Respecting Choices® and Related Models of Advance Care Planning: A Systematic Review of Published Evidence

Abstract

All individuals should receive care consistent with their expressed preferences during serious and chronic illnesses. Respecting Choices® is a well-known model of advance care planning (ACP) intended to assist individuals consider, choose and communicate these preferences to health care providers. In this systematic review, we evaluated the published literature on the outcomes of the Respecting Choices® and derivative models utilizing criteria developed by the Cochrane Collaborative.

Eighteen articles from 16 studies were included, of which nine were randomized control trials, six were observational and one was a pre-posttest study. Only two specifically included a minority population (African American). Fourteen were conducted in the United States, primarily in the Wisconsin/Minnesota region (n=8). Seven studies examined the Respecting Choices® model, while nine examined derivative models. There was significant heterogeneity of outcomes examined.

We found that there is a low level of evidence that Respecting Choices® and derivative models increase the incidence and prevalence of Advance Directive and Physician Orders for Life-Sustaining Treatment (POLST) completion. There is a high level of evidence that Respecting Choices® and derivative models increase patient-surrogate congruence in Caucasian populations. The evidence is mixed, inconclusive and too poor in quality to determine whether Respecting Choices® and derivative models change the consistency of treatment with wishes and overall healthcare utilization in the end of life. We urge further studies be conducted, particularly with minority populations and focused on the outcomes of preference-congruent treatment and health care utilization.

Introduction

The goal of advance care planning (ACP) is to ensure that individuals receive care consistent with their expressed preferences during serious and chronic illnesses.¹ Unfortunately, communication about preferences frequently either does not occur, or occurs very late in the course of an individual’s illness trajectory, sometimes mere days or weeks before death.^2,3 This leaves many individuals and families struggling to make decisions in a pressured, emotional and time-sensitive situation and leads to preference-discordant care.^4,5

Thus a key recommendation of the 2014 Institute of Medicine (National Academy of Medicine, Engineering and Science) report, Dying in America: Improving Quality and Honoring Individual Preferences Near the End of Life, was to normalize end of life (EOL) conversations through engagement of persons and families in care planning over time.⁶ The report also indicated a clear distinction between an advance directive—a static document for expressing end-of-life wishes and ACP—a process of care planning over time⁶ (Table 1). Similarly, a recent Delphi survey of palliative care experts defined ACP as a process of care “that supports adults at any age or stage of health in understanding and sharing their personal values, life goals, and preferences regarding future medical care.” ^{1 (p. 821)} ACP acknowledges that individuals’ preferences often evolve as their health status worsens and the conversation around preferences should also evolve.^7–9

Table 1.

Advance Care Planning Terms

Advance Care Planning1 “The whole process of discussion of end-of-life care, clarification of related values and goals, and embodiment of preferences through written documents and medical orders. This process can start at any time and be revisited periodically, but it becomes more focused as health status changes. Ideally, these conversations (1) occur with a person’s health care agent and primary clinician, along with other members of the clinical team; (2) are recorded and updated as needed; and (3) allow for flexible decision making in the context of the patient’s current medical situation.”.
Terms Related to Advance Care Planning1
Advance Directives	“Refers to several types of patient-initiated documents, especially living wills and documents that name a health care agent”. “Living will—a written (or video) statement about the kinds of medical care a person does or does not want under certain specific conditions (often “terminal illness”) if no longer able to express those wishes”. Durable power of attorney for health care—identifies the person (the health care agent) who should make medical decisions in case of the patient’s incapacity.
Medical Orders	“Are created with and signed by a health professional, usually a physician (in some states, a nurse practitioner or physician assistant), for someone who is seriously ill. Because they are actual doctor’s orders, other health professionals, including emergency personnel, are required to follow them.” “Physician Orders for Life-Sustaining Treatment (POLST)a—physician orders covering a range of topics likely to emerge in care of a patient near the end of life.. The orders cross care settings and are honored in the community in an emergency.” aThe names of similar forms in different states vary.”

