally, she received funding to travel for the Shared Decision Making Advisory Board Meeting in Vejle, Denmark, working with Safer Care Victoria in Melbourne, Australia, meetings and training for the Joint Commission of Taiwan.

New