Communication Tools for End-of-Life Decision-Making in Ambulatory Care Settings: A Systematic Review and Meta-Analysis

Background Patients with serious illness, and their families, state that better communication and decision-making with healthcare providers is a high priority to improve the quality of end-of-life care. Numerous communication tools to assist patients, family members, and clinicians in end-of-life decision-making have been published, but their effectiveness remains unclear. Objectives To determine, amongst adults in ambulatory care settings, the effect of structured communication tools for end-of-life decision-making on completion of advance care planning. Methods We searched for relevant randomized controlled trials (RCTs) or non-randomized intervention studies in MEDLINE, EMBASE, CINAHL, ERIC, and the Cochrane Database of Randomized Controlled Trials from database inception until July 2014. Two reviewers independently screened articles for eligibility, extracted data, and assessed risk of bias. Grading of Recommendations Assessment, Development, and Evaluation (GRADE) was used to evaluate the quality of evidence for each of the primary and secondary outcomes. Results Sixty-seven studies, including 46 RCTs, were found. The majority evaluated communication tools in older patients (age >50) with no specific medical condition, but many specifically evaluated populations with cancer, lung, heart, neurologic, or renal disease. Most studies compared the use of communication tools against usual care, but several compared the tools to less-intensive advance care planning tools. The use of structured communication tools increased: the frequency of advance care planning discussions/discussions about advance directives (RR 2.31, 95% CI 1.25–4.26, p = 0.007, low quality evidence) and the completion of advance directives (ADs) (RR 1.92, 95% CI 1.43–2.59, p<0.001, low quality evidence); concordance between AD preferences and subsequent medical orders for use or non-use of life supporting treatment (RR 1.19, 95% CI 1.01–1.39, p = 0.028, very low quality evidence, 1 observational study); and concordance between the care desired and care received by patients (RR 1.17, 95% CI 1.05–1.30, p = 0.004, low quality evidence, 2 RCTs). Conclusions The use of structured communication tools may increase the frequency of discussions about and completion of advance directives, and concordance between the care desired and the care received by patients. The use of structured communication tools rather than an ad-hoc approach to end-of-life decision-making should be considered, and the selection and implementation of such tools should be tailored to address local needs and context. Registration PROSPERO CRD42014012913


Protocol and Registration
The protocol for this review is available in the PROSPERO registry at [http://www.crd.york.ac. uk/PROSPERO/display_record.asp?ID=CRD42014012913]

Eligibility criteria
We included randomized controlled trials (RCTs) or prospective observational studies with a control group (including pre-post post studies in which participants functioned as their own control) published as articles in peer-reviewed journals, restricted to the English language. To be eligible for this review, studies must have included patients over the age of 18, and evaluated a communication tool to assist patients in EoL decision-making, in comparison to a control group. For this study, our definition of "structured communication tool" included traditional decision aids in any format (paper, video, computer, etc.), and other structured approaches to help with decision-making, including organized meeting plans; patient education interventions on EoL care options; or reminders or mailing of ADs. Interventions designed solely for information-sharing (eg. breaking bad news, providing emotional support) were excluded, because although such interventions may affect EoL decision-making, it is not their sole or explicit purpose to do so. Control groups had to receive either use care, a sham intervention, or a minimal/ low intensity intervention.
In this paper, we report findings from eligible studies that were conducted in ambulatory care settings. Studies conducted in the inpatient setting, intensive care unit setting, and studies of educational interventions for improving clinicians' competencies in EoL communication and decision-making will be analyzed and reported separately.
Our primary outcomes were 1) completion of ACP, defined as either completion of an AD or a documented discussion about EoL preferences; 2) concordance between ADs and medical orders for care; and 3) concordance between care desired by patients and the care actually received at EoL. Secondary outcomes included: 1) quality of communication between the patient and family/SDMs; 2) quality of communication between the patient and health care providers (HCPs); 3) patient and family knowledge about EoL care, including options for palliative or intensive care, and knowledge about ADs; 4) health care resource utilization; 5) patient and family satisfaction with EoL care; and (6) for study participants who were exposed to the structured communication tool, the acceptability of the intervention. We subsequently added 'patient preference for life-sustaining treatments' as a secondary outcome as it was reported in many studies, and had relevance as a potential surrogate measure of the actual future use or non-use of life-sustaining treatments.