¹Definitions from, IOM. Dying in America: Improving quality and honoring individual preferences near the end of life. Washington, DC: The National Academies Press; 2014.

Respecting Choices® is one well-known model intended to normalize end-of-life communication.⁶ The Respecting Choices® model started in LaCrosse, Wisconsin in 1991 as a community-wide initiative of integrating ACP.¹⁰ The key features of the program include a stepped-approach to ACP (First Steps®, Next Steps®, Last Steps®) and in-depth counseling by trained non-physician facilitators. The program features include treating ACP as an ongoing process; shifting focus from completion of the documents and towards facilitating discussion about values and goals; shifting the locus of ACP away from hospitals and physicians into the community and family; training of non-physician community volunteers; refocusing discussion of preferences in terms of personal relationships and away from individual autonomy; and ensuring that completed advance directives are available in patients’ charts.¹¹ Later, the model was expanded to include the Physician’s Order for Life-Sustaining Treatment (POLST) paradigm.¹² POLST is a communication process that emphasizes shared decision-making based on the patient’s current health status, prognosis, and goals and results in the completion of a set of medical orders. POLST is not an advance directive based (AD) on future care preferences.³⁴

The Respecting Choices® program showed very promising early results. Of 540 decedents in the Gundersen Lutheran Medical Center two years post-implementation, 85% had a written advance directive which was found in the patients’ medical record in 95% of cases.¹³ Over the years, several studies of Respecting Choices® and its adaptations have been reported in the literature and there have been efforts to scale the model to the national and international levels¹⁴, including partnering with the Coalition to Transform Advanced Care (C-TAC) as part of its Service Affiliate C-TAC Innovations.¹⁵ However, despite the promise of widespread integration, the published evidence for the Respecting Choices® model has not been adequately synthesized to inform its broad adoption across care settings.

The purpose of this systematic review (1991–2017) is to synthesize the published evidence on Respecting Choices® model and related models adapting Respecting Choices® for types of ACP outcomes studied, population and settings of implementation, and effectiveness of the model in improving various ACP and end of life outcomes. We also evaluate the published findings for methodological rigor and study characteristics.

Respecting Choices Model Description:

The Respecting Choices® model is a copyrighted systematic program licensed by Gundersen Health System designed to promote ACP. The model utilizes First Steps®, Next Steps® and Last Steps® components corresponding to different phases of life and illness trajectory (wellness, illness diagnosis and advanced, serious illness). To be certified as a Respecting Choices® trainer or facilitator, a 20 hour training is required³⁵. Facilitators can be trained only on First Steps®, Next Steps® or Last Steps® for the Respecting Choices® model and are trained only in Next Steps® and Last Steps® for DS-ACP and PC-ACP as these models are intended to be used in people who already have a serious life-limiting illness.

Methods

Search Methods and Databases:

The systematic review methods were adapted from the process developed by the Cochrane Collaboration.^16–19 The search was conducted between November 2016 to April 2017 using PubMed, CINAHL and Google Scholar electronic databases. PubMed is one of the leading biomedical databases, while CINAHL has a focus more inclusive of nursing and psychology. To provide the broadest results, Google Scholar was also utilized. The keywords included, “Respecting Choices”, “Honoring Choices”, “Disease-Specific Advance Care Planning”, Patient-Centered Advance Care Planning, combined with “Advance Care Planning”, “End of life”, and “Goals of Care”. The reference lists of articles were reviewed to identify any additional articles.

Criteria for Inclusion of Studies:

Published, peer-reviewed, empirical studies testing the Respecting Choices® (RC) model and close derivatives including Honoring Choices, Disease-Specific Advance Care Planning (DS-ACP) and Patient-Centered Advance Care Planning (PC-ACP) were included in the review. Honoring Choices and DS-ACP are known variants of Respecting Choices®. DS-ACP was developed by the RC program to tailor the RC model to patients with specific life-limiting diseases, including heart failure and end-stage renal disease.²⁰ The PC-ACP is also tailored towards patients with life-limiting diseases and combines the RC model with the theories of interactive decision-making and the representational approach to patient education.²¹ Only articles published in English language were included. We excluded review articles, those that did not test a patient-centered outcome (e.g. discussion of implementation without examination of results or evaluation of facilitator comfort; or those that only evaluated the feasibility and acceptability of the intervention). We excluded articles that presented a protocol only, but when a protocol was found, we followed up to see if the final study had been published. We also excluded book chapters, unpublished dissertations and conference proceedings.

