Interventions for increasing the use of shared decision making by healthcare professionals

France Légaré; Rhéda Adekpedjou; Dawn Stacey; Stéphane Turcotte; Jennifer Kryworuchko; Ian D Graham; Anne Lyddiatt; Mary C Politi; Richard Thomson; Glyn Elwyn; Norbert Donner‐Banzhoff

doi:10.1002/14651858.CD006732.pub4

. 2018 Jul 19;2018(7):CD006732. doi: 10.1002/14651858.CD006732.pub4

Interventions for increasing the use of shared decision making by healthcare professionals

France Légaré ^1,^✉, Rhéda Adekpedjou ², Dawn Stacey ³, Stéphane Turcotte ⁴, Jennifer Kryworuchko ⁵, Ian D Graham ⁶, Anne Lyddiatt ⁷, Mary C Politi ⁸, Richard Thomson ⁹, Glyn Elwyn ¹⁰, Norbert Donner‐Banzhoff ¹¹

Editor: Cochrane Effective Practice and Organisation of Care Group

PMCID: PMC6513543 PMID: 30025154

Abstract

Background

Shared decision making (SDM) is a process by which a healthcare choice is made by the patient, significant others, or both with one or more healthcare professionals. However, it has not yet been widely adopted in practice. This is the second update of this Cochrane review.

Objectives

To determine the effectiveness of interventions for increasing the use of SDM by healthcare professionals. We considered interventions targeting patients, interventions targeting healthcare professionals, and interventions targeting both.

Search methods

We searched CENTRAL, MEDLINE, Embase and five other databases on 15 June 2017. We also searched two clinical trials registries and proceedings of relevant conferences. We checked reference lists and contacted study authors to identify additional studies.

Selection criteria

Randomized and non‐randomized trials, controlled before‐after studies and interrupted time series studies evaluating interventions for increasing the use of SDM in which the primary outcomes were evaluated using observer‐based or patient‐reported measures.

Data collection and analysis

We used standard methodological procedures expected by Cochrane.

We used GRADE to assess the certainty of the evidence.

Main results

We included 87 studies (45,641 patients and 3113 healthcare professionals) conducted mainly in the USA, Germany, Canada and the Netherlands. Risk of bias was high or unclear for protection against contamination, low for differences in the baseline characteristics of patients, and unclear for other domains.

Forty‐four studies evaluated interventions targeting patients. They included decision aids, patient activation, question prompt lists and training for patients among others and were administered alone (single intervention) or in combination (multifaceted intervention). The certainty of the evidence was very low. It is uncertain if interventions targeting patients when compared with usual care increase SDM whether measured by observation (standardized mean difference (SMD) 0.54, 95% confidence interval (CI) ‐0.13 to 1.22; 4 studies; N = 424) or reported by patients (SMD 0.32, 95% CI 0.16 to 0.48; 9 studies; N = 1386; risk difference (RD) ‐0.09, 95% CI ‐0.19 to 0.01; 6 studies; N = 754), reduce decision regret (SMD ‐0.10, 95% CI ‐0.39 to 0.19; 1 study; N = 212), improve physical (SMD 0.00, 95% CI ‐0.36 to 0.36; 1 study; N = 116) or mental health‐related quality of life (QOL) (SMD 0.10, 95% CI ‐0.26 to 0.46; 1 study; N = 116), affect consultation length (SMD 0.10, 95% CI ‐0.39 to 0.58; 2 studies; N = 224) or cost (SMD 0.82, 95% CI 0.42 to 1.22; 1 study; N = 105).

It is uncertain if interventions targeting patients when compared with interventions of the same type increase SDM whether measured by observation (SMD 0.88, 95% CI 0.39 to 1.37; 3 studies; N = 271) or reported by patients (SMD 0.03, 95% CI ‐0.18 to 0.24; 11 studies; N = 1906); (RD 0.03, 95% CI ‐0.02 to 0.08; 10 studies; N = 2272); affect consultation length (SMD ‐0.65, 95% CI ‐1.29 to ‐0.00; 1 study; N = 39) or costs. No data were reported for decision regret, physical or mental health‐related QOL.

Fifteen studies evaluated interventions targeting healthcare professionals. They included educational meetings, educational material, educational outreach visits and reminders among others. The certainty of evidence is very low. It is uncertain if these interventions when compared with usual care increase SDM whether measured by observation (SMD 0.70, 95% CI 0.21 to 1.19; 6 studies; N = 479) or reported by patients (SMD 0.03, 95% CI ‐0.15 to 0.20; 5 studies; N = 5772); (RD 0.01, 95%C: ‐0.03 to 0.06; 2 studies; N = 6303); reduce decision regret (SMD 0.29, 95% CI 0.07 to 0.51; 1 study; N = 326), affect consultation length (SMD 0.51, 95% CI 0.21 to 0.81; 1 study, N = 175), cost (no data available) or physical health‐related QOL (SMD 0.16, 95% CI ‐0.05 to 0.36; 1 study; N = 359). Mental health‐related QOL may slightly improve (SMD 0.28, 95% CI 0.07 to 0.49; 1 study, N = 359; low‐certainty evidence).

It is uncertain if interventions targeting healthcare professionals compared to interventions of the same type increase SDM whether measured by observation (SMD ‐0.30, 95% CI ‐1.19 to 0.59; 1 study; N = 20) or reported by patients (SMD 0.24, 95% CI ‐0.10 to 0.58; 2 studies; N = 1459) as the certainty of the evidence is very low. There was insufficient information to determine the effect on decision regret, physical or mental health‐related QOL, consultation length or costs.

Twenty‐eight studies targeted both patients and healthcare professionals. The interventions used a combination of patient‐mediated and healthcare professional directed interventions. Based on low certainty evidence, it is uncertain whether these interventions, when compared with usual care, increase SDM whether measured by observation (SMD 1.10, 95% CI 0.42 to 1.79; 6 studies; N = 1270) or reported by patients (SMD 0.13, 95% CI ‐0.02 to 0.28; 7 studies; N = 1479); (RD ‐0.01, 95% CI ‐0.20 to 0.19; 2 studies; N = 266); improve physical (SMD 0.08, ‐0.37 to 0.54; 1 study; N = 75) or mental health‐related QOL (SMD 0.01, ‐0.44 to 0.46; 1 study; N = 75), affect consultation length (SMD 3.72, 95% CI 3.44 to 4.01; 1 study; N = 36) or costs (no data available) and may make little or no difference to decision regret (SMD 0.13, 95% CI ‐0.08 to 0.33; 1 study; low‐certainty evidence).

It is uncertain whether interventions targeting both patients and healthcare professionals compared to interventions of the same type increase SDM whether measured by observation (SMD ‐0.29, 95% CI ‐1.17 to 0.60; 1 study; N = 20); (RD ‐0.04, 95% CI ‐0.13 to 0.04; 1 study; N = 134) or reported by patients (SMD 0.00, 95% CI ‐0.32 to 0.32; 1 study; N = 150 ) as the certainty of the evidence was very low. There was insuffient information to determine the effects on decision regret, physical or mental health‐related quality of life, or consultation length or costs.

Authors' conclusions

It is uncertain whether any interventions for increasing the use of SDM by healthcare professionals are effective because the certainty of the evidence is low or very low.

Plain language summary

A review of activities to help healthcare professionals share decisions about care with their patients

What is the aim of this review?

Healthcare professionals often do not involve their patients in decision making about their care. With shared decision making, healthcare professionals inform patients about their choices and invite them to choose the option that reflects what is important to them, including the option not to proceed with treatment. Shared decision making is said to be desirable because patient involvement is accepted as a right and patients in general want more information about their health condition and prefer to take an active role in decisions about their health. The aim of this review was to find out if activities to increase shared decision making by healthcare professionals are effective or not. Examples of these activities are training programs, giving out leaflets, or email reminders. Cochrane researchers collected and analyzed all relevant studies to answer this question, and found 87 studies.

Key messages

A great variety of activities exist to increase shared decision making by healthcare professionals, but we cannot be confident about which of these activities work best because the certainty (or the confidence) of the evidence has been assessed as very low.

What was studied in the review?

Our review examined the 87 studies that tested what kind of activities work best to help healthcare professionals involve their patients more in decision making about their care. We also examined the effect of these activities on decision regret, physical or mental health‐related quality of life, length of the consultation, and cost.

The studies were so different that these activities were difficult to compare.

First, we divided the studies into ones that used outside observers to measure shared decision making and ones that used patients to measure shared decision making.

We then divided studies into ones that looked at activities a) for healthcare professionals only (e.g. training), b) for patients only (e.g. giving them a decision aid, which is a pamphlet explaining options and inviting them to think about their values and preferences), and c) for both healthcare professionals and patients (e.g. training plus a decision aid).

Finally, we subdivided each of these three categories into studies that compared the activity with usual care and studies that compared the activity with another activity.

We also looked at how certain the evidence was for our primary outcome (the extent to which healthcare professionals involve their patients more in decision making about their care) and secondary outcomes (decision regret, physical or mental health‐related quality of life, length of the consultation, and cost) of interest.

What are the main results of the review?

Forty‐four studies looked at activities for patients only, while 28 studies looked at activities for both healthcare professionals and patients, and 15 studies looked at activities for healthcare professionals only.

While studies in all three categories had tested many different activities to increase shared decision making by healthcare professionals, overall we cannot be confident in the effectiveness of these activities because the certainty of the evidence was weak. This is because there were many possible sources of error (e.g. not making sure the tested activities were not also provided to the comparison groups), and often poor reporting of results (i.e. not providing enough information to judge the quality of the evidence).

Although it was hard to come to any firm conclusions, we can say that compared to no activity at all, activities for healthcare professionals may slightly improve mental health‐related quality of life, but make little or no difference to physical health‐related quality of life (two studies). We can also say that activities targeting both healthcare professionals and patients may make little or no difference to decision regret (one study).

How up‐to‐date is this review?

We searched for studies published up to June 2017.

Summary of findings

Background

Description of the condition

There is increasing recognition of the ethical imperative to share important decisions with patients (Salzburg Global Seminar 2011). Shared decision making (SDM) can be defined as an interpersonal, interdependent process in which health professionals, patients and their caregivers relate to and influence each other as they collaborate in making decisions about a patient’s health (Charles 1997; Légaré 2011; Légaré 2013; Towle 1999). It is considered the crux of patient‐centered care (Weston 2001). Briefly, SDM depends on knowing and understanding the best available evidence about the risks and benefits across all available options while ensuring that the patient's values and preferences are taken into account (Charles 1997; Elwyn 1999; Towle 1999).

Although SDM represents a complex set of behaviors that must be achieved by both members of the patient‐healthcare professional dyad (LeBlanc 2009), it is possible to specify behaviors that both parties must adopt for SDM to occur in clinical practice (Frosch 2009; Légaré 2007a). A systematic review of SDM as a concept identified 161 definitions and summarized the key elements into one integrative model of SDM in medical encounters (Makoul 2006). This model identifies nine essential elements that can be translated into specific SDM‐related behaviors that healthcare professionals need to demonstrate during consultations with patients:

define and explain the healthcare problem,
present options,
discuss pros and cons (benefits, risks, costs),
clarify patient values and preferences,
discuss patient ability and self‐efficacy,
present what is known and make recommendations,
check and clarify the patient's understanding,
make or explicitly defer a decision, and
arrange follow‐up.

Description of the intervention

A variety of interventions have been designed to change healthcare professionals' behavior. Based on the Effective Practice and Organisation of Care (EPOC) taxonomy of interventions (EPOC 2015), these interventions aim at changing the performance of healthcare professionals through interactions with patients, or information provided by or to patients. Interventions may include, but are not limited to, the distribution of printed educational materials, educational meetings, audit and feedback, reminders, educational outreach visits and patient‐mediated interventions. In the context of SDM it is possible to identify three overarching categories of implementation intervention: 1) interventions targeting patients, 2) interventions targeting healthcare professionals, and 3) interventions targeting both.

How the intervention might work

Theoretical and empirical evidence about behavior change in healthcare professionals (Godin 2008) and complex behavior change frameworks (Michie 2009) allow us to make certain hypotheses regarding the mechanisms by which interventions might promote SDM. For example, the distribution of printed educational materials may improve professionals' attitudes to SDM by reinforcing their intention to engage in SDM (Giguère 2012). The training of professionals in SDM through educational meetings may increase professionals' perceptions of self‐efficacy, or their belief in their ability to succeed in a situation, which is one of the key determinants of behavior (Godin 2008). Patient‐mediated interventions could be a discussion with a nurse, a patient education program, or a decision aid, for example. Decision aids are tools (they can be pamphlets or online modules) that help patients become involved in decision making. They help patients clarify the decision that needs to be made, and give information about options and outcomes. They also invite patients to articulate their personal values and preferences regarding the options (Stacey 2017). In turn, the habits of healthcare professionals may change when patients themselves take the initiative to engage more in the decision‐making process, as this may increase health professionals’ knowledge and use of emerging evidence in their area of expertise (Brouwers 2010).

Regarding the association between SDM and patient outcomes, some authors have shown that communication between healthcare professionals and patients, including SDM, can lead to improved health outcomes in direct but also in indirect ways (Street 2009). Thus, according to an adapted conceptual framework linking clinician‐patient communication to health outcomes, SDM can have an impact on affective‐cognitive outcomes (e.g. knowledge, understanding, satisfaction, trust), behavioral outcomes (treatment decisions, adherence to recommended treatments and adoption of health behaviors), as well as health outcomes (e.g. quality of life, self‐rated health and biological measures of health) (Shay 2015).

Why it is important to do this review

Policy makers perceive SDM as desirable (Shafir 2012) because: a) patient involvement is accepted as a right (Straub 2008); b) patients in general want more information about their health condition and prefer to take an active role in decisions about their health (Alston 2012; Kiesler 2006); c) SDM may reduce the overuse of options not clearly associated with benefits for all and increase the use of options clearly associated with benefits for the vast majority of the concerned population (Mulley 2012); d) SDM may reduce unwarranted healthcare practice variations (Wennberg 2004); and e) SDM may foster the sustainability of the healthcare system by increasing patient ownership of their own health care (Coulter 2006).

Nonetheless, SDM has not yet been widely implemented in clinical practice. A systematic review of 33 studies using the Observing Patient Involvement in Decision Making instrument (OPTION) showed low levels of patient‐involving behaviors (Couët 2013). The rationale for this review of interventions for increasing use of SDM among healthcare professionals is to determine what kinds of intervention have been shown to increase patient‐involving behaviors among healthcare professionals.

This is the second update of a previously published Cochrane review. The review was first undertaken in 2010 (Légaré 2010) and updated in 2014 (Légaré 2014). As the demand for SDM training programs for healthcare professionals is increasing internationally (Diouf 2016), we considered a second update was important to keep abreast of developments.

Objectives

Methods

Criteria for considering studies for this review

Types of studies

This review considered:

randomized trials;
non‐randomized trials;
controlled before‐after studies (CBAs); and
interrupted time series (ITS) analyses.

To be included as a CBA, the Cochrane EPOC group (Effective Practice and Organisation of Care) requires the study to have a minimum of two intervention sites and two control sites. For ITS studies, there needs to be a clearly defined point in time when the intervention occurred and at least three data points before and three after the intervention (EPOC 2017).

Types of participants

Participants could be any healthcare professional (e.g. physicians, nurses, pharmacists, social workers), including professionals in training (for example, medical residents). We defined professionals as being licensed or registered to practice or, in the case of physicians in training, as having completed their basic pre‐licensure education. Participants could also be patients, including healthcare consumers and simulated patients. However, studies that included simulated patients were deemed eligible only if the outcome was observer‐reported.

Types of interventions

We included studies that evaluated an intervention designed to increase the use of SDM. Interventions were organized into three target categories using the EPOC taxonomy of interventions (EPOC 2015):

interventions targeting patients (for example, patient‐mediated interventions);
interventions targeting healthcare professionals (for example, distribution of printed educational material, educational meetings, audit and feedback, reminders and educational outreach visits);
interventions targeting both patients and healthcare professionals (for example, a patient‐mediated intervention combined with an intervention targeting healthcare professionals).

Patient decision aids were considered a patient‐mediated intervention since one of their purposes is to foster patient participation in decisions during the clinical encounter (Stacey 2017). Studies that evaluated patient‐mediated interventions (for example, patients' use of patient decision aids in preparation for or during their consultation with a healthcare professional) were considered only if these studies directly assessed the healthcare professional‐related outcome of interest, that is their use of SDM (see Types of outcome measures).

Types of outcome measures

Primary outcomes

Use of SDM, using objective observer‐based outcome measures (OBOMs) or patient‐reported outcome measures (PROMs). OBOMs are instruments used by a third observer to capture the decision‐making process during an encounter between a healthcare professional and a patient/family caregiver when facing health treatment or screening decisions. They are only used in the reporting of observable concepts (e.g. signs or behaviors). Unlike clinician‐reported outcome measures, OBOMs are reported by people (e.g. teachers or caregivers) who do not have professional training relevant to the measurement being made (Velentgas 2013). PROMs are instruments that collect information directly from patients. The measurement is recorded without amendment or interpretation by a clinician or other observer. The measurement can be recorded by the patient directly, or recorded by an interviewer, provided that the interviewer records the patient's response exactly (Velentgas 2013).

Secondary outcomes

Patient outcomes

Affective‐cognitive outcomes

Knowledge
Satisfaction (satisfaction with care, with the choice, with the decision‐making process, with the intervention, helpfulness of the intervention)
Decisional conflict
Decision regret
Patient‐clinician communication
Self‐efficacy
Empowerment

Behavioral outcomes

Match between preferred and actual level of participation in decision making
Match between preferred option and decision made
Adherence to decision made

Health outcomes

Health status (generic instrument types)
Health‐related quality of life (generic instrument types)
Anxiety
Depression
Stress
Distress

Process outcomes

Consultation length
Costs
Equity

Adverse effects (potential harms of interventions)

Search methods for identification of studies

Electronic searches

We searched for studies published up to 15 June 2017. Searches were not restricted by language. The following electronic databases were searched for primary studies.

Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 5) in the Cochrane Library
Health Technology Assessment Database (HTA; 2016, Issue 4) in the Cochrane Library
NHS Economic Evaluation Database (NHSEED; 2015, Issue 2) in the Cochrane Library
PubMed
Embase Ovid (1974 to 14 June 2017)
CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1980 to 15 June 2017)
PsycINFO Ovid (1967 to June Week 1 2017)

All search strategies used are provided in Appendix 1.

Searching other resources

Trial registries

We searched:

ClinicalTrials.gov, US National Institutes of Health (NIH) at http://clinicaltrials.gov/ (search performed in week 1, August 2017);
World Health Organization International Clinical Trials Registry Platform http://apps.who.int/trialsearch/ (search performed in week 1, August 2017).

We also:

handsearched the proceedings of the International Conference on Shared Decision Making (from 2003 to 2017)(Appendix 2);
handsearched the proceedings of the annual North American meetings of the Society for Medical Decision Making (from 2004 to 2016) (Appendix 3); we intended to search the European Association for Communication in Healthcare (EACH) but were unable to obtain detailed information either online or in paper form);
reviewed reference lists of all included studies, relevant systematic reviews (Appendix 4) and primary studies (Appendix 5); and
contacted authors of relevant studies or reviews to clarify reported published information and to seek unpublished data.

Data collection and analysis

Selection of studies

Review author Rhéda Adekpedjou (RA), and graduate students Jessica Hébert (JH), Élodie Chenard (EC), Alexandrie Boucher (AB), Lionel Adisso (LA) ‐ (see Acknowledgements) independently screened each title and abstract to find studies that met the inclusion criteria. Studies were only selected if published in English or French. We retrieved full‐text copies of all studies that might be relevant or for which the inclusion criteria were not clear in the title or abstract. In this update, when more than one publication described the same study but each presented new and complementary data, we included them all. Any disagreements about selection were resolved by discussion with two review authors (ST, FL). For more details about study selection, see Figure 1.

Flow diagram of Cochrane update on interventions for increasing the use of shared decision making by healthcare professionals (up to 15 June 2017)

Data extraction and management

To extract data, we designed a form derived from the EPOC Review Group data collection checklist (EPOC 2017b). At least two review authors (including RA and ST) independently extracted data from eligible studies. We reached consensus about discrepancies, and any disagreement was adjudicated by discussion among the review authors (FL, RA, DS, ST, JK, IDG, AL, MCP, RT, GE, NDB). We entered data into Review Manager Software (RevMan 5) and checked for accuracy. When information regarding any of the above was unclear, we attempted to contact the study authors of to ask them to provide further details.

In addition to EPOC's standardized data collection checklist, we extracted the following characteristics of the settings and interventions.

Level of care: primary or specialized care (as defined by the type of provider).
Setting of care: ambulatory or non‐ambulatory care (e.g. hospitalized patients in acute‐care or long‐term care facilities).
Conceptual or theoretical underpinnings of the intervention (i.e. study authors stated that the intervention was based on a theory or at least referred to a theory).
Barriers assessment (i.e. study authors stated that a barriers assessment was conducted and the intervention was designed to overcome identified barriers).
Number of components included in the intervention based on the EPOC taxonomy (when a barriers assessment was mentioned it was considered a component of the intervention).

Assessment of risk of bias in included studies

At least two review authors (including RA and ST) independently assessed the risk of bias in each included study using the criteria outlined in the suggested risk of bias criteria for EPOC reviews (EPOC 2017c) and the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) for ITS designs. For blinding, incomplete data, and baseline outcome measurement, we assessed the primary outcomes and secondary outcomes that were selected for inclusion in the 'Summary of findings' table (see below for details of selection process). Any disagreement was resolved through discussion with FL. Each risk of bias criterion was assessed as 'Low risk', 'High risk' or 'Unclear'. The 10 standard criteria as suggested for all randomized trials and CBA studies are listed below.

Random sequence generation (protection against selection bias)
Concealment of allocation (protection against selection bias)
Protection against contamination.
Blinded assessment (protection against detection bias)
Baseline outcome measurement
Patient baseline characteristics
Healthcare professional baseline characteristics
Selective reporting outcome
Incomplete data outcome
Other risk of bias.

Measures of treatment effect

We structured data analysis using statistical methods developed for EPOC by Grimshaw and colleagues (Grimshaw 2004). For each study, we reported results for categorical and continuous primary outcomes separately and in natural units. When included studies assessed SDM using an adaptation of the Control Preference Scale (Degner 1992), we dichotomized into SDM versus no SDM (Légaré 2012).

For categorical measures, we calculated the difference in risk between the intervention of interest and the control intervention. We calculated standardized mean difference (SMD) for continuous measures by dividing the mean score difference of the intervention and comparison groups in each study by the pooled estimate standard deviation for the two groups. When possible, for categorical and continuous outcomes, we constructed 95% confidence intervals (CIs) to compare groups before and after the intervention, according to the recommendations in RevMan 5. The absence of a '0' value in the CI indicated that the baselines differed or that the intervention had a positive effect compared to the control intervention or to usual care. When the baseline was different between the two groups, we used the size of the difference and its associated standard error to compare them. When there were not enough quantitative data available to make these calculations, we extracted a direct quote from the primary study on the effectiveness of the intervention and on confounding factors, if available. When no baseline was reported, we considered groups to be similar prior to the intervention.

For the analysis, we divided the studies into six comparison categories: 1) interventions targeting patients compared with usual care; 2) interventions targeting healthcare professionals compared with usual care; 3) interventions targeting both patients and healthcare professionals compared with usual care; 4) interventions targeting patients compared with other types of interventions targeting patients; 5) interventions targeting healthcare professionals compared with other types of interventions targeting healthcare professionals; 6) interventions targeting both patients and healthcare professionals compared with other types of interventions targeting both patients and healthcare professionals.

We performed a meta‐analysis if there were enough studies in each of the six comparison categories. When the study reported repeated measurements for an outcome for the same participants, only the measure closest in time to the consultation was kept in the meta‐analysis. When studies with more than two arms reported several comparisons of different outcomes or different interventions, we kept only the comparisons that most reduced the heterogeneity of the comparison group in the meta‐analysis. We considered a SMD of 0.2 as small, 0.5 as medium, and 0.8 as large (Cohen 1988).

Unit of analysis issues

We included cluster‐randomized trials in the analyses along with individually‐randomized trials. Comparisons that randomize or allocate clusters (groups of healthcare professionals or organizations) but do not account for clustering during the analysis have potential unit of analysis errors that can produce artificially significant P values and overly narrow CIs (Ukoumunne 1999). Therefore, when possible, we contacted study authors for missing information and attempted to re‐analyze studies with potential unit of analysis errors. When missing information was unavailable, we reported only the point estimate.

Assessment of heterogeneity

To explore heterogeneity, we designed tables that compared SMDs of the studies and their risk differences. We considered the following variables as potential sources of heterogeneity in the results of the included studies: type of intervention; characteristics of the intervention (e.g. duration); clinical setting (primary care versus specialized care); type of healthcare professional (physicians versus other healthcare professionals); level of training of healthcare professionals (e.g. in training versus in practice).

Data synthesis

We estimated a weighted intervention effect with 95% confidence intervals. For continuous measures, we used SMDs; for dichotomous outcomes, we calculated the risk difference. We analyzed all data with a random‐effects model because of the diverse nature of the studies being combined and then anticipated variability in the populations and interventions of the included studies. We summarized all of the results for the primary and selected secondary outcomes and rated the strength of evidence using GRADE (Andrews 2013), and then presented these results in the 'Summary of findings' tables (Higgins 2011). As the non‐randomized evidence has a high level of uncertainty and that there are few non‐randomized trials, we reported only the results of randomized trials in the Summary of findings' tables. For studies not included in the quantitative synthesis, we assessed how their results could have impacted the pooled estimate of the effect size regarding the direction of the effect (Appendix 6).

'Summary of findings' tables

We evaluated the certainty of the evidence according to the GRADE guidelines (Guyatt 2011) and the methods described in Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011). We assessed primary and selected secondary outcomes in all six comparison categories. For each outcome, we rated conclusions as follows:

high: this research provides a very good indication of the likely effect; the likelihood that the effect will be substantially different is low;
moderate: this research provides a good indication of the likely effect; the likelihood that the effect will be substantially different is moderate;
low: this research provides some indication of the likely effect; however, the likelihood that it will be substantially different is high;
very low: this research does not provide a reliable indication of the likely effect; the likelihood that the effect will be substantially different is very high).

From starting score of certainty of evidence according to the study design, we downgraded the rating if one or more of the five following criteria were present: study limitation, indirect evidence, inconsistency, imprecision of the observed effect and publication bias. A review author (RA) and a graduate student (AB) independently assessed the certainty of the evidence and reached consensus in collaboration with FL.

As the use of SDM is the only primary outcome of this review, we assessed this outcome using the GRADE approach and included it in the 'Summary of findings' tables. We used the method proposed in EPOC Worksheets (EPOC 2017d) to determine which secondary outcomes should be assessed and included in the 'Summary of findings' tables. First, the study co‐authors generated a list of relevant secondary outcomes for the review. Then we independently selected outcomes important enough to be included in the 'Summary of findings' tables by rating them on a 9‐point scale ranging from 1 (not important) to 9 (critical), and came to a consensus. Then we calculated the median of the scores we had attributed to each secondary outcome and agreed to include all that scored above 7. The selected secondary outcomes were: decision regret, health‐related quality of life (mental and physical), consultation length and cost.

Subgroup analysis and investigation of heterogeneity

Analysis was pre‐defined using a subgroup analysis approach, and we did not combine data from observer‐based outcome measures (OBOMs) with patient‐reported outcome measures (PROMs) as they measured different concepts. In addition, within each comparison category, we explored how individually‐randomized trials compared to cluster‐randomized trials regarding our primary outcome when applicable. We further investigated heterogeneity by exploring how the study design (cluster‐randomized trials versus individually‐randomized trials) affected statistical heterogeneity (Higgins 2011).

Sensitivity analysis

No sensitivity analysis were performed.

Results

Description of studies

See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of ongoing studies.

Results of the search

We identified 23,057 new potentially relevant records and excluded 18,072 during abstract screening. We retrieved 752 full‐text publications for more detailed screening and excluded another 613 records based on the identified inclusion criteria. We were not able to find the full text of 60 records for assessment . We included 87 studies in the review, 48 of which were newly identified for this update (Figure 1).

Included studies

This update search added 48 new studies (Adarkwah 2016; Almario 2016; Ampe 2017; Barton 2016; Branda 2013; Causarano 2014; Cooper 2013; Cox 2017; Coylewright 2016; Davison 2002; Eggly 2017; Epstein 2017; Feng 2013; Fiks 2015; Hamann 2011; Hamann 2014; Hamann 2017; Härter 2015; Hess 2016; Jouni 2017; Kennedy 2013; Koerner 2014; Köpke 2014; Korteland 2017; LeBlanc 2015a; LeBlanc 2015b; Maclachlan 2016; Maindal 2014; Maranda 2014; Mathers 2012; Perestelo‐Perez 2016; Pickett 2012; Rise 2012; Sanders 2017; Schroy 2016; Sheridan 2012; Sheridan 2014; Smallwood 2017; Tai‐Seale 2016; Thomson 2007; Tinsel 2013; van der Krieke 2013; van Roosmalen 2004; van Tol‐Geerdink 2016; Vestala 2013; Warner 2015; Wilkes 2013; Wolderslund 2017) to the 39 original studies for a total of 87 studies.

We identified 89 ongoing studies through trial registration databases, proceedings of conferences and protocols published in electronic databases (see Characteristics of ongoing studies). For two studies, we were unable to decide whether to include them or not because not enough information was available (See Characteristics of studies awaiting classification).

All the studies were randomized trials except for four: three non‐randomized controlled trials (Almario 2016; Barton 2016; Deinzer 2009) and a controlled before‐after study (CBA) (Ampe 2017). Among the randomized trials, 21 were cluster‐randomized trials (Branda 2013; Cox 2017; Elwyn 2004; Epstein 2017; Feng 2013; Hamann 2007; Haskard 2008; Kennedy 2013; Koerner 2014; LeBlanc 2015a; Légaré 2012; Loh 2007; Mathers 2012; O'Cathain 2002; Perestelo‐Perez 2016; Sanders 2017; Tai‐Seale 2016; Tinsel 2013; van Roosmalen 2004; Wetzels 2005; Wilkes 2013).

Settings and participants

Of the 87 included studies, 44 evaluated interventions targeting patients, 15 evaluated interventions targeting health professionals, and 28 targeted both patients and health professionals. The four most represented countries were the USA (37 studies), Germany (15 studies) and Canada (eight studies) and the Netherlands (eight studies). There were two studies by international collaborations. The level of care was primary care in 44 studies, specialized care in 36 studies and both primary and specialized care in one study. In six studies, the level of care was unclear. In 49 studies, the healthcare professionals involved were licensed; in 16 studies they were licensed and in training; in 22 studies their level of training was unclear. The three most frequent clinical conditions studied were cancer (22 studies), cardiovascular diseases (14 studies) and psychiatric conditions (11 studies) (see Characteristics of included studies).

Target categories

Interventions targeting patients (44 studies)

Most of the 44 studies of interventions targeting patients were conducted in Europe or the USA (36 studies). There was one study from Africa. Specialized care was the most frequent care setting (22 studies), and all but eight studies were conducted in and recruited patients in an ambulatory setting. Studies varied greatly regarding the number of patients involved, ranging from 26 (Lalonde 2006) to 913 (Hess 2016). Most of the studies did not report the number of healthcare professionals involved. The most common clinical conditions were oncologic (14 studies), cardiovascular (eight studies) and psychiatric (six studies).

Interventions targeting healthcare professionals (15 studies)

The majority of the studies of interventions targeting healthcare professionals were conducted in Europe or the USA (14 studies). There was one study by an international collaboration. The care setting was mainly primary care (11 studies), with most of the participants recruited in ambulatory care (11 studies). Among the 12 studies that used non‐simulated patients, the number of patients involved ranged from 298 (Cox 2017) to 10,070 (O'Cathain 2002). Two studies did not report the number of patients involved, and three did not report the number of healthcare professionals involved. The clinical condition was different in every study.

Interventions targeting both patients and healthcare professionals (28 studies)

The majority of the 28 studies of interventions targeting both patients and healthcare professionals were conducted in Europe or the USA (27 studies). There was one study by an international collaboration. The most common care setting was primary care (16 studies), with most of the participants recruited in ambulatory care (23 studies). Among the 26 studies that used non‐simulated patients, a total of 12,078 patients were enrolled, with a minimum of 60 (Fiks 2015) and a maximum of 4349 (Wolderslund 2017). Twenty‐five studies reported participating healthcare professionals, ranging from 10 per study (Bieber 2006) to 156 per study (Haskard 2008).The most common clinical condition was cancer (seven studies), followed by cardiovascular diseases (four studies), psychiatric conditions (four studies) and type‐2 diabetes (four studies).

Characteristics of interventions and comparisons

Some studies reported more than one comparison. For such studies, we extracted only data for the comparisons that corresponded to one or more of the six comparison categories in our review. In each category of comparison, no study was counted twice for the analysis. For details, see Characteristics of included studies.

Interventions targeting patients

Twenty‐four studies compared interventions targeting patients with usual care (Almario 2016; Cooper 2011; Deen 2012; Eggly 2017; Hamann 2014; Haskard 2008; Korteland 2017; Krist 2007; Landrey 2012; LeBlanc 2015a; LeBlanc 2015b; Maclachlan 2016; Maranda 2014; Murray 2001; van Peperstraten 2010; Perestelo‐Perez 2016; Pickett 2012; Sheridan 2014; Tai‐Seale 2016; van der Krieke 2013; van Tol‐Geerdink 2016; Vestala 2013; Vodermaier 2009; Wolderslund 2017). All but one study compared patient‐mediated interventions to usual care. Patient‐mediated interventions included decision aids, patient activation, question prompt lists and training for patients. The interventions were administered alone (single interventions) or in combination (multifaceted intervention).

Twenty‐eight studies presented comparisons of interventions targeting patients with other interventions targeting patients ( Adarkwah 2016; Barton 2016; Butow 2004; Causarano 2014; Davison 1997; Davison 2002; Deen 2012; Deschamps 2004; Dolan 2002; Eggly 2017; Hamann 2011; Hamann 2017; Jouni 2017; Kasper 2008; Köpke 2014; Krist 2007; Lalonde 2006; Montori 2011; Nannenga 2009; Raynes‐Greenow 2010; Schroy 2011; Schroy 2016; Smallwood 2017; Stiggelbout 2008; Street 1995; Thomson 2007; van Roosmalen 2004; Wolderslund 2017). Of these, 18 studies compared a single intervention (these included decision aid, consultation preparation package, empowerment sessions, brochure, training of patients in shared decision making (SDM), interactive‐4‐hour education program, literacy‐appropriate medication guide) to another single intervention (these included decision aid, booklets, information packages, patient activation, pamphlets, cognitive training, 4‐hour Multiple Sclerosis specific stress management program, existing medication guide); 10 studies compared a multifaceted intervention (these included decision aid and patient activation, decision aid and literacy‐appropriate medication guide, decision aid and risk assessment tool, question prompt list and assistance of a communication coach) to a single intervention (these included decision aid, question prompt list, literacy‐appropriate medication guide, existing medication guide) and three studies compared a multifaceted intervention (these included decision aid and information booklet about immunotherapy, conventional risk and genetic risk information and decision aid) to another multifaceted intervention (these included decision aid and standard information package, conventional risk information and decision aid).