Information sources and search strategy
We searched the following databases from database inception until July 2014: Medline (1946-July 2014); Embase (1980-July 2014); CINAHL (1982-July 2014); Cochrane Database of Clinical Controlled Trials (2005-July 2014); and ERIC (1966-July 2014). Search terms included: "communication," "decision-making," "end-of-life," "cardiopulmonary resuscitation" (complete electronic search strategies for each database can be found in S1 Table). We also hand searched the reference lists of eligible articles and our personal files to identify further articles for screening. all articles which passed initial screening by either reviewer were then assessed independently and in duplicate for final eligibility using standardized, piloted eligibility forms. Disagreement about study eligibility was resolved by consulting with a third reviewer (JY). When screening for eligibility, reviewers were not blinded to article authors, journal, or results. Kappa statistics were calculated to assess inter-rater reliability of the screening and eligibility phases [3]. Eligible studies were then divided based on study type into outpatient, inpatient, or intensive care unit settings; or educational interventions for clinicians.

Data collection process & data items
Study data were collected using standardized, piloted online forms by the two reviewers (HC and SO). Study authors were contacted to clarify study outcomes and methods when they were unclear in the published document. Data collected included study publication information, study dates and population characteristics, study interventions, our primary and secondary outcome measurements, and study methods required to assess the risk of bias in individual studies.

Risk of bias in individual studies
For RCTs, we assessed risk of bias using the Cochrane risk of bias tool with regard to random sequence generation, allocation concealment, blinding of participants and personnel, incomplete outcome data, and selective reporting [4]. Each domain was assessed independently by both reviewers and reported as being at "high", "low", or "uncertain" risk of bias. Studies were considered to be of "low" risk of bias if assessed as being "low" risk of bias in all domains; "uncertain" if uncertain bias in at least one domain, with no domains at high risk of bias; and "high" if there was high risk of bias in any domain. For studies at 'uncertain' risk of bias, we attempted to contact study authors to clarify the relevant issue(s), and revised the overall study risk of bias accordingly. Disagreement between reviewers about risk of bias was resolved by consulting with a third reviewer (JY). For observational cohort and case-control studies, we used the Newcastle-Ottawa scale to assess risk of bias [5]. For uncontrolled before-after studies, the National Institutes of Health rating system was applied [6].

Synthesis of results & sensitivity analysis
We used Revman 5.3 software to conduct our analyses. For each outcome, similar studies were pooled, with a priority given to randomized trials i.e. data was sought from RCTs first and non-randomized studies were only used in the absence of randomized data. Summarized outcomes (standardized mean difference (SMD) or mean difference (MD) for continuous variables, relative risk (RR) for dichotomous variables) and 95% confidence intervals (95% CI) were calculated using a random-effects model. In our primary analyses, when calculating pooled effect estimates, we restricted to RCTs at low and unclear risk of bias. In sensitivity analyses, we included all studies, including those at high risk of bias, to assess the robustness of our pooled estimates of effect if all studies were included, regardless of risk of bias.

Publication Bias
Publication bias was assessed using visual inspection of funnel plots generated in Revman 5.3, where sufficient numbers of studies existed to permit interpretation [7].

Rating of Quality of Evidence
We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to assess the quality of evidence for each outcome [8]. Outcomes for which the majority of evidence was derived from RCTs was considered to initially be of 'high' quality while those from which the majority of evidence was from observational studies started at 'low' quality, with both types rated up or down after considering the risk of bias across studies (eg. publication bias); potential biases and their direction within each study; and the imprecision, inconsistency, and indirectness of the evidence. GRADE summary of findings tables were generated using the online GradePRO software [9].