Study Quality and Assessment of Bias:

The study quality was evaluated using GRADE criteria, and the tool for assessing risk of bias developed by the Cochrane Collaborative was utilized for assessing the risk of bias in each study.^16–19,22

Results

Description of Studies

The search of the electronic databases (conducted November 2016-April 2017) initially yielded 745 unique articles. The study selection and screening process is denoted in Figure 1. Eighteen articles from 16 studies were included in the final review.

Figure 1.

Respecting Choices.

Study characteristics

Of the included studies, nine were randomized control trials (RCTs) ^21,23–33 (five of which were pilot studies or smaller studies with less than 32 participants or dyads^{21,24,29,30,32,33}), six were observational^20,34–38 and one was a pre-post test study design.³⁹ Fourteen were conducted in the United States, primarily in Wisconsin and/or Minnesota (n=8) and four even more specifically in La Crosse, Wisconsin.^{21,25,26,33,34} Two were conducted internationally (Australia and Germany).^23,35

Participant characteristics

Participant characteristics such as race and ethnicity were not consistently reported; two studies were predominantly or exclusively with Blacks/African Americans,^24,32 while five were predominantly or exclusively with Whites/Caucasians.^{21,25,26,33,36} Two of the studies were conducted with adolescent samples,^27–30 while the majority (n=14) were performed with adults or older adults. Eight were focused on a particular diagnosis group such as cardiovascular or renal disease (n=5),^{20,21,25,26,32,33,40} HIV (n=1)²⁸ or cancer (n=2)^29,30,38 while six studies did not limit by specific diagnosis (Table 2).

Table 2.