Four studies reported basing their intervention on a barriers assessment (Hamann 2011; Jouni 2017; Korteland 2017; van Peperstraten 2010).

Interventions targeting healthcare professionals

Fifteen studies compared interventions targeting the healthcare professionals with usual care ( Ampe 2017; Bernhard 2011; Cooper 2011; Cox 2017; Fossli 2011; Kennedy 2013; Koerner 2014; LeBlanc 2015b; Légaré 2012; O'Cathain 2002; Sanders 2017; Shepherd 2011; Stacey 2006; Tinsel 2013; Wilkes 2013). Of these, seven studies compared a single intervention (educational meeting, distribution of educational material, educational outreach visit, and reminder) to usual care and eight studies compared a multifaceted intervention (educational meeting and audit and feed‐back; educational meeting and distribution of educational material; educational meeting and audit and feedback and distribution of educational material; distribution of educational materials and educational meeting and audit and feedback and barriers assessment) to usual care.

Two studies compared an intervention targeting the healthcare professional (educational meeting, reminder) with one targeting the patient (decision aid, patient coaching by community health workers) (Cooper 2011; LeBlanc 2015b).

Three studies compared interventions targeting the healthcare professional with other interventions targeting the healthcare professional (Elwyn 2004; Feng 2013; Krones 2008 (ARRIBA‐Herz)). Of these, one study compared a multifaceted intervention (educational meeting and audit and feedback focusing on SDM skills) to another multifaceted intervention (educational meetings and audit and feedback focusing on risk communication skills), one study compared a single intervention (distribution of educational material) to another single intervention (distribution of educational material), and one study compared a multifaceted intervention (educational meeting, audit and feedback, distribution of educational material, and an educational outreach component) to a single intervention (educational meeting).

Four studies reported the performance of a barriers assessment and based their interventions on the identified barriers (Ampe 2017; Bernhard 2011; Murray 2010; Stacey 2006).

Interventions targeting both patients and healthcare professionals

Seventeen studies compared an intervention targeting patients and healthcare professionals with usual care (Branda 2013; Cooper 2011; Coylewright 2016; Epstein 2017; Hamann 2007; Härter 2015; Haskard 2008; Hess 2012; Hess 2016; Leighl 2011; Loh 2007; Mathers 2012; Murray 2010; Rise 2012; Tai‐Seale 2016; Wetzels 2005; Wilkes 2013). Of these, 11 studies presented interventions that used educational meetings and patient‐mediated interventions; one study presented a patient‐mediated intervention with educational outreach visits; one study presented an arm with an intervention that used a combination of a patient‐mediated intervention, distribution of educational material and educational meetings (Haskard 2008); one study presented an arm with a patient‐mediated intervention and a distribution of educational material (Wilkes 2013); one study presented an arm with a patient‐mediated intervention and a reminder; one study presented an arm with a combination of educational meeting, audit and feedback, distribution of educational material, educational outreach visit and barriers assessment; one study presented interventions that used educational meetings, patient‐mediated interventions and distribution of educational material.

Seven studies compared interventions targeting both patients and healthcare professionals with interventions targeting patients alone (Bieber 2006; Cooper 2011; Deinzer 2009; Mullan 2009; Sheridan 2012; Tai‐Seale 2016; Warner 2015). Of these, six studies compared educational meetings and patient‐mediated interventions with patient‐mediated interventions alone; one study compared interventions that used educational meetings, patient‐mediated interventions and distribution of educational material with patient‐mediated interventions alone.

Five studies compared interventions targeting both patients and healthcare professionals with interventions targeting healthcare professionals alone (Cooper 2011; Feng 2013; Fiks 2015; Maindal 2014; Roter 2012). Of these, two studies compared patient‐mediated interventions and the distribution of educational materials with the distribution of educational materials alone; two study compared educational meetings and patient‐mediated interventions with educational meetings alone; and one study compared patient‐mediated interventions and reminders with reminders alone.

Three studies compared an intervention targeting both patients and healthcare professionals with another intervention targeting both patients and healthcare professionals (Cooper 2013; Myers 2011; Tai‐Seale 2016). Of these, one study compared a multifaceted intervention including a patient‐mediated intervention, educational outreach visit, distribution of educational material and audit and feedback with another multifaceted intervention including a patient‐mediated intervention, educational outreach visit and distribution of educational material; one study compared a multifaceted intervention including a patient‐mediated intervention and reminders with another multifaceted intervention also including a patient‐mediated intervention and reminders; and one study compared a multifaceted intervention including educational meetings, patient‐mediated interventions and distribution of educational material with another multifaceted intervention including educational meetings, patient‐mediated interventions and distribution of educational material.

Three studies reported the performance of a barriers assessment and based its interventions on identified barriers (Cooper 2013; Coylewright 2016; Epstein 2017).

Conceptual framework

Thirty‐one out of the 87 studies included in this update used or referred to a conceptual framework. Six studies referred to the Ottawa Decision Support Framework (Causarano 2014; Murray 2010; Raynes‐Greenow 2010; Schroy 2011; Stacey 2006; van Peperstraten 2010); two studies referred to the RE‐AIM framework (Reach, Effectiveness, Adoption, Implementation, Maintenance) (Branda 2013; LeBlanc 2015a); two studies referred to the Four Habits model (Fossli 2011; Tai‐Seale 2016); and two studies referred to the UKMRC framework (Medical Research Council guidance) (Köpke 2014; Mathers 2012). Five studies used a conceptual model but did not describe it (Bernhard 2011; Butow 2004; Hamann 2011; Hamann 2014; Loh 2007). The 14 other studies each referred to a different conceptual model, including the 4E Model (Haskard 2008); the Empowerment Model by Conger and Kanungo (Davison 1997); the LEAPS (Listen, Educate, Assess, Partner and Support) framework (Roter 2012); the Markov Model (Stiggelbout 2008); the Model of Interpersonal Interaction (Elwyn 2004); the SWOT analysis (Strengths, Weaknesses, Opportunities and Threats) (Wetzels 2005); the Theory of Planned Behaviour (Légaré 2012); the Framework for Accountable Decision‐Making (FADM) (Maranda 2014); the Integrative Theory, Protection Motivation Theory and Self‐Determination Theory (Sheridan 2014); the WISE (Whole System Informing Self‐management Engagement) Model (Kennedy 2013); the NIH PROMIS framework (Almario 2016); the three‐step model for SDM (Ampe 2017); the Systems Engineering Initiative for Patient Safety (SEIPS) Model (Cox 2017); and the Bandura’s social cognitive theory of self‐efficacy (Maclachlan 2016) .

Outcome measures

Primary outcome (use of shared decision making)

Of 87 studies, 59 reported patient‐reported outcome measures (PROMs), 19 reported observer‐based outcome measures (OBOMs), and nine reported both OBOMs and PROMs. PROMs were used to measure patient or family caregiver’s self‐reported experiences of participating in the decision‐making process when facing health treatment or screening decisions. Among 68 studies using PROMs, 30 unique scales or subscales were used to measure the use of SDM from a patient'S perspective. In 29 studies, PROMs were the “perceived level of control in decision making” or “role assumed during the consultation” (adaptation of the Control Preference Scale (Degner 1992). Two other PROMs were the SDMQ‐9 (Kriston 2010), and the Patient Activation Measure (PAM) (Hibbard 2004; Hibbard 2005). Twenty‐seven additional unique scales or subscales were used in the studies analyzed. For more details, see Characteristics of included studies. Among the 28 studies that used OBOMs, 16 unique scales or subscales were used to measure the use of SDM from an observer‐based perspective. Study authors reporting observer‐based outcomes used the OPTION scale (Elwyn 2003) in 15 studies, and OPTION‐5 (Barr 2015) in two studies. Fourteen additional unique scales or subscales were used in the studies analyzed. For more details, see Characteristics of included studies.

Secondary outcomes

Study authors reported most on affective‐cognitive outcomes, followed by health outcomes, behavioral outcomes and process outcomes. Adverse events were seldom reported. None of the studies assessed distress or equity.

Excluded studies

After full‐text assessment of articles for eligibility, we initially excluded 613 articles. The reasons for exclusion were related to the design of the study, the type of participants, the type of outcome measure, the content of the intervention, and the language. Main reasons for exclusion of the 39 studies listed in Excluded studies are presented under Characteristics of excluded studies.

Risk of bias in included studies

Further details on the ratings and rationale for risk of bias are in the 'Risk of bias' tables in the Characteristics of included studies tables and displayed in Figure 2 and Figure 3. The 'Risk of bias' assessment reported there was based on the primary outcome only.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

'Risk of bias' summary for each included study.

Allocation

Allocation concealment was rated as being at low risk of bias in 44 of 87 studies (51%), unclear risk of bias in 38 studies (44%) and high risk of bias in five studies (6%) .

Blinding

For assessing risk of detection bias in the 28 studies that used observer‐based outcome measures (OBOMs), blinding was rated as being at low risk of bias in 11 studies (39%), unclear risk in 16 studies (57%) and high risk in one study (4%). In the 68 studies that used patient‐reported outcome measures (PROMs), blinding was rated as being at low risk of bias in 14 studies (21%), unclear risk in 36 studies (53%) and high risk in 18 studies (26%).

Incomplete outcome data

Of the 28 studies that used OBOMs, incomplete outcome data were rated as being at low risk of bias in 15 studies (53%), unclear risk in eight studies (29%) and high risk in five studies (18%). Of the 68 studies that used PROMs, incomplete outcome data were rated as being at low risk of bias in 27 studies (40%), unclear risk in 29 studies (42%) and high risk in 12 studies (17%).

Selective reporting

For assessing risk of reporting bias, selective outcome reporting was rated as being at low risk of bias in 25 of 87 studies (29%), unclear risk in 44 studies (50%) and high risk in 18 studies (21%).

Other potential sources of bias

Among the 87 studies, in 79 studies other risks of bias were rated as low (91%), in five studies they were unclear (6%) and in three studies they were high (3%).

Effects of interventions

See: Table 1; Table 2; Table 3

Summary of findings for the main comparison. Interventions targeting patients compared to usual care or interventions of the same type for shared decision making.