Study selection
Initial database searches retrieved 5727 articles. After exclusion of duplicate references and conference abstracts, title and abstract screening resulted in 366 articles selected for full text review (κ = 0.648; 95% CI 0.601-0.695). A total of 121 articles were eligible for our systematic review after full-text review and additional manual reference screening. Of these, 67 reported findings from studies conducted in the outpatient setting and are the subject of this article. (Fig  1)

Study characteristics
Study settings & populations. Studies ranged in publication date from 1991-2014. Sixtytwo of the studies were conducted in North America (93%), 2 in Asia (3%), 2 in Europe (3%), and 1 in Australia (1%). Most studies (n = 39, 58%) were of adult participants with no specific medical condition, and the remaining studies focused on participants with cancer (n = 12 studies), cardiac disease (n = 8 studies), renal disease (n = 8 studies), advanced COPD (n = 7 studies), neurologic disease (n = 3 studies), dementia (n = 2 studies), or HIV (n = 2). Some studies included participants with multiple conditions (n = 8). Just under half of studies explicitly included older adults greater than age 50 (n = 29, 43%). A table describing the characteristics in all included studies can be seen in S2 Table. Interventions. Interventions in the eligible studies included verbal discussions alone (n = 9 studies), paper tools alone (n = 9 studies), verbal discussion with paper tool (n = 20 studies), videos (n = 12 studies), computer programs (n = 4 studies), complex multimodal interventions (n = 10 studies), and interventions directed at HCPs rather than patients or SDMs (n = 3). Studies aimed at HCPs were included in this review because either the unit of randomization was the patient (rather than the HCP), or the studies were not educational in nature (eg. chart reminders to discuss ADs with patients).
Characteristics of excluded studies. Of the 366 studies which underwent full-text review, 224 were excluded. 156 were not relevant to EoL decision-making, 44 were conference abstracts only, 7 were study protocols only, 7 were duplicate articles, 5 did not report any outcomes of interest, 3 were pediatric studies, 2 studies included no comparison arm, and 32 for a combination of reasons, including duplicate or irrelevant studies. A further 101 articles which were purely qualitative in nature, or not based in ambulatory settings, were not included in this analysis and will be reported elsewhere.
3. Concordance between care desired by patients and care received by patients. Only two studies, one considered to be at 'low' risk of bias, the other at 'unclear' risk of bias, reported  Communication Tools for End-of-Life Decision-Making Secondary outcomes (Table 2) 1. Patient preferences for life-prolonging treatments. Seven RCTs considered to be of 'low' or 'uncertain' risk of bias reported patient preferences for life-supporting treatments [11,15,17,19,21,42,43], and found that communication tools reduced patients' stated desire for life-supporting treatment (RR 0.62, 95% CI = 0.41-0.94, p = 0.02; I 2 = 2%). (Fig 7) For this outcome, publication bias was suspected based upon visual inspection of the funnel plot. (Fig 8) In a sensitivity analysis that included three 'high' risk of bias trials, the magnitude of effect was smaller but still statistically significant [25,32,36]. (S1 Fig) 2. Quality of communication between the patient and family/SDM. Measurement of quality of communication between patients and families/SDMs was reported in two ways in the included studies: some reported concordance between patients and SDMs about preferences for EoL care while others used rating scales to assess the quality of communication.
Two RCTs, both considered to be of 'high' risk of bias reported scores assessing the quality of communication between patients and SDM [27,29]. Neither study found a statistically significant difference between intervention and control groups for quality of communication between patients and SDMs.
3. Quality of communication between the patient and HCPs. Only two RCTs of 'low' or 'uncertain' risk of bias reported quality of communication scores for patients and HCPs [16,56], both using the same quality of communication score. In both, the intervention was associated with a statistically significant improvement in the quality of communication score (MD 3.02, 95% CI = 1.26-4.78, P<0.001; I 2 = 51%). (Fig 10) When two 'high' risk of bias studies were included in the sensitivity analysis, the magnitude of effect decreased but remained statistically significant [22,29]  Communication Tools for End-of-Life Decision-Making treatments, such as CPR or mechanical ventilation [11,21,41,42], with a statistically significant improvement in knowledge scores (SMD 0.56, 95% CI = 0.26-0.86, p<0.001; I 2 = 52%). (Fig  11) Two studies considered to be of 'high' risk of bias were also found, and their inclusion in a sensitivity analysis did not appreciably change the estimate of effect [24,34]. (S1 Fig) • VERY LOW 4 5 6 CI: Confidence interval; RR: Risk ratio; OR: Odds ratio. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect. * The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). 1 Suspected publication bias based upon visual inspection of funnel plot.
2 Most information is from studies at 'uncertain' rather than 'low' risk of bias.
3 Large amount of statistical heterogeneity, but between large and small positive treatment effects rather than between positive and negative treatment effects.
4 Most information is from studies at high risk of bias.
5 Most information is from surrogate or variable outcomes rather than from objective and direct outcomes. 6 Insufficient sample to meet optimal information size criteria and 95% CI close to or crosses line of no effect. Four separate RCTs also evaluated the effects of communication tools on patient and family knowledge of advance care planning, two at 'low' risk of bias [43,51] and two of 'uncertain' risk of bias [52,55]. A statistically significant improvement in knowledge scores was seen (SMD = 0.30, 95% CI = 0.12-0.49, p = 0.001: I 2 = 0%), with no change in a sensitivity analysis when one high risk of bias study was included [27]. (Fig 12, S1 Fig) 5. Patient and family satisfaction with EoL care. Only one RCT considered to be of 'low' risk of bias [39] reported patient and family satisfaction with EoL care, finding no differences with the use of communication tools. Four RCTs considered to be of 'high' risk of bias [24,26,27,29] also reported measures of patient and family satisfaction with EoL care, and likewise did not find any significant improvement or worsening in patient or family satisfaction with EoL care.
6. Health care resource utilization. There were two RCTs at low risk of bias that reported health care utilization, one finding a non-statistically significant reduction in the total number of ICU admissions between the intervention and control groups (0 (0%) v. 3 (12%), p = 0.093) [11], and the other finding a reduction in the number of hospitalizations ( [32].
No pooled results were generated due to the heterogeneity of outcomes reported. The overall quality of evidence for resource utilization was considered to be 'low' given the low quality 7. Patient and family acceptability of the intervention. Nine RCTs [11,15,21,34,38,39,42,53,54] reported on patient and/or family acceptability of the EoL communication tool studied. In all studies, the majority of participants found the communication tools acceptable. All studies which compared acceptability of the tools to usual care found them to be equally or more acceptable than usual care [11,15,34,53,54].