Study Characteristics and Summary of Findings

First Author Year	Model & RC Step	Outcomes	Study Design	Sample	Measures	Results	Location	Interventionist
Briggs 2004	PC-ACP: Next Steps	Knowledge of ACP, Patient-surrogate congruence, decisional conflict, quality of patient-clinician communication	Randomized control trial (RCT) stratified by diagnosis	N=27 dyads Clients with heart failure, end-stage renal disease or pre-open heart surgery	Statement of Treatment Preferences, Treatment Decision-making Role Preference, Decisional Conflict Scale, Knowledge instrument	Greater patient-surrogate congruence, decreased Decisional conflict, higher rating of quality of patient-clinician communication in intervention group. No differences in ACP knowledge.	La Crosse, WI	Nurse
Detering 2010	Respecting Choices – adapted to Australia: Next Steps	Consistency of treatment with wishes, satisfaction with hospital stay; surrogate mental health post-bereavement	RCT	N=309 80+ year olds admitted under internal medicine, cardiology or respiratory	Chart review Post-death surrogate interview, Hospital anxiety Depression Scale (HADS) for anxiety, depression	Higher ACP completion rate, greater likelihood of having wishes known and respected, higher satisfaction in intervention group.	Australia	Nurse or allied health worker
Hall 2014	PC-ACP: Last Steps	Number of patients with completed POLST, resident and surrogate satisfaction with process	Pilot, pre/post-test	N=8 dyads Assisted Living residents with <1 year life expectancy, cognitively intact with surrogate available	Chart Review Quality of Patient-Clinician Communication about EOL Care	All patients completed POLST form. 90% rated quality of interaction as positive.	Minnesota	Nurses
Hammes 2010	Respecting Choices: Next steps & Last Steps	AD prevalence, POLST prevalence, consistency of treatment with wishes	Retrospective, comparative, observational study comparing cohorts from 2007/8 to 1995/96	N=940 (cohort 1: 540; cohort 2: 400); Adult decedents	Chart Review	Greater AD prevalence in later cohort (90% vs. 85%). Little difference in consistency of treatment with wishes.	La Crosse County, Wisconsin	Unknown
Huang 2016	Respecting Choices +Motivational Interviewing +health literacy adapted AD: First Steps	Satisfaction with ACP; Knowledge of ACP, Intent to complete AD	Pilot RCT, mixed-method	N= 30 dyads; ≥45y/o; community-dwelling African Americans	RC Program Evaluation Questionnaire, Knowledge Instrument, 1 question on intent to complete (Y/N)	No difference in intent to complete (100% both groups),no difference in Knowledge.	Deep South	Certified Respecting Choice Facilitator
In der Schmitten 2014	Respecting Choices – adapted to Germany: Next Steps & Last Steps	Incidence of new ADs, prevalence of ADs. Incidence and prevalence of physician orders on life-sustaining treatment(German equivalent of POLST).	Cohort Case-control study between regions (intervention and control)	N=136 intervention (3 sites); 439 control (10 sites); Nursing home residents	Chart review	Increased ACP completion, number of proxy designations, and documentation of code status in intervention sites.	Germany	Trained non-physician facilitators Trained primary care physician facilitators
Kirchhoff 2010	PC-ACP: Next Steps	Knowledge about ACP, Patient-surrogate congruence, satisfaction with process	RCT, stratified by site and diagnosis	N= 313 Adults with congestive heart failure and end-stage renal disease with life expectancy < 2 years	Knowledge about ACP tool, Statement of Treatment Preferences, Quality of Patient-Clinician Communication (Intervention only)	Greater congruence and knowledge in intervention than control. Intervention group highly satisfied (satisfaction not evaluated in control)	Wisconsin (La Crosse & Madison)	Nurses, social workers, and a chaplain
Kirchhoff 2012	PC-ACP: Next Steps	Consistency between treatment and expressed wishes	RCT, stratified by site and diagnosis	N= 313 Adults with congestive heart failure and end-stage renal disease with life expectancy < 2 years	Phone interview with surrogate	No significant differences in consistency of treatment with wishes between two groups; greater dialysis withdrawal for ESRD patients in intervention group.	Wisconsin (La Crosse & Madison)	Nurses, social workers, and a chaplain
Lyon 2009	Adapted DS-ACP +Lyon ACP Survey+5Wishes: Next Steps & Last Steps	Patient-surrogate congruence, Decisional Conflict, Satisfaction with ACP process	RCT	N= 38 dyads HIV+ adolescents and adult guardians	Statement of Treatment Preferences, Decisional Conflict Scale, Quality of Patient-Interviewer Communication Scale,	Increased congruence, satisfaction and adolescent knowledge in intervention group; lack of clear reporting on Decisional Conflict.	Washington DC & Memphis, TN	Trained research assistant (discipline unspecified)