Interventions targeting patients compared to usual care or to interventions of the same type for shared decision making
Patient or population: patients, including healthcare consumers and simulated patients  Settings: Australia, Canada, Germany, Namibia, Spain, Sweden, the Netherlands, UK, USA  Interventions: interventions designed to improve shared decision making among healthcare professionals that target patients (for example, patient‐mediated interventions)  Comparison: usual care or interventions of the same type
Outcomes	*Illustrative comparative risks (95% CI)**		Risk difference (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	[control]	[experimental]
a‐ Intervention targeting patients compared to usual care
Shared decision making (observer based outcome measure (OBOM), continuous measures)  (follow‐up: up to 6 months)	‐	SMD 0.54 higher  (0.13 lower to 1.22 higher)	‐	424  (4 randomized trials)	⨁◯◯◯  VERY LOW ^a,b,c,d	Scales are: OPTION (0‐100) and RIAS. Higher score indicates more shared decision making use. One study was not included in the quantitative synthesis and was consistent with the pooled result
Shared decision making (patient reported outcome measure (PROM), continuous measures)  (follow‐up: up to 3 years)	‐	SMD 0.32 higher  (0.16 higher to 0.48 higher)	‐	1386  (9 randomized)	⨁◯◯◯  VERY LOW ^a,b,c	Scales are: Patient activation measure (0‐100), patient self‐advocacy (1‐5), COMRADE (0‐100), decision evaluation scale (1‐5), clinicians’ participatory decision making (1‐5), satisfaction with decision making process (0‐100), CollaboRATE (0‐100), patient role in treatment decision (1‐5). Higher score indicates more shared decision making use. One study was not included in the quantitative synthesis. It is unlikely that it would change the direction of the effect size estimate given that its sample size was not very large.
Shared decision making (PROM), dichotomous measures  (follow‐up : up to 3 months)	Study population		‐0.09 (‐0.19 to 0.01)	754  (6 randomized trials)	⨁◯◯◯  VERY LOW ^a,c,f	Three studies were not included in the quantitative synthesis.The first study did not support the pooled result but given that the pooled estimate of the effect size is in favor of the control group, it is likely that adding that study would move the pooled estimate of the effect size toward a null effect. The second study did not support the pooled result but given its very large sample size, it is likely that adding this study would move the pooled estimate of the effect size toward a positive effect. The third study supported the pooled result toward the null effect.
	56 per 100	46 per 100
	Low risk population
	33 per 100^e	33 per 100
	High risk population
	88 per 100^e	60 per 100
Decision regret  (follow‐up : 6 months)	‐	SMD 0.10 lower  (0.39 lower to 0.19 higher)	‐	212  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,h	Decision regret scale (0‐100). Higher score indicates more regret after decision
Health‐related quality of life (physical)  (follow‐up: 3 months post‐operatively)	‐	SMD 0.00  (0.36 lower to 0.36 higher)	‐	116  (1 randomized trial)	⨁◯◯◯  VERY LOW ^c,g,h	Physical component scale of SF‐36 (0‐100). Higher score indicate better quality of life.
Health‐related quality of life (mental)  (follow‐up: 3 months post‐operatively)	‐	SMD 0.10 higher  (0.26 lower to 0.46 higher)	‐	116  (1 randomized trial)	⨁◯◯◯  VERY LOW ^c,g,h	Mental component scale of SF‐36 (0‐100). Higher score indicate better quality of life.
Consultation length (minutes)	‐	SMD 0.10 higher  (0.39 lower to 0.58 higher)	‐	224  (2 randomized trials)	⨁◯◯◯  VERY LOW ^c,f,g,h
Cost (£)	‐	SMD 0.82 higher  (0.42 higher to 1.22 higher)	‐	105  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,h
b‐ Intervention targeting patients compared to intervention of the same type
Shared decision making (OBOM, continuous)  (post‐visit)	‐	SMD 0.88 higher  (0.39 higher to 1.37 higher)	‐	271  (3 randomized trials)	⨁◯◯◯  VERY LOW ^b,c,g,h	OPTION scale (0‐100). Higher score indicates more shared decision making use.    Decision aid study increase the use of shared decision making compared to booklet or pamphlet
Shared decision making (PROM, continuous)  (follow‐up: up to 6 months)	‐	SMD 0.03 higher  (0.18 lower to 0.24 higher)	‐	1906  (11 randomized trials)	⨁◯◯◯  VERY LOW ^b,c,g,h	Scales are: Decision making subscale of the Modified Perceived involvement in care scale (4‐20), Patient Activation Measure (0‐100), 1‐item question on “who makes decisions about medical treatment” (1‐5), Satisfaction With Decision Making Process scale (12‐60), Problem‐Solving Decision‐Making Scale (1‐5), SDM‐Q9 (0‐100), patient role in treatment decision (1‐5), SDM‐Q (0‐11), Patient‐reported shared decision making (0‐4). Higher score indicates more shared decision making use. Two studies were not included in the quantitative synthesis but supported the pooled results.
Shared decision making (PROM, categorical or dichotomous)  (follow‐up : up to 6 weeks)	Study population		0.03 (‐0.02 to 0.08)	2272  (10 randomized trials)	⨁◯◯◯  VERY LOW ^a,c,f	Three studies were not included in the quantitative synthesis.Two of them were consistent with the pooled results, but the third reported an increase in the use of shared decision making for the intervention group.
	38 per 100	40 per 100
	Low risk population
	18 per 100^e	22 per 100
	High risk population
	73 per 100^e	52 per 100
Decision regret	‐	‐	‐	‐	‐	No data available for this outcome
Health‐related quality of life (physical)	‐	‐	‐	‐	‐	No data available for this outcome
Health‐related quality of life (mental)	‐	‐	‐	‐	‐	No data available for this outcome
Consultation length (minutes)	‐	SMD 0.65 lower  (1.29 lower to 0.00 )	‐	39  (1 randomized trial)	⨁◯◯◯  VERY LOW ^c,g,h
Cost	‐	‐	‐	‐	‐	No data available for this outcome
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Interventions targeting healthcare professionals compared to usual care or interventions of the same type for shared decision making
Patient or population: healthcare professionals responsible for patient care  Settings: Australia, Austria, Belgium, Canada, Germany, the Netherlands, New Zealand, Norway, Switzerland, USA, UK  Interventions: interventions designed to improve shared decision making among healthcare professionals that target healthcare professionals (for example, distribution of printed educational material, educational meetings, audit and feedback, reminders and educational outreach visits)  Comparison: usual care or interventions of the same type
Outcomes	*Illustrative comparative risks (95% CI)**		Risk difference (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	[control]	[experimental]
a‐ Intervention targeting healthcare professionals compared to usual care
Shared decision making (OBOM, continuous)  (follow‐up: up to 3 months post‐intervention)	‐	SMD 0.70 higher  (0.21 higher to 1.19 higher)	‐	479  (6 randomized trials)	⨁◯◯◯  VERY LOW ^a,b,c,d	Scales are: Fours Habits Coding Scheme (23‐115), OPTION (0‐100), Decision Support Analysis Tool (0‐100), Control Preference Scale (0‐4), and Family engagement (number of utterances or decision‐making events that families engaged). Higher score indicates more shared decision making use. Two studies were not included in the quantitative synthesis. The first study and one sub‐sample of the second study were consistent with the pooled result. The other sub‐sample of the second study reported no difference between the study groups.
Shared decision making (PROM, continuous)  (follow‐up : up to 12 months)	‐	SMD 0.03 higher  (0.15 lower to 0.20 higher)	‐	5772  (5 randomized trials)	⨁◯◯◯  VERY LOW ^a,b,c,d	Scales are: Physicians’ participatory decision making style (0‐4), short‐form healthcare climate questionnaire (0‐100), SDM‐Q9 (0‐100), and Overall PSA SDM perception (5‐20). Higher score indicates more shared decision making use. One study was not included in the quantitative synthesis and reported an increase in the use of shared decision making for the intervention group.
Shared decision making (PROM, categorical or dichotomous)  (follow‐up: up to 8 weeks after delivery of pregnant women)	Study population		0.01 (‐0.03 to 0.06)	6303  (2 randomized trials)	⨁◯◯◯  VERY LOW ^a,b,c	One study was not included in the quantitative synthesis and was consistent with the pooled results.
	21 per 100	22 per 100
	Low risk population
	19 per 100^e	17 per 100
	High risk population
	36 per 100^e	45 per 100
Decision regret  (follow‐up: 2 weeks)	‐	SMD 0.29 higher  (0.07 higher to 0.51 higher)	‐	326  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d	Decision regret scale (0‐100). Higher score indicates more regret after decision.  The slight effect observed on patient decisional regret was not clinically significant.
Health‐related quality of life (physical)  (follow‐up: 2 weeks)	‐	SMD 0.16 higher  (0.05 lower to 0.36 higher)	‐	359  (1 randomized trial)	⨁⨁◯◯  LOW ^a,c	Scale are : Physical scale of SF‐12 (0‐100) and SF12v2 (0‐100). Higher score indicate better quality of life.
Health‐related quality of life (mental)  (follow‐up: 2 weeks)	‐	SMD 0.28 higher  (0.07 to 0.49 higher)	‐	359  (1 randomized trial)	⨁⨁◯◯  LOW ^a,c	Scale are : Mental scale of SF‐12 (0‐100) and SF12v2 (0‐100). Higher score indicate better quality of life.
Consultation length (minutes)	‐	SMD 0.51 higher  (0.21 higher to 0.81 higher)	‐	175  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d
Cost	‐	‐	‐	‐	‐	No data available for this outcome
b‐ Intervention targeting healthcare professionals compared to intervention of the same type
Shared decision making (OBOM, continuous)  (post‐visit)	‐	SMD 0.30 lower  (1.19 lower to 0.59 higher)	‐	20  (1 randomized trial)	⨁◯◯◯  VERY LOW ^c,f	OPTION scale (0‐100). Higher score indicates more shared decision making use.    Intervention group included: education meeting (in shared decision‐making skill) + audit and feed‐back. Control group included: educational meeting (in risk communication skills) + audit and feed‐back. One study was not included in the quantitative synthesis but reported significant positive results.
Shared decision making (PROM, continuous)  (follow‐up: up to 4 weeks)	‐	SMD 0.24 higher  (0.10 lower to 0.58 higher)	‐	1459  (2 randomized trials)	⨁◯◯◯  VERY LOW ^a,b,c,d	Scales are: COMRADE (0‐100) and SDM‐Q. Higher score indicates more shared decision making use.    In one study multifaceted intervention like educational meeting, audit and feedback, educational material and educational outreach visit increase the use of shared decision making compared to educational meeting alone (control group). In the second study, no differences were found between intervention and control groups.
Decision regret	‐	‐	‐	‐	‐	No data available for this outcome
Health‐related quality of life (physical)	‐	‐	‐	‐	‐	No data available for this outcome
Health‐related quality of life (mental)	‐	‐	‐	‐	‐	No data available for this outcome
Consultation length	‐	‐	‐	‐	‐	No data available for this outcome
Cost	‐	‐	‐	‐	‐	No data available for this outcome
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Interventions targeting healthcare professionals and patients compared to usual care or interventions of the same type for shared decision making
Patient or population: healthcare professionals and patients  Settings: Australia, Canada, Denmark, Germany, Norway, the Netherlands, UK, USA  Interventions: intervention designed to improve shared decision making among healthcare professionals that target both healthcare professionals and patients (for example, a patient‐mediated intervention combined with an intervention targeting healthcare professionals)  Comparison: usual care or interventions of the same type
Outcomes	*Illustrative comparative risks (95% CI)**		Risk difference (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	[control]	[experimental]
a‐ Interventions targeting healthcare professionals and patients compared to usual care
Shared decision making (OBOM, continuous)  (follow‐up : up to 3 months)	‐	SMD 1.10 higher  (0.42 higher to 1.79 higher)	‐	1270  (6 randomized trials)	⨁◯◯◯  VERY LOW ^a,b,c,d,e	OPTION scale (0‐100). Higher score indicates more shared decision making use.
Shared decision making (PROM, continuous)  (follow‐up: up to 6 weeks)	‐	SMD 0.13 higher  (0.02 lower to 0.28 higher)	‐	1479  (7 randomized trials)	⨁◯◯◯  VERY LOW ^a,c,f	Scales are: Physicians’ participatory decision making style (0‐4), SDM‐Q9 (0‐100), COMRADE (0‐100), Patient activation measure (0‐100), Overall PSA SDM perception (5‐20), CollaboRATE (0‐100), Healthcare Climate Questionnaire (0‐100). Higher score indicates more shared decision making use. Two studies were not included in the quantitative synthesis. One study reported an increase in the use of shared decision making for the intervention group and the second study did not report any differences between the study groups.
Shared decision making (PROM, categorical or dichotomous)  (post‐visit)	Study population		‐0.01 (‐0.20 to 0.19)	266  (2 randomized trials)	⨁◯◯◯  VERY LOW ^a,c,d,f	One study was not included in the quantitative synthesis and the results were consistent with the pooled results.
	41 per 100	36 per 100
	Low risk population
	36 per 100^g	27 per 100
	High risk population
	48 per 100^g	58 per 100
Decision regret  (follow‐up : 3 months)	‐	SMD 0.13 higher  (0.08 lower to 0.33 higher)	‐	369  (1 randomized trial)	⨁⨁◯◯  LOW ^a,c	Decision regret scale (0‐100). Higher score indicates more regret after decision
Health‐related quality of life (physical) (follow‐up: 6 weeks)	‐	SMD 0.08 higher  (0.37 lower to 0.54 higher)	‐	75  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d	Scale are : Physical scale of SF‐12 (0‐100) and SF12v2 (0‐100). Higher score indicate better quality of life.
Health‐related quality of life (mental) (follow‐up: 6 weeks)	‐	SMD 0.01 higher  (0.44 lower to 0.46 higher)	‐	75  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d	Scale are : Mental scale of SF‐12 (0‐100) and SF12v2 (0‐100). Higher score indicate better quality of life.
Consultation length (minutes)	‐	SMD 3.72 higher  (3.44 higher to 4.01 higher)	‐	536  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d
Cost	‐	‐	‐	‐	‐	No data available for this outcome
b‐ Interventions targeting healthcare professionals and patients compared to intervention of the same type
Shared decision making (OBOM, continuous)  (post‐visit)	‐	SMD 0.29 lower  (1.17 lower to 0.6 higher)	‐	20  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d	OPTION scale (0‐100). Higher score indicates more shared decision making use.
Shared decision making (OBOM, categorical or dichotomous)  (post‐visit)	Study population		‐0.04 (‐0.13 to 0.04)	134  (1 randomized trial)	⨁◯◯◯  VERY LOW ^c,d,h
	8 per 100	4 per 100
	Low risk population
	N/A	N/A
	High risk population
	N/A	N/A
Shared decision making (PROM, continuous)  (post‐visit)	‐	SMD 0.00  (0.32 lower to 0.32 higher)	‐	150  (1 randomized trial)	⨁◯◯◯  VERY LOW ^a,c,d	CollaboRATE (0‐100). Higher score indicates more shared decision making use.
Decision regret	‐	‐	‐	‐	‐	No data available for this outcome
Health‐related quality of life (physical)	‐	‐	‐	‐	‐	No data available for this outcome
Health‐related quality of life (mental)	‐	‐	‐	‐	‐	No data available for this outcome
Consultation length	‐	‐	‐	‐	‐	No data available for this outcome
Cost	‐	‐	‐	‐	‐	No data available for this outcome
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Observer‐based outcome measure ‐ Continous data								Meta‐analysis
Study	Intervention	Control	Outcome	N	SMD	SMD (95% CI)	I²	Meta‐analysis
LeBlanc 2015a	Patient‐mediated intervention (decision aid)	Usual care	OPTION	96	0.93 (0.50 to 1.36)	0.54 (‐0.13 to 1.22)	84%	X
Maclachlan 2016	Patient‐mediated intervention (educational meeting for patient)	Usual care	RIAS for patients (Patient activation and engagement)	289	‐0.05 (‐0.28 to 0.18)			X
LeBlanc 2015b	Patient‐mediated intervention (decision aid)	Usual care	OPTION (/100)	19	1.23 (0.17 to 2.29)			x
Tai‐Seale 2016	Patient‐mediated intervention (one‐page ASK handout)	Usual care	OPTION5 (/100)	20	0.32 (‐0.56 to 1.21)			x
Observer‐based outcome measure ‐ Categorical data								Meta‐analysis
Study	Intervention	Control	Outcome	N	RD	RD (95% CI)	I²	Meta‐analysis
No study
Observer‐based outcome measure ‐ Qualitative statement								Meta‐analysis
Study	Intervention	Control	Outcome	Direct quote				Meta‐analysis
Hamann 2014	Patient‐mediated intervention (question prompt sheet)	Usual care	Who made the decision in today's consultation	Intervention: Median (2) ‐ Range (1‐4); Control: Median (2) ‐ Range (1‐4) ‐ No difference				na
Haskard 2008	Patient‐mediated intervention	Usual care	Physician informative and participatory	Unit of error analysis				na
Haskard 2008	Patient‐mediated intervention	Usual care	Patient active	Unit of error analysis				na
Haskard 2008	Patient‐mediated intervention	Usual care	Physician‐patient interaction	Unit of error analysis				na
Patient‐reported outcome measure ‐ Continous data								Meta‐analysis
Study	Intervention	Control	Outcome	N	SMD	SMD (95% CI)	I²	Meta‐analysis
Deen 2012	Patient‐mediated intervention (decision aid)	Usual care	Patient Activation Measure (PAM)	138	‐0.07 (‐0.40 to 0.26)	0.28 (0.13 to 0.44)	57%
Deen 2012	Patient‐mediated intervention (Patient Activation)	Usual care	Patient Activation Measure (PAM)	142	0.09 (‐0.24 to 0.41)			X
Deen 2012	Patient‐mediated intervention (decision aid + Patient Activation)	Usual care	Patient Activation Measure (PAM)	137	0.04 (‐0.29 to 0.38)
Maranda 2014	Patient‐mediated intervention (patient activated intervention)	Usual care	Patient Activation Measure (PAM)	132	0.12 (‐0.22 to 0.47)			x
Pickett 2012	Patient‐mediated intervention (training for patients)	Usual care	Patient self‐advocacy (immediately after)	Pre: 428; Post: 342	0.20 (‐0.01 to 0.41)			x
Pickett 2012	Patient‐mediated intervention (training for patients)	Usual care	Patient self‐advocacy (6 months after)	Pre: 428; Post: 318	0.02 (‐0.20 to 0.24)
van der Krieke 2013	Patient‐mediated intervention (web‐based information and decision tool)	Usual care	COMRADE (communication)	73	0.94 (0.46 to 1.43)			x
van Peperstraten 2010	Patient‐mediated intervention (decision aid + support call)	Usual care	Decision Evaluation scale	252	0.53 (0.28 to 0.78)			x
Cooper 2011	Patient‐mediated intervention	Usual care	Participatory Decision making (PDM)	83	0.21 (‐0.22 to 0.64)			x
Perestelo‐Perez 2016	Patient‐mediated intervention (decision aid)	Usual care	Satisfaction with Decision making process (SDMP)	153	0.50 (0.18 to 0.82)			X
Tai‐Seale 2016	Patient‐mediated intervention (one‐page ASK handout)	Usual care	CollaboRATE (%)	150	0.31 (‐0.01 to 0.63)			X
Almario 2016	Patient‐mediated intervention (GI PROMIS)	Usual care	SDMQ‐9	303	0.02 (‐0.22 to 0.25)			X
Eggly 2017	Patient‐mediated intervention (QPL‐only)	Usual care	Patient role in treatment decision	59	0.05 (‐0.46 to 0.56)			x
Eggly 2017	Patient‐mediated intervention (QPL‐plus‐Coach)	Usual care	Patient role in treatment decision	55	‐0.14 (‐0.67 to 0.39)
Patient‐reported outcome measure ‐ Categorical data								Meta‐analysis
Study	Intervention	Control	Outcome	N	RD	RD (95% CI)	I²	Meta‐analysis
Krist 2007	Patient‐mediated intervention (decision aid brochure)	Usual care	Modified Control Preference Scale	237	0.00 (‐0.14 to 0.14)	‐0.09 (‐0.19 to 0.01)	48%	x
Krist 2007	Patient‐mediated intervention (decision aid web)	Usual care	Modified Control Preference Scale	261	‐0.01 (‐0.14 to 0.13)
Landrey 2012	Patient‐mediated intervention (mailed flyer)	Usual care	Modified Control Preference Scale	152	‐0.03 (‐0.19 to 0.12)			x
Murray 2001	Patient‐mediated intervention (decision aid)	Usual care	Modified Control Preference Scale	105	‐0.28 (‐0.44 to ‐0.12)			x
Sheridan 2014	Patient‐mediated intervention (decision aid)	Usual care	Shared decision	114	‐0.17 (‐0.35 to 0.002)			x
Vestala 2013	Patient‐mediated intervention (patient participation in nursing documentation)	Usual care	Modified Control Preference Scale	39	0.00 (‐0.30 to 0.30)			x
Vodermaier 2009	Patient‐mediated intervention (decision aid)	Usual care	Modified Control Preference Scale	107	‐0.01 (‐0.19 to 0.17)			x
Patient‐reported outcome measure ‐ Qualitative statement								Meta‐analysis
Study	Intervention	Control	Outcome	Direct quote				Meta‐analysis
Hamann 2014	Patient‐mediated intervention (question prompt sheet)	Usual care	Who made the decision in today's consultation	Intervention: Median (3) ‐ Range (1‐5); Control: Median (3) ‐ Range (1‐5) ‐ No difference				na
van der Krieke 2013	Patient‐mediated intervention (web‐based information and decision tool)	Usual care	COMRADE (confidence)	COMRADE confidence in decision: F(1.67) = 0.086. P=0.77. see also Table 3				na
Vodermaier 2009	Patient‐mediated intervention (decision aid)	Usual care	Man‐Son‐Hing Instrument	No data				na
van Tol‐Geerdink 2016	Patient‐mediated intervention (decision aid)	Usual care	Patient participation	"At t2, 95% of the patients in the decision aid group indicated that they actually had been involved in the decision, compared to 83% in the usual care group (P = 0.002). As such, the decision aid had the expected effect on patient participation."Page 466				na
Wolderslund 2017	Patient‐mediated intervention (question prompt list) + other (digital audio recording)	Usual care	Involvement in decision making	"Intention‐to‐treat analyses of the participants’ perception of the consultation showed that on average, 4.5% more patients (range: 4.0‐5.5), regardless of intervention group, rated their satisfaction with the treatment, confidence in and relationship with the health professional, and involvement in decision making in the highest reply category compared with patients in the control group. This increase was significant (p = 0.001) except for decision making (p = 0.044)." Page 247				na
Wolderslund 2017	Other (digital audio recording)	Usual care	Involvement in decision making	"Intention‐to‐treat analyses of the participants’ perception of the consultation showed that on average, 4.5% more patients (range: 4.0‐5.5), regardless of intervention group, rated their satisfaction with the treatment, confidence in and relationship with the health professional, and involvement in decision making in the highest reply category compared with patients in the control group. This increase was significant (p = 0.001) except for decision making (p = 0.044)." Page 247				na
Korteland 2017	Patient‐mediated intervention (decision aid)	Usual care	Involvement in decision making	There was no difference between intervention and control group regarding those who "totally agree" that the doctor has involved them in the decision (37.3% vs 39.1%; fig 2)				na