Discussion
Our systematic review identified 46 RCTs and 21 observational studies which assessed the effects of structured communication tools on EoL decision-making in a wide variety of adult outpatient populations. Although many of the studies addressed populations with specific comorbidities, the majority of studies addressed populations with no specific medical condition. We found low quality evidence that structured communication tools to assist with EOL decision making in ambulatory care settings may increase the completion of ACP (discussions or AD documentation), very low quality evidence (one non-randomized before-after study) that they may increase concordance of patient preferences with medical orders for the use/ non-use of life sustaining treatments, and low quality evidence that they may increase concordance between care desired and care received at EOL (Table 1).
It is unknown whether this is achieved by i) increased translation of AD documents into the acute care setting; ii) improved patient and/or SDM knowledge of the limits of acute care, resulting in more realistic expectations for aggressive care; iii) improved verbal communication between patients, SDMs, and health care providers; or a combination of the above factors. Our secondary outcomes provide support for the latter two mechanisms, with moderate quality evidence suggesting that communication tools result in improved health literacy for patients and SDMs, and reduce the proportion of patients with preferences for life-prolonging care. Very low quality evidence suggests that the use of structured communication tools improves quality of communication between patients, SDMs, and health care providers (Table 2).
Our review found only very low quality evidence about other 'downstream' outcomes of ACP-namely patient satisfaction with EoL care and health care resource utilization. Only a small number of studies of limited quality reported the long-term effects of the communication tools on these important outcomes, with no major effects seen on patient satisfaction, and variable results upon resource use. The lack of evidence for the effectiveness for the interventions upon these outcomes may be due in part to their use in the ambulatory setting-we would anticipate that a very large sample size and prolonged follow-up period would be required to find a significant effects from what is, in effect, a 'preventative' treatment aimed at avoiding potentially unwanted, invasive care in the future.
We believe that eliciting patient wishes for EoL care planning is an inherently valuable practice. Our review provides evidence (albeit of low quality) that the use of such tools have a "class effect" and may increase completion of ACP (discussions or documentation of ADs). The wide variety of populations and interventions studied in the articles we reviewed make it difficult to identify a single 'best' tool to adopt, especially considering not all of the interventions we studied had large or positive effects. It seems reasonable that implementation of tools should be tailored to the local context (disease population, severity of illness, etc.). Given the low confidence in the effect of these tools on ACP completion and other related outcomes, we suggest that those who choose to implement structured EOL communication tools in clinical practice track performance related to outcome(s) of interest, such as completion of ACP. By doing so, endusers can ensure that these resources are having the desired effect. Finally, one of the difficulties in our review was the limited reporting of important clinical outcomes. We would support a standardization of outcomes for studies of advance care planning. A structured approach to ACP outcomes should include measures of knowledge (patient and SDM understanding of ACP), process (increased completion of ADs, improved communication between patients, SDMs, and HCPs) and outcome (concordance between care desired and received; patient satisfaction; and resource use). These outcomes should also be considered at the patient level (eg. patient preferences), SDM level (eg. concordance of SDM with patient preferences), and system level (eg. concordance between care desired and care received). Future research should focus on simple interventions, and study their use across a wide variety of patient populations. As the documentation for ACP can vary from region to region, and change over time, the interventions should ideally focus on eliciting values and preferences for EoL care in general, rather than focus on completing the specific document used in the investigator's region.