Lyon 2013	RC+Lyon Family-Centered ACP Survey+5 Wishes: Next Steps & Last Steps	Patient-surrogate congruence, Decisional conflict, quality of communication	RCT	N=30 dyads; Adolescents with cancer & their adult guardians	Statement of Treatment Preferences, Decisional Conflict Scale, Quality of Patient-Interviewer Communication Scale,	Greater patient-surrogate congruence in intervention group; adolescents felt better informed about ACP in intervention group; no difference in decisional conflict or quality of communication	Washington DC	Trained research assistant (discipline unspecified)
Lyon 2014	RC+Lyon Family-Centered ACP Survey+5 Wishes: Next Steps & Last Steps	Quality of life, accessibility of ACP in chart, spirituality, satisfaction	RCT	N=30 dyads; Adolescents with cancer & their adult guardians	Satisfaction questionnaire, Beck Anxiety & Depression Inventories, Pediatric Quality of Life Inventory, Spiritual Well-being Scale, chart review	Intervention families were satisfied (no comparison to control group), Anxiety increased in intervention group vs. control over time; depression decreased in intervention vs. control over time. 100% completion of 5 Wishes document in Intervention vs. 0% in control. Increased spirituality in intervention.	Washington DC	Trained graduate students (discipline unspecified)
Moorman 2012	Respecting Choices: Next Steps	Social diffusion of RC: AD prevalence in those with residential access to La Crosse area health systems, those with social access and those with none.	Secondary analysis of survey data – Wisconsin Longitudinal Survey	N=5, 106; >65 y/o Wisconsin high-school graduates – all white.	Dichotomous Survey questions	High levels of ACP overall (55% Living wills; 52% power of attorney, 74% informal discussions). Those with residential access less likely to have ACP. No evidence for social diffusion.	Wisconsin	Unspecified
Pecanac 2014	Respecting Choices: Next Steps	Prevalence of ADs; consistency between treatment and expressed wishes	Retrospective, comparative chart review of pre-RC implementation (2005–2007) and post-implementation (2008–2010)	N= 732 Adult decedents who died in hospital (excluded Emergency Department deaths)	Chart review	Most decedents had ADs prior to hospitalization.. No change in AD prevalence when examining all decedents or white decedents only. Significant increase in ADs in racial minorities post-RC implementation. No change in consistency between wishes and care provided.	Milwaukee, Wisconsin	Chaplains, social workers, and ethicists
Rocque 2017	Respecting Choices: First Steps (although population appropriate for Next Steps)	Number of documented ACP conversations; End of life healthcare utilization	Mixed-method prospective evaluation	N=8,704 Adult cancer patients	Chart review and claims data for hospitalizations, Emergency Department (ED) visits, Intensive care (ICU) admissions, chemotherapy and hospice use	Navigators initiated ACP discussions with 15% of patients. Decreased hospitalization rates in those who had ACP conversation. No difference in hospice. No significant difference in ED/ICU/Chemotherapy use	Southern US	Lay navigators
Schellinger 2011	DS-ACP: Next Steps	Type of documentation related to ACP noted in Electronic Health Record; healthcare utilization	Longitudinal, comparative (participants vs. non-participants)	N=1894 (602 participants, 1292 non-participants) Adults with heart failure	Chart review	31.8% of referrals participated. Participants has 94.3% ACP completion rate vs. 24.8% of non-participants. No difference in hospital 30 or 60 day readmissions. Among decedents, increased hospice enrollment and longer hospice length of stay for those who participated.	Minnesota & Wisconsin	Nurses and social workers
Schwartz 2002	Respecting Choices: Next Steps	Patient-surrogate congruence; Knowledge on ACP; surrogate’s comfort with role; patient preferences	RCT	N=61 dyads Older adults	Medical Directive Questionnaire; Knowledge Questionnaire; Agent Comfort Questionnaire; Beliefs & Values Questionnaire	Increased congruence, knowledge and surrogate comfort in intervention group.	Massachusetts	Nurse
Song 2005	PC-ACP: Next Steps	Patient-surrogate congruence, Knowledge on ACP; Anxiety around ACP; Decisional Conflict	RCT	N=32 dyads; Clients from cardiothoracic surgery clinic	Statement of Treatment Preferences; Spielberger’s State Anxiety Inventory; Decisional Conflict Scale	Increased congruence in intervention group; no difference in anxiety or knowledge; decreased decisional conflict in intervention.	La Crosse, Wisconsin	Nurse
Song 2010	PC-ACP: Next Steps	Decisional conflict; Patient-surrogate congruence; surrogate comfort with role; psychosocial well-being of dyad	RCT	N=17 dyads; African Americans with end-stage renal disease recieving dialysis	Statement of Treatment Preferences; Decisional Conflict Scale; Decision-making Confidence Scale	Increased congruence in intervention group. Intervention group had greater preference for life-sustaining treatment than Control group. No differences in decisional conflict or surrogate comfort or psychosocial well-being.	United States	Nurse