Continous Data								Meta‐analysis
Study	Intervention	Intervention	Outcome	N	SMD	SMD (95% CI)	I²	Meta‐analysis
Murray 2010	Educational meeting + audit and feedback + distribution of educational material + educational outreach visit + barriers assessement	Usual care	Knowledge	70	0.55 (0.07 to 1.03)	0.26 (‐0.16 to 0.69)	68%	x
Tinsel 2013	Distribution of educational material + educational meeting	Usual care	Knowledge (time 1)	Pre: 1083; Post: 899	0.10 (‐0.03 to 0.23)			x
Tinsel 2013	Distribution of educational material + educational meeting	Usual care	Knowledge (time 2)	Pre: 1083; Post: 794	0.09 (‐0.05 to 0.23)
Tinsel 2013	Distribution of educational material + educational meeting	Usual care	Knowledge (time 3)	Pre: 1083; Post: 714	0.17 (0.02 to 0.31)
Wilkes 2013	Distribution of educational material	Usual care	Satisfaction with consultation	479	0.00 (‐0.18 to 0.18)	0.00 (‐0.18 to 0.18)	na	na
Légaré 2012	Educational meeting and distribution of Educational material	Usual care	Decision regret (2 weeks)	Pre: 329  Post: 326	0.29 (0.07 to 0.51)	0.29 (0.07 to 0.51)	na	na
Kennedy 2013	Educational meeting	Usual care	Self‐efficacy (6 months)	Pre: 5599; Post: 4475	‐0.03 (‐0.09 to 0.03)	‐0.03 (‐0.09 to 0.03)	na	na
Kennedy 2013	Educational meeting	Usual care	Self‐efficacy (12 months)	Pre: 5599; Post: 4005	‐0.04 (‐0.10 to 0.03)	‐0.04 (‐0.10 to 0.03)	na	na
Tinsel 2013	Distribution of educational material + educational meeting	Usual care	Adherence to medication (time 1)	Pre: 1070; Post: 827	‐0.08 (‐0.21 to 0.06)	‐0.08 (‐0.21 to 0.06)	na	na
Tinsel 2013	Distribution of educational material + educational meeting	Usual care	Adherence to medication (time 2)	Pre: 1070; Post: 744	‐0.01 (‐0.16 to 0.13)	‐0.01 (‐0.16 to 0.13)	na	na
Tinsel 2013	Distribution of educational material + educational meeting	Usual care	Adherence to medication (time 3)	Pre: 1070; Post: 675	0.10 (‐0.05 to 0.25)	0.10 (‐0.05 to 0.25)	na	na
Kennedy 2013	Educational meeting	Usual care	General health	Pre: 5599; Post: 4056	0.02 (‐0.04 to 0.08)	0.02 (‐0.04 to 0.08)	na	na
Kennedy 2013	Educational meeting	Usual care	Psychological well‐being	4052	0.00 (‐0.06 to 0.06)	0.00 (‐0.06 to 0.06)	na	na
Légaré 2012	educational meeting and distribution of educational material	Usual care	Health‐related QoL (physical)	359	0.16 (‐0.05 to 0.36)	0.16 (‐0.05 to 0.36)	na	na
Légaré 2012	Educational meeting and distribution of Educational material	Usual care	Health‐related QoL (mental)	359	0.28 (0.07 to 0.49)	0.28 (0.07 to 0.49)	na	na
Kennedy 2013	Educational meeting	Usual care	Health‐related QoL (6 months)	Pre: 5598; Post: 4449	0.00 (‐0.06 to 0.06)	0.00 (‐0.06 to 0.06)	na	na
Kennedy 2013	Educational meeting	Usual care	Health‐related QoL (12 months)	Pre: 5598; Post: 3991	0.00 (‐0.06 to 0.06)	0.00 (‐0.06 to 0.06)	na	na
Sanders 2017	Educational meeting + audit and feed‐back	Usual care	Consultation length (minutes)	175	0.51 (0.21 to 0.81)	0.51 (0.21 to 0.81)	na	na
Cox 2017	Distribution of educatuonal material (FCR checklist) + Educational meeting	Usual care	Safety (parent perception of hospital safety)	Pre: 144; Post: 154	0.00 (‐0.32 to 0.32)	0.00 (‐0.32 to 0.32)	na	na
Categorical data								Meta‐analysis
Study	Intervention	Intervention	Outcome	N	RD	RD (95% CI)	I²	Meta‐analysis
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Knowledge of risk without medication	40	‐0.10 (‐0.42 to) 0.23]	‐0.10 (‐0.42 to 0.23)	na	na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Knowledge of risk risk post‐treatment	40	‐0.17 (‐0.49 to 0.16)	‐0.17 (‐0.49 to 0.16)	na	na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Satisfaction with amount of information (postnatal period)	Pre: 1671 Post: 1492	0.02 (‐0.02 to 0.07)	0.02 (‐0.02 to 0.07)	na	na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Satisfaction with the decision making process (how choice had been made) (postnatal period)	Pre: 1666 Post: 1488	‐0.03 (‐0.07 to 0.02)	‐0.03 (‐0.07 to 0.02)	na	na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Satisfaction with discussion with HCP (View of whether they had had sufficient discussion with HCP) (postnatal period)	Pre: 1657 Post: 1483	‐0.00 (‐0.05 to 0.05)	‐0.00 (‐0.05 to 0.05)	na	na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Anxiety (more anxious) ‐ antenatal period	Pre: 1195 Post: 1527	‐0.00 (‐0.03 to 0.03)	‐0.00 (‐0.02 to 0.02)	0%	x
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Anxiety (more anxious) ‐ postnatal period	Pre: 1651 Post: 1476	0.00 (‐0.03 to 0.03)	‐0.00 (‐0.02 to 0.02)	0%	x
Wilkes 2013	Distribution of Educational material	Usual care	Consultation length (10‐20 minutes)	479	‐0.04 (‐0.13 to 0.05)	‐0.04 (‐0.13 to 0.05)	na	na
Qualitative statement								Meta‐analysis
Study	Intervention	Intervention	Outcome	Direct quote				Meta‐analysis
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Knowldege (antenatal period)	Adjusted mean difference 95%CI : 0.20 (‐0.09 to 0.49) (Page 3, table 3)				na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Knowldege (postnatal period)	Adjusted mean difference 95%CI : 0.20 (‐0.05 to 0.44) (Page 3, table 3)				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Knowledge	"Although there were some tendencies in the expected direction, there was no overall effect by the training in the secondary patient outcomes in either language cohort." Pages 1269‐1270				na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Knowledge (overall)	No differences between group (table 2; supplementary material)				na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Knowledge (Items not addressed in DA ) (4 items)	No differences between group (table 2; supplementary material)				na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Knowledge (Items addressed in DA ) (9 items)	Significant differences between group (table 2; supplementary material)				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Satisfaction with consultation	"There was higher satisfaction with doctor’s consultation skills for patients in ANZ than for patients in SGA, and all cohorts showed improved satisfaction with doctor’s consultation skills except for the ANZ control group (Figure 3c), although not statistically significant (p = 0.08 for SGA and p = 0.26 for ANZ)." Page 1270				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Satisfaction with decision (2 weeks)	"Overall, patients were satisfied with their treatment decision (Figure 3a). Although there were some tendencies in the expected direction, there was no overall effect by the training in the secondary patient outcomes in either language cohort." Pages 1269‐1270				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Satisfaction with decision (4 months)	"Although there were some tendencies in the expected direction, there was no overall effect by the training in the secondary patient outcomes in either language cohort." Pages 1269‐1270				na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Satisfaction with the decision making process (how choice had been made) (antenatal period)	No difference between groups (Page 3; table 2)				na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Satisfaction with discussion with HCP (View of whether they had had sufficient discussion with HCP) (antenatal period)	No difference between groups (Page 3; table 2)				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Satisfaction with doctor communication	"Although there were some tendencies in the expected direction, there was no overall effect by the training in the secondary patient outcomes in either language cohort." Pages 1269‐1270				na
Fossli 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Satisfaction (Patient global satisfaction)	"The duration of encounters (min:sec) did not change significantly (1:03 (p = 0.69, 95% CI 6:13; 4:07)) from pre to post, and neither did patient global satisfaction (0.3 (p = 0.38, 95% CI 0.3; 0.8))." Page 166				na
Murray 2010	Educational meeting + audit and feedback + distribution of Educational material + Educational outreach visit + barriers assessement	Usual care	Satisfaction with the intervention (acceptability and utility of intervention components)	All the intervention group members who logged to the autotutorial and completed the satisfaction survey (n=8), rated the tutorial as excellent to very good, seven indicated that the tutorial was helpful, and six reported that it was very easy to complete. "In all, 37 members of the intervention group (97%) commented on the acceptability of the skills  building workshop. Most (n=35 (94%)) said they would recommend the workshop to others, and 26 (68%) gave it an overall rating of excellent. Workshop participants rated the workshop very good at helping them to understand and use decision support…" Overall, the POC PtDA was rated as acceptable and clinically usefull by the 38 participants who provided data...31(81%) agreed that the PtDA would be acceptable to patients, while 24  (63%) agreed that it would be acceptable to practitioners.” Page 117 “All 36 who participated in the Educational outreach call indicated an interested in using the POC  PtDa and express frustration that it was not available for use in their clinical practice setting.” Page 118				na
O'Cathain 2002	Educational meeting + distribution of Educational material	Usual care	Satisfaction with amount of information (antenatal period)	Difference between groups: adjusted OR 1.40 (1.05 to 1.88) (Page 3; table 2)				na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Decisional conflict	"Decision conflict was low for both groups, and was lower in the Decision Aid arm, but no significant difference was found in the overall scale or in its subscales across arms (Table 2)." page 8				na
Légaré 2012	Educational meeting + distribution of Educational material	Usual care		"The training had no statistically significant  effect on decisional conflict,…" Page E731				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care		"Overall, patients in the SGA cohort reported higher conflict scores than those in the ANZ cohort… Post‐randomisation, although there was no change in conflict in the SGA cohort, in the ANZ cohort there was an increased level of conflict in the control group (estimate = 0.28; ES = 0.17: p = 0.003). This change exceeded the improvement in the training group (estimate = 0.14; ES = 0.09; p = 0.13). In summary, the training appeared to have no overall effect on decisional conflict in either language cohort." Pages 1268‐1269				na
Légaré 2012	Educational meeting + distribution of Educational material	Usual care	Adherence to decision	No differences between group (page E733; table 4)				na
Cooper 2011	Educational meeting	Usual care	Adherence to medication	Changes in patient‐reported adherence to medications at 12 months did not differ for any of  the intervention groups compared to the patient+physician minimal intervention group.”Page 1300				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Health‐related QoL	"The quality of life indicators showed similar findings (data not shown)." Page 1270				na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care	Health‐related QoL	"We found no difference between the FRAX and usual care arms, nor were overall results significantly impacted by analyses comparing the three arms versus only two arms (Decision aid arm vs. FRAX/usual care arms together, see Tables A, B, C, D, and E in S1 File. Therefore, the results comparing the FRAX arm and the Usual Care arm were combined and all subsequent results are presented as Decision Aid vs. FRAX/Usual Care arm (i.e. different forms of usual care)." Page 7 There were no difference between group regarding quality of life [Median (IQR); Decision aid: 85 (80, 95) FRAX/UC: 85 (73, 90); p=0.19 (table 4)]				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Anxiety (state) ‐ 2 weeks	"Anxiety slightly decreased over time for all cohorts. Patients in the SGA (Figure 4a) and ANZ (Figure 4b) cohorts reported comparable anxiety levels at each time point." Page 1270				na
Bernhard 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Anxiety (state) ‐ 4 months	"Anxiety slightly decreased over time for all cohorts. Patients in the SGA (Figure 4a) and ANZ (Figure 4b) cohorts reported comparable anxiety levels at each time point." Page 1270				na
Fossli 2011	Educational meeting + audit and feedback + distribution of Educational material	Usual care	Consultation length	"The duration of encounters (min:sec) did not change significantly ( 1:03 (p = 0.69, 95% CI 6:13; 4:07)) from pre to post, and neither did patient global satisfaction (0.3 (p = 0.38, 95% CI 0.3; 0.8))." Page 166				na
Murray 2010	Educational meeting + audit and feedback + distribution of Educational material + Educational outreach visit + barriers assessement	Usual care		“At baseline there was no significant difference. However, in the post‐calls, the mean call duration was longer in the intervention group at 13,47 minutes (95% confidence interval 11.8;14.21), than in the control group at 10.29 minutes (95% CI 8.79 to 11.79 P = 0.004)” Page 117				na
Shepherd 2011	Educational outreach visit	Usual care		“These effects occurred without any significant difference in consultation length, mean consultation lengths were 26 minutes for control and intervention visits.” Page 381				na
LeBlanc 2015b	Reminder (copy of patient's estimated risk of fracture as computed by the FRAX)	Usual care		"We found no difference between the FRAX and usual care arms, nor were overall results significantly impacted by analyses comparing the three arms versus only two arms (Decision aid arm vs. FRAX/usual care arms together, see Tables A, B, C, D, and E in S1 File..." Page 7				na

Date	Event	Description
15 June 2017	New search has been performed	This is the second update of this Cochrane review first published in 2010. A new search was conducted and other content updated. Forty‐eight new studies were added to the review.
15 June 2017	New citation required but conclusions have not changed	Forty‐eight new studies were added to the review. The review includes 87 studies.

Date	Event	Description
15 September 2014	New search has been performed	Included observer‐reported and patient‐reported outcomes to 2012
30 November 2011	Amended
29 September 2011	New search has been performed	Updated observer‐reported outcomes to 2010

No.	Search terms	Results
#1	(shared decision[tiab] or sharing decision[tiab] or informed decision[tiab] or informed choice[tiab] or decision aid[tiab] or ((share[ti] or sharing[ti] or informed[ti]) and (decision[ti] or deciding[ti] or choice*[ti])))	14116
#2	(decision making[mh:noexp] or decision support techniques[mh:noexp] or decision support systems, clinical[mh] or choice behaviour[mh:noexp] or decision making[tiab] or decision support[tiab] or choice behaviour[tiab] or ((decision[ti] or choice[ti]) and (making[ti] or support[ti] or behaviour[ti])))	195864
#3	(patient participation[mh] or patient participation[tiab] or consumer participation[tiab] or patient involvement[tiab] or consumer involvement[tiab] or ((patient[ti] or consumer[ti]) and (involvement[ti] or involving[ti] or participation[ti] or participating[ti])))	29921
#4	(professional‐patient relations[mh] or ((nurses[mh] or physicians[mh] or nurse[ti] or physician[ti] or clinician[ti] or doctor[ti] or general practitioner[ti] or gps[ti] or health care professional[ti] or healthcare professional[ti] or health care provider[ti] or healthcare provider[ti] or resident[ti]) and (patients[mh] or patient[ti] or consumer[ti] or people*[ti])))	160812
#5	(clinical trial[pt:noexp] or randomized controlled trial[pt] or controlled clinical trial[pt] or evaluation studies[pt] or comparative study[pt] or intervention studies[mh] or Evaluation Studies as Topic[mh:noexp] or program evaluation[mh:noexp] or random allocation[mh] or random[tiab] or double blind[tiab] or controlled trial[tiab] or clinical trial[tiab] or pretest[tiab] or pre test[tiab] or posttest[tiab] or post test[tiab] or prepost[tiab] or pre post[tiab] or controlled before[tiab] or "before and after"[tiab] or interrupted time[tiab] or time serie[tiab] or intervention[tiab])	4001044
#6	((#1 OR (#2 AND #3) OR (#2 AND #4) OR (#3 AND #4)) AND #5)	8177
#7	((#1 OR (#2 AND #3) OR (#2 AND #4) OR (#3 AND #4)) AND #5) Filters: Publication date from 2015/01/01 to 2017/12/31	1760

No.	Search terms	Results
1	(shared decision or sharing decision or informed decision or informed choice or decision aid).ti,ab. or ((share* or sharing* or informed) and (decision or deciding* or choice*)).ti.	12773
2	exp clinical decision making/ or exp decision making/ or exp decision support system/ or exp ethical decision making/ or exp family decision making/ or exp medical decision making/ or exp patient decision making/ or (decision making or decision support or choice behaviour).ti,ab. or ((decision* or choice) and (making or support* or behaviour*)).ti.	355815
3	exp patient participation/ or (patient participation or consumer participation or patient involvement or consumer involvement).ti,ab. or ((patient* or consumer) and (involvement or involving* or participation* or participating*)).ti.	33527
4	exp doctor patient relation/ or exp nurse patient relationship/ or ((exp nurse/ or exp physician/ or (nurse* or physician* or clinician* or doctor* or general practitioners or gps or health care professionals or healthcare professionals or health care providers or healthcare providers or resident)).ti. and (exp patient/ or (patient or consumer* or people*).ti.))	412940
5	1 or (2 and 3) or (2 and 4) or (3 and 4)	46996
6	exp clinical trial/ or exp randomized controlled trial/ or exp controlled clinical trial/ or exp controlled trial/ or exp pretest posttest control group design/ or exp comparative study/ or exp evaluation research/ or exp intervention study/ or exp randomization/ or (random* or double blind or controlled trial or clinical trial or pretest* or pre test or pre tests or posttest* or post test or post tests or prepost* or pre post or controlled before or "before and after" or interruped time or time serie? or intervention*).ti,ab.	8427337
7	(2015* or 2016* or 2017*).yr.	3508644
8	5 and 6 and 7	4460
9	limit 8 to embase	2206

No.	Search terms	Results
#1	((shar* or inform) near/3 (decision or aid* or deciding* or choice*)):ti,ab,kw	2463
#2	((decision* or choice) near/3 (making or support* or behaviour*)):ti,ab,kw	11115
#3	((patient* or consumer) near/3 (involvement or involving* or participation* or participating*)):ti,ab,kw	8416
#4	((nurse* or physician* or clinician* or doctor* or general practitioner* or gps or health care professional* or healthcare professional* or health care provider* or healthcare provider* or resident) near/3 (patient or consumer* or people*)):ti,ab,kw	15782
#5	#1 or (#2 and #3) or (#2 and #4) or (#3 and #4)	4248
#6	#1 or (#2 and #3) or (#2 and #4) or (#3 and #4) Publication Year from 2015 to 2017	1360

No.	Search terms	Results
S1	AB Shared Decision* OR TI Shared Decision* OR AB Sharing Decision* OR TI Sharing Decision* OR AB Informed Decision* OR TI Informed Decision* OR AB Informed Choice* OR TI Informed Choice* OR AB Decision Aid* OR TI Decision Aid* OR ((TI Share* OR TI Sharing OR TI Informed) AND (TI Decision OR TI Deciding* OR TI Choice*))	7,369
S2	MH "Decision Making+" OR MW Decision Support OR AB Decision Making* OR TI Decision Making* OR AB Decision Support* OR TI Decision Support* OR AB Choice Behaviour* OR TI Choice Behaviour* OR ((TI Decision* OR TI Choice) AND (TI Making OR TI Support* OR TI Behaviour*))	85,579
S3	MH Consumer Participation OR AB Patient Participation* OR TI Patient Participation* OR AB Consumer Participation* OR TI Consumer Participation* OR AB Patient Involvement* OR TI Patient Involvement* OR AB Consumer Involvement* OR TI Consumer Involvement* OR ((TI Patient* OR TI Consumer) AND (TI Participating OR TI Participation* OR TI Involving* OR TI Involvement*))	17,733
S4	MH Professional Patient Relations OR MH Nurse Patient Relations OR MH Physician Patient Relations OR ((MH Nurses+ OR MH Physicians+ OR TI Nurse* OR TI Physician* OR TI Clinician* OR TI Doctor* OR TI General Practitioner* OR TI GPs OR TI Health Care Professional* OR TI Healthcare Professional* OR TI Health Care Provider* OR TI Healthcare Provider* OR TI Resident) AND (MH Patients+ OR TI Patient OR TI Consumer* OR TI People*))	47,297
S5	MH Experimental Studies+ OR MH Quasi‐Experimental Studies OR MH Comparative Studies OR MH Evaluation Research OR AB Random* OR TI Random* OR AB Double Blind* OR TI Double Blind* OR AB Controlled Trial* OR TI Controlled Trial* OR AB Clinical Trial* OR TI Clinical Trial* OR AB Pretest* OR TI Prestest* OR AB Pre Test* OR TI Pre Test* OR AB Posttest* OR TI Posttest* OR AB Post Test* OR TI Post Test* OR AB Prepost* OR TI Prepost* OR AB Pre Post* OR TI Pre Post* OR AB Controlled Before* OR TI Controlled Before* OR AB "Before and After" OR TI "Before and After" OR AB Interruped Time* OR TI Interrupted Time* OR AB Time Serie* OR TI Time Serie* OR AB Intervention* OR TI Intervention*	521,147
S6	(S1 OR (S2 AND S3) OR (S2 AND S4) OR (S3 AND S4)) AND S5	3,439
S7	S6 Limiters ‐ Published Date: 20150101‐20171231; Exclude MEDLINE records	341