Strengths
The strength of our study lies in its rigorous search strategies; the use of two separate authors in assessing studies for screening, eligibility, and risk of bias, with secondary checking and verification of data extraction; as well as in our use of GRADE to assess the overall quality of evidence for each outcome. Our study also explicitly included a wide variety of interventions to assist in EoL decision-making, including traditional decision aids, structured meetings, and educational interventions, allowing us to review the full spectrum of tools which have been published in peer-reviewed journals, some of which may not have been traditionally identified as a 'decision-aid.'

Limitations
Our study has two major limitations, one related to our methods, and the second related to the studies we found. Firstly, identifying studies of interventions to facilitate EoL decision-making is challenging due to a lack of consistent terminology for such interventions. In many cases, only by carefully reviewing a study's methods and outcomes could it be determined that the intervention's purpose was to facilitate EoL decision-making. Evidence of this difficulty is seen in the large number of studies found by hand-searching in addition to our computerized literature searches. Given these difficulties, it is possible that our review failed to identify some potentially relevant articles, despite our rigorous search.
Secondly, our review is limited by the highly heterogenous nature of the populations and interventions studied. Across studies, 'usual care' varied between no intervention and complex encouragement of ADs, depending on local practice. Despite the wide variety of populations and interventions, the studies generally revealed either positive or neutral effects for all of our outcomes of interest, suggesting that the use of a structured communication tool has a "class effect" and is overall more likely to lead to improved outcomes compared to less-structured approaches used during usual care.

Conclusions
A wide variety of communication tools for EoL decision-making have been evaluated in many outpatient populations. Overall, the available evidence suggests that structured communication tools to assist in end-of-life decision-making may improve communication processes and some downstream patient level outcomes, but uncertainty about the true magnitude of effect remains because of the low quality of the existing evidence. While awaiting more rigorous evaluation, use of structured communication tools, rather than ad-hoc discussions about the end-of-life, should be considered. Given the heterogeneity of populations, interventions, and effects, more work is needed to guide the selection, adaptation, and tailored implementation of such tools to local care settings and contexts. Effectiveness of implementation efforts needs to be monitored to assess the success of such interventions.