Interventions

Seven completed studies examined the RC model,^{23,31,34–38} while 9 examined RC adaptations, which included RC combined with Five Wishes or Motivational Interviewing (n=2),^28–30 PC-ACP (n=4)^{21,25,26,32,33} and DS-ACP (n=3)^20,27–30. None of the eligible studies examined the Honoring Choices® model.

The Use of First Steps®, Next Steps® and Last Steps®.

Only one study included healthy adults and appropriately utilized the First Steps® component.²⁴ A second study involved adults with cancer, who would be appropriate to Next Steps®, but the interventionists were trained with First Steps®.³⁸ One study included nursing home residents with a less than 1 year life expectancy and utilized the Last Steps® component.³⁹ The other studies enrolled patients with life-limiting illnesses and used either Next Steps® exclusively (n=10 studies)^{20,21,23,25,26,31–33,36–37} or a combination of Next Steps® and Last Steps® (n= 5 studies)^{28–30,34,35}.

Interventionist Type and Training

There was wide variability in interventionist discipline and training and in many cases, there was inadequate information provided on the interventionist. Registered nurses (RNs) were utilized as interventionists in five studies;^{21,31–33,39} in another five studies, the interventionist was a nurse or an allied health staff member (social workers, chaplains).^{20,23,25,26,37} Lay navigators were utilized in one study.³⁸ In six studies, the discipline of the interventionist was not specified or was unknown.^{24,28–30,34,36} In nine articles, it was specified that the interventionist received official Respecting Choices® training,^{23–26,28–30,35,38} but it was noted in one study that not all facilitators were fully trained³⁷ and another six articles did not specify whether and how facilitators were trained.^{20,21,31,34,36,39}

Study Outcomes

Table 2 specifies the outcomes investigated in each study. There was significant heterogeneity in the outcomes employed across studies. The most commonly evaluated outcomes were patient-surrogate congruence in treatment choices (n=8 studies)^{21,25,26,28,29,31–33}, patient and/or surrogate satisfaction with the ACP intervention or process (n=7 studies)^{24–26,28,30,35,39}, and AD completion rates (n=5 studies)^30,34–37.

Risk of Bias in Included Studies

Table 3 summarizes the risks of bias identified in each study. The most frequent bias identified was selection bias (n=14; 87.5%). Among the RCTs (including pilots), one study had noted differences between intervention and control groups that were not adjusted for statistically;²¹ while several others did not specify whether there were significant differences between the two groups or did not provide enough information on the two groups.^25,26 In one study, the intervention group had a higher likelihood of previous ACP exposure,³³ which could strongly bias the outcome. In addition, among the observational studies, two compared participants to non-participants^20,38 and in one, there was a racial difference in who was offered the opportunity to participate.³⁸

Table 3:

Risk of Bias

Study	Selection Bias	Attrition Bias	Performance Bias	Measurement Bias	Reporting Bias	Overall Risk Judgment
Briggs	+	−	+	+	−	High
Detering	−	+	+	−	+	High
Hall	+	+	−	+	−	High
Hammes	+	−	−	?	−	Low
Huang	−	−	+	?	−	Moderate
In der Schmitten	+	+	−	+	−	Moderate
Kirchhoff 2010 & 2012	?	+	+	−	−	High
Lyon 2009	−	−	−	+	−	Low
Lyon 2013/2014	+	?	+	−	−	Moderate
Moorman	+	−	−	−	−	Low
Pecanac	+	−	−	?	−	Low
Rocque	+	−	−	+	−	High
Schellinger	+	−	+	+	−	High
Schwartz	?	?	+	+	−	High
Song 2005	+	−	+	+	−	High
Song 2010	?	?	+	−	−	High

Legend: (+) = positive bias, (−)= no identified bias, (?) = lack of information

Performance bias was also a major issue for 13 of the 16 studies. The major issue contributing to performance bias was that the control group in the RCTs were generally only offered an educational pamphlet on ACP or nothing at all,^{21,23–26,29–31,33} versus in-person sessions with an interventionist received by the intervention group. One study offered a time and attention-equivalent to the control group on topics other than ACP.²⁸ There were no comparative effectiveness studies directly comparing RC or a derivative model to another form of ACP intervention. In addition, completing an AD or POLST was part of the intervention in three studies and yet AD and/or POLST completion was also measured as an outcome, making it an unavoidable confounder.^39,30,35