Search terms	Results
"informed choice"	101
"decision making"	435
"decision support"	295
"informed decision"	85
"decision aid"	297
"sharing decision"	141
"shared decision"	122
	1,476

"informed choice"	16
"decision making"	672
"decision support"	317
"informed decision"	33
"decision aid"	269
"sharing decision"	0
"shared decision"	153
	1,460

[Search in ISDM proceedings]	[Results]
References	700

[Search in SMDM proceedings]	[Results]
References	1607

[Reference lists of systematic reviews]	[Results]
Reference	587

[Reference list of primary studies]	[Results]
Reference	976

Comparison 1: Interventions targeting patients compared with usual care
Observer‐based outcome measure ‐ continuous data		The study not included in the quantitative synthesis supported the pooled result
Studies included in the quantitative synthesis (n = 424)	SMD (95% IC)
LeBlanc 2015a	0.54 (‐0.13 to 1.22)
Maclachlan 2016
LeBlanc 2015b
Tai‐Seale 2016
Observer‐based outcome measure ‐ continuous data
Studies not included in the quantitative synthesis	Results
Hamann 2014	n= 100, no difference between the study groups, unclear risk of bias
Haskard 2008	Unit of error analysis
Patient reported outcome measure ‐ continuous data		It is unlikely that the study not included in the quantitative synthesis would change the direction of the effect size estimate given that its sample size is not very large
Studies included in the quantitative synthesis (n = 1386)	SMD (95% IC)
Deen 2012	0.32 (0.16 to 0.48)
Maranda 2014
Pickett 2012
van der Krieke 2013
van Peperstraten 2010
Cooper 2011
Perestelo‐Perez 2016
Tai‐Seale 2016
Eggly 2017
Patient‐reported outcome measure ‐ continuous data
Studies not included in the quantitative synthesis	Results
Hamann 2014	n = 100, no difference between the study groups, unclear risk of bias
Patient‐reported outcome measure ‐ categorical data		Van Tol‐Geerdink 2016 does not support the pooled result but given that the pooled estimate of the effect size is in favor of the control group, it is likely that adding this study would move the pooled estimate of the effect size towards a null effect. Wolderslund 2017 does not support the pooled result but given its very large sample size, it is likely that adding this study would move the pooled estimate of the effect size towards a positive effect. Korteland 2017 supports the pooled result (null effect).
Studies included in the quantitative synthesis (n = 754)	RD (95% IC)
Krist 2007	‐0,09 (‐0.19 to 0.01)
Landrey 2012
Murray 2001
Sheridan 2014
Vestala 2013
Vodermaier 2009
Patient‐reported outcome measure ‐ categorical data
Studies not included in the quantitative synthesis	Results
van Tol‐Geerdink 2016	n = 240, significant difference in favor of the intervention group, high risk of bias
Wolderslund 2017	n = 4349, significant difference in favor of the intervention group, high risk of bias
Korteland 2017	n = 155, no difference between the study groups, high risk of bias

Comparison 2. Effect of intervention: Interventions targeting healthcare professionals compared to usual care
Observer‐based outcome measure ‐ continuous data		Murray 2010 and Bernhard 2011 ‐ ANZ (Australia and New Zealand sub‐sample) support the pooled result. Bernhard 2011 ‐ SGA (Switzerland, Germany and Austria sub‐sample) does not support the pooled result but it is unlikely that adding these data to the quantitative synthesis would move the pooled estimate of the effect size toward a null effect or in the opposite direction, given its very small sample size. Moreover, the different results observed in the two sub‐samples is a possible indication of an effect modification
Studies included in the quantitative synthesis (n = 479)	SMD (95% IC)
Fossli 2011	0.70 (0.21 to 1.19)
Shepherd 2011
Stacey 2006
Sanders 2017
Cox 2017
LeBlanc 2015b
Observer‐based outcome measure ‐ continuous data
Studies not included in the quantitative synthesis	Results
Bernhard 2011‐ SGA	n = 32, no difference between the study groups, high risk of bias
Bernhard 2011‐ ANZ	n = 21, significant difference in favor of the intervention group, high risk of bias
Murray 2010	n = 88, significant difference in favor of the intervention group, low risk of bias
Patient‐reported outcome measure ‐ continuous data		Légaré 2012 does not support the pooled result. However, it is unlikely that adding these data to the quantitative synthesis would move the pooled estimate of the effect size from the null effect to a significant positive effect.
Studies included in the quantitative synthesis (n = 5772)	SMD (95% IC)
Cooper 2011	0.03 (‐0.15 to 0.20)
Kennedy 2013
Koerner 2014
Tinsel 2013
Wilkes 2013
Patient reported outcome measure ‐ continuous
Studies not included in the quantitative synthesis	Results
Légaré 2012	n = 359, significant difference in favor of the intervention group, unclear risk of bias
Patient ‐reported outcome measure ‐ categorical data		Bernhard 2011 supports the pooled result.
Studies included in the quantitative synthesis (n = 6303)	RD (95% IC)
Légaré 2012	0,01 (‐0.03 to 0.06)
O'Cathain 2002	0,01 (‐0.03 to 0.06)
Patient‐reported outcome measure ‐ categorical
Studies not included in the quantitative synthesis	Results
Bernhard 2011	n = 694, no difference between the study groups, high risk of bias

Comparison 3: Interventions targeting both patients and healthcare professionals compared to usual care
Patient‐reported outcome measure ‐ continous data		Although the confidence interval of the pooled estimate of the effect size contains the null effect, the estimate is in favor of the intervention group. Adding data from Loh 2007 would move the pooled estimate towards a positive effect. Adding Wetzels 2005 would move the result towards a null effect.
Studies included in the quantitative synthesis (n = 1479)	SMD (95% IC)
Cooper 2011	0.13 (‐0.02 to 0.28)
Härter 2015
Hamann 2007
Rise 2012
Wilkes 2013
Tai‐Seale 2016
Epstein 2017
Patient reported outcome measure ‐ continuous data
Studies not included in the quantitative synthesis	Results
Loh 2007	n = 405, significant difference in favor of the intervention group, unclear risk of bias
Wetzels 2005	n = 263, no difference between the study groups, high risk of bias
Patient‐reported outcome measure ‐ categorical data		Leighl 2011 supports the pooled result
Studies included in the quantitative synthesis (n = 266)	RD (95% IC)
Mathers 2012	‐0.01 (‐0.20 to 0.19)
Härter 2015	‐0.01 (‐0.20 to 0.19)
Patient‐reported outcome measure ‐ categorical data
Studies not included in the quantitative synthesis	Results
Leighl 2011	n = 207, no difference between the study groups, unclear risk of bias

Comparison 4: Interventions targeting patients compared to other interventions targeting patients
Patient‐reported outcome measure ‐ continuous data		Lalonde 2006 and Street 1995 support the pooled results
Studies included in the quantitative synthesis (n = 1906)	SMD (95% IC)
Causarano 2014	0.03 (‐0.18 to 0.24)
Deen 2012
Hamann 2011
Schroy 2011
van Roosmalen 2004
Schroy 2016
Adarkwah 2016
Eggly 2017
Hamann 2017
Jouni 2017
Smallwood 2017
Patient‐reported outcome measure ‐ continuous data
Studies not included in the quantitative synthesis	Results
Lalonde 2006	n = 26, no difference between study groups, high risk of bias
Street 1995	n = 60, no difference between study groups, high risk of bias
Patient‐reported outcome measure ‐ categorical data		Kopke 2014 and Butow 2004 support the pooled results. Wolderslund 2017 does not support the pooled result but given its very large sample size, it is likely that adding this study would move the pooled estimate of the effect size toward a positive effect.
Studies included in the quantitative synthesis (n = 2272)	RD (95% IC)
Butow 2004	0,03 (‐0.02 to 0.08)
Davison 1997
Deschamps 2004
Dolan 2002
Kasper 2008
Krist 2007
Raynes‐Greenow 2010
Stiggelbout 2008
Thomson 2007
Davison 2002
Patient‐reported outcome measure ‐ categorical data
Studies not included in the quantitative synthesis	Results
Köpke 2014	n = 192, no difference between the study groups, high risk of bias
Butow 2004	n =164, no difference between the study groups, high risk of bias
Wolderslund 2017	n = 4349, significant difference in favor of the intervention group, high risk of bias

Comparison 5: Interventions targeting healthcare professionals compared to other interventions targeting healthcare professionals
Observer‐based outcome measure ‐ continuous data		Contrary to Elwyn 2004, which reported null results, Feng 2013 reported significant positive results. However it is unlikely that combining these two studies would move the estimate of the effect towards a significant positive result. More studies are needed to draw robust conclusions.
Studies included in the quantitative synthesis (n = 20)	SMD (95% IC)
Elwyn 2004	‐0,30 (‐1.19 to 0.59)
Observer‐based outcome measure ‐ continuous data
Studies not included in the quantitative synthesis	Results
Feng 2013	n =118, significant difference in favor of the intervention group, high risk of bias

[Previous review (Légaré 2014]	[Results]
References	20,792

Name of outcome	1. Certainty of evidence	2. Starting score for certainty of evidence according to study design	3. Criteria to downgrade score
Name of outcome	1. Certainty of evidence		3a. Study limitation	3b. Indirectness of evidence	3c. Inconsistency	3d. Imprecision	3e. Publication bias
Comparison 1: Interventions targeting patients compared with usual care
OBOM ‐ Continuous	1	4	‐2	‐1	‐2	‐2	N/A
PROM ‐ Continuous	1	4	‐2	‐1	‐2	0	0
PROM ‐ Continuous ‐ (Non‐randomized control trial)	1	2	‐2	‐1	N/A	‐1	N/A
PROM ‐ Categorical	1	4	‐2	‐1	‐1	0	0
Comparison 2: Interventions targeting healthcare professionals compared to usual care
OBOM ‐ Continuous	1	4	‐1	‐1	‐2	‐1	0
OBOM ‐ Continuous ‐ (Controlled before‐after study)	1	2	‐2	‐1	N/A	‐1	N/A
PROM ‐ Continuous	1	4	‐1	‐1	‐2	‐1	0
PROM ‐ Categorical	1	4	‐1	‐1	‐2	0	N/A
Comparison 3: Interventions targeting both patients and healthcare professionals compared to usual care
OBOM ‐ Continuous	1	4	‐1	‐1	‐2	‐1	‐1
PROM ‐ Continuous	1	4	‐1	‐1	‐1	0	0
PROM ‐ Categorical	1	4	‐1	‐1	‐1	‐1	N/A
Comparison 4: Interventions targeting patients compared to other interventions targeting patients
OBOM ‐ Continuous	1	4	‐1	‐1	‐2	‐1	N/A
PROM ‐ Continuous	1	4	‐1	‐1	‐2	‐1	0
PROM ‐ Continuous ‐ (Non‐randomized control trial)	1	2	‐2	‐1	N/A	‐1	N/A
PROM ‐ Categorical	1	4	‐2	‐1	‐1	0	0
Comparison 5: Interventions targeting healthcare professionals compared to other interventions targeting healthcare professionals
OBOM ‐ Continuous	1	4	0	‐1	N/A	‐2	N/A
PROM ‐ Continuous	1	4	‐1	‐1	‐2	‐1	N/A
Comparison 6: Interventions targeting both patients and healthcare professionals compared to other interventions targeting both patients and healthcare professionals
OBOM ‐ Continuous	1	4	‐1	‐1	N/A	‐1	N/A
OBOM ‐ Categorial	1	4	‐2	‐1	N/A	‐1	N/A
PROM ‐ Continuous	1	4	‐1	‐1	N/A	‐1	N/A