Poor reporting made it difficult to assess measurement bias across the studies. In some studies, the pre- and post-intervention measures were collected by research assistants other than the interventionist – but it was not always clear if they were blinded.^28–30 In addition, one study measured outcomes twice for the intervention group (pre and post-intervention), but only once for the control group (if they did not receive anything).^25,26

The reporting on attrition was not clear enough in three studies to determine the risk of attrition bias.^29–32 In one article, it was unclear if an outcome was statistically significantly different between the intervention and control groups as p-values were not reported.²³

Evidence Synthesis

The goal of ACP is to improve communication about end of life preferences and ultimately, to improve the rate of preference-concordant treatment. In addition, greater preference-concordant treatment is often linked to decreased EOL healthcare utilization and this is seen as a benefit of ACP. Thus, there were four outcomes that we judged to be of key importance to this body of literature: the AD and/or POLST completion rate (incidence and/or prevalence), patient-surrogate congruence, consistency between treatment and expressed wishes, and EOL healthcare utilization.

There were five studies that examined the incidence and/or prevalence of ADs as an outcome^30,34–37 – four assessed this outcome via chart review, and one measured it via self-report on survey questions. One study was an RCT, which included AD completion as part of the protocol³⁰ and one was a well-designed cohort case-control study with a high risk of bias.³⁵ Three of the five studies showed an increase in AD incidence and/or prevalence in the intervention group or post-intervention.

Three studies measured POLST completion rate.^34,35,39 One was a well-designed cohort case-control study with a high risk of bias;³⁵ one was observational³⁴ and one was a pre-post-test study which included POLST completion as part of the study protocol.³⁹ All three studies demonstrated a positive impact of RC and derivatives on the prevalence of POLST orders. We conclude that there is low-level evidence that RC or derivative models positively impact the completion of ADs and POLST orders when compared to an inactive control when used in Caucasian populations.

There were seven studies that examined the outcome of patient-surrogate congruence.^{21,25,26,28,29,31–33} Of these, six measured it using the Statement of Treatment Preferences while one utilized the Medical Directive Questionnaire.³¹ These tools measure congruence by asking patients and surrogates separately to make choices in hypothetical scenarios. All of these were RCTs and all consistently demonstrated greater patient-surrogate congruence in the intervention group compared to the control group. We conclude that there is high-level evidence that RC or derivative models positively impact patient-surrogate congruence when compared to inactive control when used in Caucasian populations. There is low-level evidence (one study)³² for the same conclusion among African Americans.

There were four studies that examined the outcome of consistency between treatment and expressed wishes – three with chart review,^23,34,37 while one utilized a phone interview with the family surrogate.²⁶ Two of these were randomized control trials^23,26 and two were retrospective chart reviews.^34,37 Of these, one (performed in Australia) found an increase in consistency in the intervention versus control group²³ and the other three (performed in the United States) found no differences in consistency of treatment with expressed wishes. We conclude that there is mixed evidence that RC and its derivative models have any effect on the consistency of end of life treatment with expressed patient wishes.

Only two studies examined healthcare utilization;^20,38 both compared those who participated in ACP with those who did not. Both measured this utilizing chart review and/or claims data. One found no difference in hospitalizations, but greater hospice admission and longer hospice length of stay for those who participated.²⁰ The other found decreased hospitalization rates among participants with no significant differences for emergency department visits, intensive care stays or chemotherapy use.³⁸ Both studies had a high risk of bias. We conclude that there is mixed, very low-level available evidence that RC and its derivative models have any effect on healthcare utilization at the EOL.

Discussion

Respecting Choices® and its derivative models have been widely implemented in Wisconsin and Minnesota. However, in synthesizing the published evidence, we found that the evidence for RC is not as strong as might be thought. For the immediate outcome of increasing AD and POLST completion, there is a low level of evidence in support of RC and derivative models. For the outcome of patient-surrogate congruence in making choices in hypothetical scenarios, there is a high level of evidence in support of RC and related models among Caucasians (and low level of evidence among African Americans). However, for the long-term outcomes of AD and ACP on consistency of treatment with wishes and overall EOL healthcare utilization, the evidence was mixed, inconclusive and generally low quality.

Respecting Choices® models are generally found to increase the likelihood that an AD or a POLST will be completed and that it will be able to be located in the patient’s chart. The incidence and prevalence of ADs have been one of the most common indicators that healthcare systems have used to evaluate their ACP programs.⁴¹ This is not surprising, as having a written document is one of the easiest tools to measure the outcomes of the ACP process. It is somewhat unclear, however, which patients are likely to willingly participate versus those unwilling to participate. It is also unclear if there is a subset of population that is more likely to benefit from the time and resource-intensive nature of the RC model.

Based on this review, the evidence that RC models increase patient-surrogate congruency is the strongest of the outcomes that were examined. The idea behind patient-surrogate congruence is that the better the surrogate knows the patient’s wishes, the more effective they will be as a surrogate – and potentially, the more comfortable they will feel in the surrogate role. Like the presence of ADs, patient-surrogate congruence is relatively easily measured. Little is known about whether patient-surrogate congruence measured on hypothetical scenarios translates to more effective surrogate decisions when confronted with real-life decisions or whether surrogate comfort level is maintained amidst real-life decisions. Given that individual preferences evolve over time and with illness and treatment exposure, it is unclear whether patient-surrogate congruence also evolves.^7–9 We must also note that there is little data on patient-surrogate congruence among other racial or ethnic groups besides Caucasians.

A major challenge to the use of ADs is the charge that they “fail” or do not work – either in protecting a patient’s wishes at the end of life or in reducing unnecessary or futile healthcare utilization at the end of life. Based on the reviewed evidence, Respecting Choices® and its derivative models are liable to the same charge. There is mixed, low-level evidence on the impact of RC on either the consistency of treatment with expressed wishes or overall end of life healthcare utilization. This is not surprising, as a systematic review of how healthcare systems evaluate ACP initiatives found this to be a weakness generally.⁴¹ Further studies are needed to better examine whether or not ADs in general – and RC models in particular – can have a clinically meaningful difference in honoring patient choices and reducing unwanted treatment in the EOL.

Importantly, we note that RC models were never tested against another method of ACP. This is important as RC is a resource-intensive model and the program implementation fee can range from $190,000 to $500,000, making it difficult to scale the model.⁴² Furthermore, there was some variance in which components of RC (e.g. First Steps®, Next Steps®, Last Steps®) were tested. Further testing of RC is needed to determine which aspects of the model are most impactful – and for whom. In addition, further testing in diverse geographical regions of the United States and with other minority populations, such as Latinos and Asians, are needed.

Strengths and Limitations

Although we used a comprehensive search strategy with broad inclusion criteria, it is possible that relevant studies may have been missed. We utilized the criteria developed by the Cochrane Collaborative for assessing the risk of bias, but there is still some subjectivity involved in determining risk of bias.²² In addition, we focused on patient/family centered outcomes, which may or may not be relevant to various stakeholders. Due to scarcity of published evidence on RC model, we included RC and derivative models in this analysis. Future studies may focus specifically on the outcomes of the RC model as evidence accumulates.

Conclusions

The Respecting Choices® and derivative models of ACP likely increase patient-surrogate congruence on hypothetical EOL scenarios and are also likely to increase the prevalence of AD and POLST completion. However, there is not enough evidence that either the Respecting Choices® and derivative models or the completion of ADs and POLST orders actually change the quality of EOL care or the congruence between actual treatment received and patient expressed preferences. We urge that further testing of the Respecting Choices® and derivative models be conducted, particularly with a focus on outcomes that reflect quality of EOL care received (eg consistency between care received and expressed preferences, healthcare utilization, surrogate distress post-death). The model should also be evaluated as part of comparative effectiveness studies, especially against lower cost models to see if comparable ACP outcomes may be arrived at using low resource interventions.