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Shared decision making (OBOM, continuous)	4	424	Std. Mean Difference (IV, Random, 95% CI)	0.54 [‐0.13, 1.22]
1.1 Cluster RCT	2	116	Std. Mean Difference (IV, Random, 95% CI)	0.75 [0.22, 1.29]
1.2 Individual RCT	2	308	Std. Mean Difference (IV, Random, 95% CI)	0.48 [‐0.76, 1.72]
2 Shared decision making (PROM, continuous)	9	1386	Std. Mean Difference (IV, Random, 95% CI)	0.32 [0.16, 0.48]
2.1 Cluster RCT	2	303	Std. Mean Difference (IV, Random, 95% CI)	0.40 [0.18, 0.63]
2.2 Individual RCT	7	1083	Std. Mean Difference (IV, Random, 95% CI)	0.29 [0.10, 0.49]
3 Shared decision making (PROM, continuous) ‐ NRCT	1	303	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.21, 0.25]
4 Shared decision making (PROM, categorical)	6	754	Risk Difference (M‐H, Random, 95% CI)	‐0.09 [‐0.19, 0.01]
5 Knowledge	3	565	Std. Mean Difference (IV, Random, 95% CI)	0.38 [0.16, 0.61]
6 Knowledge (categorical)	2	312	Risk Difference (M‐H, Random, 95% CI)	0.17 [0.05, 0.29]
7 Satisfaction	1	107	Std. Mean Difference (IV, Random, 95% CI)	0.14 [‐0.24, 0.52]
8 Decisional conflict	3	367	Std. Mean Difference (IV, Random, 95% CI)	‐0.30 [‐0.68, 0.09]
9 Decision regret	1	212	Std. Mean Difference (IV, Random, 95% CI)	‐0.10 [‐0.39, 0.19]
10 Patient‐physician communication (number of topics raised by patients)	1	100	Std. Mean Difference (IV, Random, 95% CI)	0.26 [‐0.13, 0.65]
11 Patient‐physician communication (patient raised discussion)	1	157	Risk Difference (M‐H, Random, 95% CI)	0.29 [0.14, 0.44]
12 Patient‐physician communication (patient participation in discussion)	1	157	Risk Difference (M‐H, Random, 95% CI)	0.27 [0.13, 0.42]
13 Decision self‐efficacy	2	274	Std. Mean Difference (IV, Random, 95% CI)	0.16 [‐0.08, 0.40]
14 Empowerment	1	342	Std. Mean Difference (IV, Random, 95% CI)	0.26 [0.05, 0.48]
15 Empowerment (categorical)	1	262	Risk Difference (M‐H, Random, 95% CI)	0.18 [0.09, 0.27]
16 Adherence	2	598	Risk Difference (M‐H, Random, 95% CI)	‐0.03 [‐0.07, 0.02]
17 Health‐related quality of life (physical)	1	116	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.36, 0.36]
18 Health‐related quality of life (mental)	1	116	Std. Mean Difference (IV, Random, 95% CI)	0.10 [‐0.26, 0.46]
19 Anxiety	2	419	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.33, 0.37]
20 Anxiety (categorical)	1	127	Risk Difference (M‐H, Random, 95% CI)	0.04 [‐0.07, 0.15]
21 Depression	1	127	Risk Difference (M‐H, Random, 95% CI)	0.16 [0.05, 0.28]
22 Consultation length	2	224	Std. Mean Difference (IV, Random, 95% CI)	0.10 [‐0.39, 0.58]
23 Cost	1	105	Std. Mean Difference (IV, Random, 95% CI)	0.82 [0.42, 1.22]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Shared decision making (OBOM, continuous)	6	479	Std. Mean Difference (IV, Random, 95% CI)	0.70 [0.21, 1.19]
1.1 Cluster RCT	2	329	Std. Mean Difference (IV, Random, 95% CI)	0.48 [‐0.25, 1.21]
1.2 Individual RCT	4	150	Std. Mean Difference (IV, Random, 95% CI)	0.85 [0.06, 1.64]
2 Shared decision making (OBOM, continuous) ‐ CBAs	1	21	Std. Mean Difference (IV, Random, 95% CI)	‐0.10 [‐0.96, 0.76]
3 Shared decision making (PROM, continuous)	5	5772	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.15, 0.20]
3.1 Cluster RCT	4	5678	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.17, 0.20]
3.2 Individual RCT	1	94	Std. Mean Difference (IV, Random, 95% CI)	0.11 [‐0.30, 0.51]
4 Shared decision making (PROM, categorical)	2	6303	Risk Difference (M‐H, Random, 95% CI)	0.01 [‐0.03, 0.06]
5 Knowledge	2	969	Std. Mean Difference (IV, Random, 95% CI)	0.26 [‐0.16, 0.69]
6 Knowledge (categorical)	1	80	Risk Difference (M‐H, Random, 95% CI)	‐0.13 [‐0.36, 0.10]
7 Satisfaction with consultation	1	479	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.18, 0.18]
8 Satisfaction with information	1	1492	Risk Difference (M‐H, Random, 95% CI)	0.02 [‐0.02, 0.07]
9 Satisfaction with decision making process	1	1488	Risk Difference (M‐H, Random, 95% CI)	‐0.03 [‐0.07, 0.02]
10 Satisfaction with discussion	1	1483	Risk Difference (M‐H, Random, 95% CI)	‐0.00 [‐0.05, 0.05]
11 Decision regret	1	326	Std. Mean Difference (IV, Random, 95% CI)	0.29 [0.07, 0.51]
12 Self‐efficacy	1	4475	Std. Mean Difference (IV, Random, 95% CI)	‐0.03 [‐0.09, 0.03]
13 Adherence	1	827	Std. Mean Difference (IV, Random, 95% CI)	‐0.08 [‐0.21, 0.06]
14 General health	1	4056	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.04, 0.08]
15 Psychological well‐being	1	4052	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.06, 0.06]
16 Health‐related quality of life (physical)	1	359	Std. Mean Difference (IV, Random, 95% CI)	0.16 [‐0.05, 0.36]
17 Health‐related quality of life (mental)	1	359	Std. Mean Difference (IV, Random, 95% CI)	0.28 [0.07, 0.49]
18 Health‐related quality of life	1	4449	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.06, 0.06]
19 Anxiety	1	3003	Risk Difference (M‐H, Random, 95% CI)	‐0.00 [‐0.02, 0.02]
20 Consultation length	1	175	Std. Mean Difference (IV, Random, 95% CI)	0.51 [0.21, 0.81]
21 Consultation length (10‐20 min)	1	479	Risk Difference (M‐H, Random, 95% CI)	‐0.04 [‐0.13, 0.05]
22 Safety	1	154	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.32, 0.32]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Shared decision making (OBOM, continuous)	6	1270	Std. Mean Difference (IV, Random, 95% CI)	1.10 [0.42, 1.79]
1.1 Cluster RCT	2	59	Std. Mean Difference (IV, Random, 95% CI)	0.67 [0.14, 1.20]
1.2 Individual RCT	4	1211	Std. Mean Difference (IV, Random, 95% CI)	1.31 [0.45, 2.17]
2 Shared decision making (PROM, continuous)	7	1479	Std. Mean Difference (IV, Random, 95% CI)	0.13 [‐0.02, 0.28]
2.1 Cluster RCT	4	890	Std. Mean Difference (IV, Random, 95% CI)	0.20 [‐0.01, 0.41]
2.2 Individual RCT	3	589	Std. Mean Difference (IV, Random, 95% CI)	‐0.01 [‐0.17, 0.16]
3 Shared decision making (PROM, categorical)	2	266	Risk Difference (M‐H, Random, 95% CI)	‐0.01 [‐0.20, 0.19]
3.1 Cluster RCT	1	169	Risk Difference (M‐H, Random, 95% CI)	‐0.09 [‐0.23, 0.05]
3.2 Individual RCT	1	97	Risk Difference (M‐H, Random, 95% CI)	0.11 [‐0.10, 0.31]
4 Knowledge	2	1004	Std. Mean Difference (IV, Random, 95% CI)	0.41 [0.28, 0.53]
5 Knowledge (categorical)	4	1260	Risk Difference (M‐H, Random, 95% CI)	0.28 [0.05, 0.51]
6 Satisfaction with care	2	362	Std. Mean Difference (IV, Random, 95% CI)	0.51 [‐0.34, 1.36]
7 Satisfaction with decision	1	424	Std. Mean Difference (IV, Random, 95% CI)	0.24 [0.05, 0.43]
8 Satisfaction with consultation	1	393	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.23, 0.23]
9 Decisional conflict	2	1065	Std. Mean Difference (IV, Random, 95% CI)	‐0.35 [‐0.71, 0.01]
10 Confidence in decision	1	414	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.17, 0.22]
11 Decision regret	1	369	Std. Mean Difference (IV, Random, 95% CI)	0.13 [‐0.08, 0.33]
12 Patient‐physician communication (patient‐centered communication)	1	265	Std. Mean Difference (IV, Random, 95% CI)	0.23 [‐0.01, 0.47]
13 Match between preferred and actual level of participation in decision making	2	185	Risk Difference (M‐H, Random, 95% CI)	‐0.03 [‐0.16, 0.10]
14 Adherence	1	489	Std. Mean Difference (IV, Random, 95% CI)	0.60 [0.36, 0.83]
15 Adherence (categorical)	2	145	Risk Difference (M‐H, Random, 95% CI)	0.00 [‐0.15, 0.15]
16 Health‐related quality of life	1	265	Std. Mean Difference (IV, Random, 95% CI)	0.08 [‐0.16, 0.33]
17 Health‐related quality of life (physical)	1	75	Std. Mean Difference (IV, Random, 95% CI)	0.08 [‐0.37, 0.54]
18 Health‐related quality of life (mental)	1	75	Std. Mean Difference (IV, Random, 95% CI)	0.01 [‐0.44, 0.46]
19 Anxiety	1	419	Std. Mean Difference (IV, Random, 95% CI)	‐0.12 [‐0.31, 0.08]
20 Depression	1	418	Std. Mean Difference (IV, Random, 95% CI)	‐0.14 [‐0.33, 0.05]
21 Consultation length	1	536	Std. Mean Difference (IV, Random, 95% CI)	3.72 [3.44, 4.01]
22 Safety	1	898	Risk Difference (M‐H, Random, 95% CI)	0.0 [‐0.00, 0.00]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Shared decision making (OBOM, continuous)	3	271	Std. Mean Difference (IV, Random, 95% CI)	0.88 [0.39, 1.37]
2 Shared decision making (PROM, continuous)	11	1906	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.18, 0.24]
2.1 Cluster RCT	1	78	Std. Mean Difference (IV, Random, 95% CI)	0.30 [‐0.14, 0.75]
2.2 Individual RCT	10	1828	Std. Mean Difference (IV, Random, 95% CI)	0.01 [‐0.22, 0.23]
3 Shared decision making (PROM, continuous) comp1 ‐ NRCT	1	97	Std. Mean Difference (IV, Random, 95% CI)	‐0.21 [‐0.61, 0.19]
4 Shared decision making (PROM, continuous) comp2 ‐ NRCT	1	110	Std. Mean Difference (IV, Random, 95% CI)	‐0.19 [‐0.56, 0.19]
5 Shared decision making (PROM, continuous) comp3 ‐ NRCT	1	99	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.37, 0.43]
6 Shared decision making (PROM, categorical)	10	2272	Risk Difference (M‐H, Random, 95% CI)	0.03 [‐0.02, 0.08]
7 Knowledge	1	596	Std. Mean Difference (IV, Random, 95% CI)	0.30 [0.13, 0.47]
8 Knowledge (categorical)	3	706	Risk Difference (M‐H, Random, 95% CI)	0.16 [‐0.10, 0.42]
9 Satisfaction with decision	1	596	Std. Mean Difference (IV, Random, 95% CI)	0.07 [‐0.10, 0.24]
10 Satisfaction with treatment	2	267	Std. Mean Difference (IV, Random, 95% CI)	‐0.09 [‐0.34, 0.16]
11 Satisfaction with consultation	1	207	Std. Mean Difference (IV, Random, 95% CI)	‐0.14 [‐0.42, 0.13]
12 Satisfaction with information provided	1	39	Std. Mean Difference (IV, Random, 95% CI)	0.11 [‐0.52, 0.73]
13 Decisional conflict	5	1088	Std. Mean Difference (IV, Random, 95% CI)	‐0.20 [‐0.48, 0.08]
14 Decision uncertainty	1	80	Std. Mean Difference (IV, Random, 95% CI)	‐0.21 [‐0.65, 0.23]
15 Decision self‐efficacy	2	100	Std. Mean Difference (IV, Random, 95% CI)	‐0.02 [‐0.41, 0.37]
16 Match between preferred and actual level of participation in decision making	4	1206	Risk Difference (M‐H, Random, 95% CI)	‐0.10 [‐0.16, ‐0.05]
17 Match between preferred option and decision made	2	363	Risk Difference (M‐H, Random, 95% CI)	‐0.20 [‐0.60, 0.20]
18 Adherence	1	100	Std. Mean Difference (IV, Random, 95% CI)	0.05 [‐0.35, 0.44]
19 Adherence (categorical)	3	301	Risk Difference (M‐H, Random, 95% CI)	0.01 [‐0.10, 0.12]
20 General health	1	88	Std. Mean Difference (IV, Random, 95% CI)	‐0.19 [‐0.61, 0.23]
21 Anxiety	2	682	Std. Mean Difference (IV, Random, 95% CI)	‐0.11 [‐0.27, 0.05]
22 Depression	1	86	Std. Mean Difference (IV, Random, 95% CI)	‐0.27 [‐0.69, 0.16]
23 Consultation length	1	39	Std. Mean Difference (IV, Random, 95% CI)	‐0.65 [‐1.29, ‐0.00]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Shared decision making (OBOM, continuous)	1	20	Std. Mean Difference (IV, Random, 95% CI)	‐0.30 [‐1.19, 0.59]
2 Shared decision making (PROM, continuous)	2	1459	Std. Mean Difference (IV, Random, 95% CI)	0.24 [‐0.10, 0.58]
2.1 Cluster RCT	1	327	Std. Mean Difference (IV, Random, 95% CI)	0.05 [‐0.17, 0.27]
2.2 Individual RCT	1	1132	Std. Mean Difference (IV, Random, 95% CI)	0.40 [0.28, 0.52]
3 Health status (mental)	1	295	Std. Mean Difference (IV, Random, 95% CI)	0.24 [0.01, 0.47]
4 Health status (physical)	1	295	Std. Mean Difference (IV, Random, 95% CI)	0.05 [‐0.19, 0.28]
5 Anxiety	1	843	Std. Mean Difference (IV, Random, 95% CI)	0.14 [0.00, 0.28]

PERMALINK

Interventions for increasing the use of shared decision making by healthcare professionals

France Légaré

Rhéda Adekpedjou

Dawn Stacey

Stéphane Turcotte

Jennifer Kryworuchko

Ian D Graham

Anne Lyddiatt

Mary C Politi

Richard Thomson

Glyn Elwyn

Norbert Donner‐Banzhoff

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Patient outcomes

Affective‐cognitive outcomes

Behavioral outcomes

Health outcomes

Process outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Trial registries

Data collection and analysis

Selection of studies

1.

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Assessment of heterogeneity

Data synthesis

'Summary of findings' tables

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

Included studies

Settings and participants

Target categories

Interventions targeting patients (44 studies)

Interventions targeting healthcare professionals (15 studies)

Interventions targeting both patients and healthcare professionals (28 studies)

Characteristics of interventions and comparisons

Interventions targeting patients

Interventions targeting healthcare professionals

Interventions targeting both patients and healthcare professionals

Conceptual framework

Outcome measures

Primary outcome (use of shared decision making)

Secondary outcomes

Excluded studies

Risk of bias in included studies

2.

3.

Allocation

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Shared decision making (OBOM, continuous)	1	20	Std. Mean Difference (IV, Random, 95% CI)	‐0.29 [‐1.17, 0.60]
2 Shared decision making (OBOM, categorical)	1	134	Risk Difference (M‐H, Random, 95% CI)	‐0.04 [‐0.13, 0.04]
3 Shared decision making (PROM, continuous)	1	150	Std. Mean Difference (IV, Random, 95% CI)	0.0 [‐0.32, 0.32]
4 Decisional conflict	1	286	Std. Mean Difference (IV, Random, 95% CI)	‐0.07 [‐0.30, 0.16]

Methods	Study design: randomized trial Unit of allocation: patient Unit of analysis: patient Power calculation: unclear
Participants	Care setting: primary care, Germany Health professionals: 32; general practitioners; fully trained Patients: 304; cardiovascular risk prevention; male and female
Interventions	Single intervention: patient‐mediated intervention (Computerised decision aid (TTE)) Quote: "Immediately after giving their informed consent, patients were randomized to consultation with the emoticons (Fig. 2) or the TTE illustration (Fig. 3). GPs entered a study ID into the decision support software, which automatically allocated each patient into one of the two conditions according to an a priori randomized sequence. GPs learned about each patient’s allocation by the illustration displayed by the software. They then started a discussion with their patients on the basis of the allocated display, i.e. either emoticons, or TTE, respectively." Page 3, figure 3 Single intervention: patient‐mediated intervention (Computerised decision aid (emoticon)). Figure 2
Outcomes	PEF‐FB‐9 (SDM‐Q9) (continuous)
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "GPs entered a study ID into the decision support software, which automatically allocated each patient into one of the two conditions according to an a priori randomized sequence." page 3
Allocation concealment (selection bias)	Low risk	Quote: "GPs entered a study ID into the decision support software, which automatically allocated each patient into one of the two conditions according to an a priori randomized sequence." page 3
Blinding (performance bias and detection bias)   Participant‐reported outcome	High risk	Performance bias. Quote: "GPs recorded the decision made, such as specific medications, dose adjustments, behavioral measures or no change at all." page 3 Quote: "GPs learned about each patient’s allocation by the illustration displayed by the software." page 3
Incomplete outcome data (attrition bias)   Participant‐reported outcome	Unclear risk	Comment: we cannot assume that all patients replied to all questions related to outcomes. Missing outcome data were not specified.
Selective reporting (reporting bias)	Low risk	Comment: relevant outcomes pre‐specified in the study protocol were reported in the results (Clinical Trials Register Platform (ICTRP, ID DRKS00004933)).
Other bias	Low risk	Comment: no evidence of other risk of biases.
Baseline measurement?   Participant‐reported outcome	Unclear risk	Comment: no baseline measure of primary outcome.
Protection against contamination?	High risk	Comment: patients were randomized.
Baseline characteristics patients	Low risk	Quote: "Both study arms were well‐balanced regarding socio demographic and clinical variables." page 6. See Table 1
Baseline characteristics healthcare professionals	Unclear risk	Comment: no report of characteristics.

Methods	Study design: non‐randomized trial Unit of allocation: patient Unit of analysis: patient Power calculation: unclear
Participants	Care setting: ambulatory care, specialized care, USA Health professionals: various types; fully trained and in training Patients: 371; gastrointestinal disorders; male and female
Interventions	Single intervention ‐ patient‐mediated (GI PROMIS)  Quote: "Intervention patients completed GI PROMIS symptom questionnaires on an e‐portal one week before their visit." page 1  Quote: "Using modern psychometric techniques, such as item response theory and computerized adaptive testing,(14, 15) PROMIS offers state‐of‐the‐art psychometrics, establishes common‐language benchmarks for symptoms across conditions, and identifies clinical thresholds for action and meaningful clinical improvement or decline. PROMIS questionnaires are administered electronically and efficiently, allowing implementation in busy clinical settings. Because of the extraordinary burden of illness from digestive diseases, the PROMIS consortium added a gastrointestinal (GI) item bank, which our group developed." page 3 Usual care Quote: "Usual care patients were managed according to customary practices." page 2
Outcomes	SDM‐Q‐9 (continuous)
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Comment: no randomization.
Allocation concealment (selection bias)	High risk	Quote: "We used a pragmatic, off‐on study design alternating weekly between the PROMIS intervention and control arms." page 4
Blinding (performance bias and detection bias)   Participant‐reported outcome	Unclear risk	Comment: insufficient information to make a judgement.
Incomplete outcome data (attrition bias)   Participant‐reported outcome	Low risk	Quote: "Specifically, the missing outcome data for this group was imputed to the corresponding mean value calculated from controls for each item. Because this assumption biases towards the null, we also performed a sensitivity analysis using a per‐protocol approach where we excluded patients without follow‐up data." page 6
Selective reporting (reporting bias)	Unclear risk	Comment: study protocol not available.
Other bias	Unclear risk	Quote: "We only evaluated patients with GI symptoms, so we cannot know whether using other PROMIS questionnaires, such as those for fatigue, physical function, or pain, among many others, would also fail to show a difference vs. usual care." page 8
Baseline measurement?   Participant‐reported outcome	Unclear risk	Comment: insufficient information to make a judgement.
Protection against contamination?	Unclear risk	Comment: insufficient information to make a judgement.
Baseline characteristics patients	Low risk	Quote: "No differences were seen in gender and race/ethnicity between groups (Appendix Table 2)." page 7
Baseline characteristics healthcare professionals	Unclear risk	Comment: no report of characteristics.

Methods	Study design: clinician‐randomized trial Unit of allocation: clinician Unit of analysis: patient Power calculation: done
Participants	Care setting: specialized care; ambulatory care; Australia, New Zealand, Switzerland, Germany, and Austria Health professionals: 62; various type of physician (medical, surgical, radiation and gynacological oncologists); fully trained Patients: 694; breast cancer; female
Interventions	Multifaceted intervention: educational meeting, audit and feedback, distribution of educational materials (interactive face‐to face workshop and two follow‐up telephone calls) Quote: "The training consisted of a 7 hours interactive face to‐face workshop with one to two follow‐up telephone calls over 2 months. The elements of this training were evidence‐based .... The training focused on four key concepts: ... The workshops were held at the participating centres and conducted in the local language by one to two clinical psychologists ... The teaching materials were in English .... Before the workshop, participants were expected to have read the strategies document." Page 1267 Usual care (control): No training workshop Quote: "Following baseline assessment and before the scheduled training workshop, they were randomly assigned to ... or control (no training workshop) group." Page 1267
Outcomes	Patient involvement preference and actual involvement; SDM framework (DAS‐O subscale) (qualitative)
Notes	Additional information Number of approached patients (eligible): SGA (Swiss/German/Austrian): 429; ANZ (Australian/New Zealand): 340 Number of patients per physician: SGA (Swiss/German/Austrian): 41; ANZ (Australian/New Zealand): 21
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "They were randomly assigned to the experimental (training workshop) or control (no training workshop) group." page 1266
Allocation concealment (selection bias)	Unclear risk	Comment: not specified in the paper
Blinding (performance bias and detection bias)   Observer‐based outcome	Unclear risk	Quote: "Two raters applied the DAS‐O coding system to the consultations." page 7 (Butow 2014). Comment: in the paper, it is not specified if the raters were unaware of the allocation of the audiotaped record.
Blinding (performance bias and detection bias)   Participant‐reported outcome	Unclear risk	Comment: Not specified in the paper.
Incomplete outcome data (attrition bias)   Observer‐based outcome	High risk	Quote: "53 (85,5%) doctors… were assessable with regard to the primary endpoint for this analysis: doctor behavior…" page 3 (Butow 2014).  In the ANZ group, 75% of the consultation requested were audiotaped. In the SGA group, 9 doctors did not provide the an audiotape (7 were gynecologists). 74 % of the requested consultation among the 32 who provided audiotape were audiotaped.    Quote: "Characteristics of doctors with... and without... audio tapes were compared... results suggest a difference in doctor speciality... Physicians without tapes were also younger" page 4‐5 (Butow 2004).    Comment: it is likely that the missing primary outcome data (non audiotaped) were related to doctors behaviors. If the missing outcomes were related to a characteristic that is related to the intervention, it may be a selection bias. However, age and specialty are similar at baseline among doctors.
Incomplete outcome data (attrition bias)   Participant‐reported outcome	Unclear risk	Comment: for our primary outcome, but low risk for the primary outcome of the study (decisional conflict).
Selective reporting (reporting bias)	High risk	Comment: outcomes of interest are reported incompletely so they cannot be entered in a meta‐analysis.
Other bias	Low risk	Comment: no evidence of other risk of biases.
Baseline measurement?   Observer‐based outcome	Unclear risk	Comment; no baseline measure of our primary outcome.
Baseline measurement?   Participant‐reported outcome	Low risk	Quote: "Within two weeks of their initial consultation discussing treatment options, patients gave informed consent and completed a baseline questionnaire gathering demographics; preferences for information (degree of detail required on a Likert scale from ‘prefer few details’ to ‘prefer as many details ... " page 1266
Protection against contamination?	Unclear risk	Comment: physicians within centre were allocated to intervention or control.
Baseline characteristics patients	Low risk	Quote: "Baseline characteristics of eligible patients are summarised for the pre‐randomisation and post‐randomisation cohorts in Table 2." page 1268
Baseline characteristics healthcare professionals	Low risk	Quote: "Baseline characteristics of eligible doctors are shown in Table 1." page 1268

Methods	Study design: patient‐randomized trial Unit of allocation: patient Unit of analysis: patient Power calculation: not clear
Participants	Care setting: specialized care and ambulatory care (Rheumatologic Outpatient Clinic of the University of Heidelberg); Germany  Health professionals: 10; internal medicine; fully trained Patients: 149; fibromyalgia syndrome; male and female
Interventions	Multifaceted intervention: educational meeting with physician (18 hours); patient‐mediated intervention (computer‐based visualized information tool). The computer‐based tool provided information on fibromyalgia syndrome, combining textual information with diagrams and short video sequences. The educational meeting involved training physicians to improve patient‐centered communication and interaction skills. Single intervention (control): patient‐mediated intervention (computer‐based visualized information tool) The tool was the same as the multifaceted intervention.
Outcomes	Doctor‐patient interaction, from the patient perspective, using the QQPPI (Questionnaire on the Quality of Physician‐Patient Interaction) (continuous); joint process between healthcare professionals and patients to make decisions.
Notes	Additional information Number of approached patients (eligible): not reported Number of patients per physician: not reported
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: patients were randomized but the method was unspecified.
Allocation concealment (selection bias)	Unclear risk	Comment: not specified in the paper.
Blinding (performance bias and detection bias)   Participant‐reported outcome	Low risk	Quote: "Patients were informed on the intervention but they were blinded to the fact in which group they were being treated." Page 359
Incomplete outcome data (attrition bias)   Participant‐reported outcome	Unclear risk	Comment: 64/149 (43%) patients were excluded after randomization (did not meet inclusion criteria, refused to complete questionnaire); Information about missing data in our primary outcome were lacking.
Selective reporting (reporting bias)	Unclear risk	Comment: no evidence that outcomes were selectively reported, but no protocol.
Other bias	Low risk	Comment: no evidence of other risk of biases.
Baseline measurement?   Participant‐reported outcome	Unclear risk	Comment: baseline measurements for the FAPI are not reported, nor were they measured.
Protection against contamination?	High risk	Comment: it was the patients who were randomized in an university outpatient setting.
Baseline characteristics patients	Low risk	Quote: "All three patient group were comparable as to socio‐economic (see Table 1) and health related variables (see Table 3)." page 359
Baseline characteristics healthcare professionals	Unclear risk	Comment: no report of characteristics.

Methods	Study design: cluster‐randomized trial Unit of allocation: practice Unit of analysis: provider and patient Power calculation: not done
Participants	Care setting: primary care; ambulatory care; USA Health professionals: 41; various type (physician, nurse practitioner, physician assistant, resident/fellow); fully trained and in training Patients: 110; Type 2 diabetes; male and female
Interventions	Multifaceted intervention: patient‐mediated intervention, educational meeting. Quote: "The intervention will consist of the use of a decision aid (Statin Choice and Aspirin Choice, or Diabetes Medication Choice) by patients and their primary care clinician during the clinical encounter ." page 3 of the study protocol. Quote: "A study team member will conduct a demonstration showing how to use the decision aid at the time of the initial in‐person discussion with clinics. The focal points of the demonstration will be that decision aids serve as guides for conversation rather than scripted discussions... Brief video clips and storyboards that demonstrate the basic use of decision aids are publicly available at http://kercards.e‐bm.info for clinicians to review at their convenience. A study team member will remain available to do one‐on‐one demonstrations after the initial group demonstration if needed." page 4 of the study protocol Usual care (control): Quote: "For patients in the usual care arm, clinicians will manage the discussion about medication regimen as usual, without using decision aids. " page 4 of the study protocol.
Outcomes	Level of patient engagement (OPTION) (continuous)
Notes	Additional information Number of approached patients (eligible): not reported Number of patients per physician: not reported
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: not described in the paper.
Allocation concealment (selection bias)	Unclear risk	Quote: "A study statistician will perform the randomization centrally after the practice has been enrolled." (Study protocol page 3, column 2). However, intervention and usual care patients were recruited within each practice (diabetes DA and UC, statin DA and UC). Investigators, clinicians were not blinded to the practice status. If the persons who enrolled the patients were not blinded too, it may have biased the selection.
Blinding (performance bias and detection bias)   Observer‐based outcome	Unclear risk	Comment: data collectors and analysts were blinded to allocation (p7 para 1) but it is not specified whether investigators who assessed videos recorded were blinded.
Incomplete outcome data (attrition bias)   Observer‐based outcome	Unclear risk	Quote: "We were able to obtain video recordings from 38% of encounters. This limits our ability to use our checklist and to obtain an OPTION score in all encounters thus reducing our confidence in the inferences related to fidelity and clinicians’ efforts to engage patients in decision making, respectively." page 6 para 3  DA: 41,5% UC: 34%
Selective reporting (reporting bias)	High risk	Comment: see study protocol: not all the study's prespecified relevant outcomes were reported (quality of life, costs and resources utilization). Morevover some relevant outcomes are specified differently in the results: e.g. decision comfort instead of decisional conflict, patient satisfaction with knowledge transfer instead of satisfaction with decision making.
Other bias	Low risk	Comment: no evidence of other risk of biases.
Baseline measurement?   Observer‐based outcome	Unclear risk	Comment: no baseline measure of our primary outcome.
Protection against contamination?	Low risk	Comment: intervention and usual care patients were recruited within each practice (diabetes DA and UC, statin DA and UC).
Baseline characteristics patients	High risk	Quote: "All patient factors were well balanced across both arms with a difference found in the type of discussion that patients had (statins vs. diabetes medication; Table 1); subsequent results adjust for this difference." page 4 column 2  Comment: there is imbalance in race and HbA1c too.
Baseline characteristics healthcare professionals	Unclear risk	Comment: there is mention of participant characteristics in table 2 but comparisons between intervention and control arms were not presented.

Methods	Study design: randomized trial (factorial design) Unit of allocation: physician and patient Unit of analysis: physician and patient Power calculation: unclear
Participants	Care setting: primary care, ambulatory care (especially low SES service), USA Health professionals: 41, physicians fully trained Patients: 279, hypertensive; 184 female
Interventions	Four arms: patient‐mediated intervention, educational meeting (physician communication skills training and patient coaching by community health workers) educational meeting: physician communication skills training patient‐mediated intervention: patient coaching by community health workers patient and physician minimal intervention: (control) Quote: "The physician communication skills program was designed to provide physicians with personalized feedback based on their videotaped performance with a simulated patient scheduled for an office appointment. ... Intervention group physicians reviewed the videotape of their personal interviews with the simulated patient and completed exercises on the CD‐ROM or in the workbook." page 1298 Quote: "Control group physicians participated in the simulated visit but did not receive any feedback until the end of the study" page 1298
Outcomes	Participatory Decision making (PDM); Patient involvement in care
Notes	Additional information Number of approached patients (eligible): 980 Number of patients per physician: 50
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Random blocks of size two and four were used, and a list of random numbers between zero and one was generated in Stata version 7.0." protocol
Allocation concealment (selection bias)	Low risk	Quote: "The study statistician generated the allocation sequence for both physicians and patients and placed the intervention assignment for each subject in opaque envelopes to be opened by research assistants." Protocol
Blinding (performance bias and detection bias)   Participant‐reported outcome	High risk	Quote: "Due to the nature of the interventions, complete masking of participants, investigators and CHWs was not possible." page 1299
Incomplete outcome data (attrition bias)   Participant‐reported outcome	Unclear risk	Comment: we do not know if missing outcomes were balanced across intervention groups.  Imputation was used by assuming that data are MAR but the mechanism and the reasons of missing data were not reported.
Selective reporting (reporting bias)	Low risk	Comment: outcomes were described in the protocol.
Other bias	Low risk	Comment: no evidence of other risk of biases.
Baseline measurement?   Participant‐reported outcome	Low risk	Comment: process measures at baseline and change at 12‐month follow‐up by intervention group.
Protection against contamination?	Unclear risk	Comment: professionals were allocated within a clinic.
Baseline characteristics patients	Low risk	Comment: see table 2.
Baseline characteristics healthcare professionals	Low risk	Comment: see table 1.

Methods	Study design: provider‐randomized trial Unit of allocation: provider Unit of analysis: patient Power calculation: not done
Participants	Care setting: primary care; USA Health professionals: 36 (but 27 contributed patients); various type (general internists, family physicians, nurse practitioners); fully trained Patients: 132; major depressive disorders; male and female
Interventions	Multifaceted intervention (Patient‐centered group): patient‐mediated intervention, educational outreach visit, distribution of educational material, audit and feedback Quote: "Table 1 provides the rationale and expected outcomes for each intervention component and a comparison of the two intervention approaches." page 154 Multifaceted intervention (Standard group): patient‐mediated intervention, educational outreach visit, distribution of educational material Quote: "Table 1 provides the rationale and expected outcomes for each intervention component and a comparison of the two intervention approaches." page 154
Outcomes	Patient rating of their clinicians participatory decision‐making skills (PDM) (categorical)
Notes	Additional information Number of approached patients (eligible): 1486 Number of patients per physician: not reported
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: within each study site (stratum), a randomization schedule was generated through computer by the  study statistician using the Moses and Oakford algorithm page 5, randomization.
Allocation concealment (selection bias)	Unclear risk	Comment: there was an allocation concealment but the method of concealment was not described in sufficient details to make a judgement.
Blinding (performance bias and detection bias)   Participant‐reported outcome	Low risk	Quote: "Interviewers who collected data at 6 and 12 months were masked to clinician and patient intervention assignment. Outcome assessors at 18 months were not blinded to intervention assignment. The 12‐month assessments remained the primary outcome." page 158 paragraph 2
Incomplete outcome data (attrition bias)   Participant‐reported outcome	Low risk	Comment: see patient flow chart of the study protocol  Quote: "Overall, 89 percent (N = 117) of the sample completed the 6‐month interview, 85 percent (N = 113) completed the 12‐month interview, and 55 percent (N = 73) completed the 18‐month interview. Follow‐up rates for standard and patient‐centered groups were as follows: 88 percent versus 90 percent at 6 months and 83 percent versus 88 percent at 12 months. There were no significant differences in characteristics between participants who completed the trials and those who were lost to follow‐up." page 162‐163  In addition: reasons for missing data were similar across groups.
Selective reporting (reporting bias)	High risk	Comment: see table 3 and 4 of the study protocol.
Other bias	Low risk	Comment: no evidence of other risk of biases.
Baseline measurement?   Participant‐reported outcome	Low risk	Comment: see Table 3.
Protection against contamination?	Unclear risk	Comment: allocation was by clinician within a clinic.
Baseline characteristics patients	Low risk	Quote: "patients in then standard group had higher mean scores on readiness for treatment (7.2 vs. 6.6, p = .02). A higher proportion of patients in the patient‐centered group received care from an African American clinician (race‐concordant relationship) (61.2 percent vs. 6.2 percent, p < .001)." page 162  Quote: "However, because race concordance between clinicians and patients is an important predictor of patient‐reported outcomes (Cooper et al. 2003a), it was included as a covariate. Clinician age and patient attitudes were also included as covariates in separate mean models to examine the robustness of main inferences." page 161
Baseline characteristics healthcare professionals	Low risk	Comment: see table 2  Clinician race was taken into account in the analysis through the race concordance.

Methods	Study design: clinician‐randomized, cross‐over trial Unit of allocation: clinician Unit of analysis: clinician Power calculation: done
Participants	Setting of care: primary care, ambulatory care; Norway Healthcare professionals: 72; various types of physician (residents, consultants, medical surgeons, neurologist, podiatrists, gynecologist), fully trained and residents Patients: not reported
Interventions	Multifaceted intervention: educational meeting, distribution of educational materials, Audit and feedback after role‐play Quote: "Doctors participated in the 20 hours (a 45 min) course over two consecutive days. …The course consisted of a 50/50 mix of theory and 45 min group sessions (3–7 participants and two teachers per group) including role‐plays, with plenary debriefs after each group." page 2 "Our course was based on the same content as the 5‐day course Communication Skills Intensive offered by Kaiser Permanente." page 2 "At the conclusion of the course, all participants received a one‐sheet overview of the Four Habits to carry in their pockets as reminder in everyday work." page 3 Usual care (control)
Outcomes	Four Habits Coding Scheme (continuous, score); SDM is assessed as the fostering by healthcare professionals of active participation of patients in the decision‐making process
Notes	Additional information Number of approached patients (eligible): not reported Number of patients per physician: not reported, planned for eight video consultations per physicians
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "The doctors were randomized to receive the intervention in the summer of 2007 or the winter of 2008." page 3
Allocation concealment (selection bias)	Unclear risk	Comment: unit of allocation is the doctor within a clinic. No details on allocation procedure.
Blinding (performance bias and detection bias)   Observer‐based outcome	Low risk	Quote: "Raters were blinded to all information about the doctors and the encounters, including whether the video was made before or after the intervention." page 3
Incomplete outcome data (attrition bias)   Observer‐based outcome	Low risk	Comment: reasons for missing outcome data unlikely to be related to outcome.
Selective reporting (reporting bias)	Unclear risk	Comment: no evidence that outcomes were selectively reported, but no protocol.
Other bias	Low risk	Comment: no evidence of other risk of biases
Baseline measurement?   Observer‐based outcome	Low risk	Quote: "All included doctors had two encounters videotaped before the first course (period A ‐ baseline)." page 2
Protection against contamination?	Unclear risk	Comment: doctor were allocated within a clinic.
Baseline characteristics patients	Unclear risk	Comment: not reported
Baseline characteristics healthcare professionals	Unclear risk	Comment: No comparison between intervention and control group

Methods	Study design: patient randomized trial Unit of allocation: patient Unit of analysis: patient Power calculation: done
Participants	Care setting: tertiary care; ambulatory care, USA Health professionals: 102; physicians, residents; fully trained and in training Patients: 204; chest pain; male and female: 120 females, 84 males
Interventions	Multifaceted intervention: patient‐mediated intervention (one brief demonstration of the use of the decision aid) and educational meeting (one hour training session) Quote: "Participating clinicians were oriented during a 1‐hour training session given by the lead investigator (E.P.H.) as well as a brief (3 min) demonstration from the study coordinator on how to use the decision aid before meeting the first enrolled patient and as needed." page 252 No intervention, standard care (control)
Outcomes	Observing Patient Involvement (OPTION) scores; The fostering by healthcare professionals of active participation of patients in the decision‐making process
Notes	Additional information Number of approached patients (eligible): 310 Number of patients per physician: 208 patients for 51 clinicians
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Patients were randomised to either usual care or shared decision making through a Web‐based, computer‐generated allocation sequence in a 1:1 concealed fashion …" page 253
Allocation concealment (selection bias)	Unclear risk	Comment: the centralized randomization scheme was unclear
Blinding (performance bias and detection bias)   Observer‐based outcome	Low risk	Quote: "Third investigator (H.H.T.), who was also blinded to allocation, reviewed all potentially positive outcomes." page 254
Incomplete outcome data (attrition bias)   Observer‐based outcome	Low risk	Comment: 3 missing outcomes in DA and 5 in control group. Reasons unlikely to be related to our outcome of interest.
Selective reporting (reporting bias)	Low risk	Comment: pre‐specified relevant outcomes were reported in the results (see clinical tria.gov NCT 01077037).
Other bias	Low risk	Comment: no other evidence of risk of biases.
Baseline measurement?   Observer‐based outcome	Unclear risk	Comment: not specified in the paper.
Protection against contamination?	Low risk	Comment: it was the patients who were randomized in an ED Hospital.
Baseline characteristics patients	Low risk	Comment: see table 1 for patient baseline characteristics, page 255
Baseline characteristics healthcare professionals	Unclear risk	Comment: no report of characteristics.