The squares and horizontal lines correspond to the study-specific risk ratio (RR) and 95% CI. The diamond represents the pooled RR of overall preference. The vertical dashed line indicates the overall pooled RR of 0.70. CPR indicates cardiopulmonary resuscitation.
The squares and horizontal lines correspond to the study-specific standardized mean difference (SMD) and 95% CI. The diamond represents the pooled SMD of patient knowledge. The vertical dashed line indicates the overall pooled SMD of 0.55.
eTable 1. Search Strategy for PubMed
eTable 2. Risk Assessment by Cochrane Risk of Bias Tool
eFigure. Flow of Studies Through the Review Process
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Becker C, Lecheler L, Hochstrasser S, et al. Association of Communication Interventions to Discuss Code Status With Patient Decisions for Do-Not-Resuscitate Orders: A Systematic Review and Meta-analysis. JAMA Netw Open. Published online June 07, 20192(6):e195033. doi:10.1001/jamanetworkopen.2019.5033
Is there an association between communication interventions and patient preference regarding do-not-resuscitate (DNR) code status decisions and knowledge regarding life-sustaining treatment?
In this systematic review and meta-analysis, the pooled meta-analysis of 11 randomized clinical trials involving 1463 patients showed a significant association between communication interventions and higher patient preference for a DNR code status. In an analysis of 5 eligible trials, communication interventions were also associated with better patient knowledge about resuscitation.
Communication interventions may be an effective decision aid for code status discussions that potentially alter patient decisions regarding DNR code status and increase patient knowledge.
Whether specific communication interventions to discuss code status alter patient decisions regarding do-not-resuscitate code status and knowledge about cardiopulmonary resuscitation (CPR) remains unclear.
To conduct a systematic review and meta-analysis regarding the association of communication interventions with patient decisions and knowledge about CPR.
PubMed, Embase, PsycINFO, and CINAHL were systematically searched from the inception of each database to November 19, 2018.
Randomized clinical trials focusing on interventions to facilitate code status discussions. Two independent reviewers performed the data extraction and assessed risk of bias using the Cochrane Risk of Bias Tool. Data were pooled using a fixed-effects model, and risk ratios (RRs) with corresponding 95% CIs are reported.
Data Extraction and Synthesis
The study was performed according to the PRISMA guidelines.
Main Outcomes and Measures
The primary outcome was patient preference for CPR, and the key secondary outcome was patient knowledge regarding life-sustaining treatment.
Fifteen randomized clinical trials (2405 patients) were included in the qualitative synthesis, 11 trials (1463 patients) were included for the quantitative synthesis of the primary end point, and 5 trials (652 patients) were included for the secondary end point. Communication interventions were significantly associated with a lower preference for CPR (390 of 727 [53.6%] vs 284 of 736 [38.6%]; RR, 0.70; 95% CI, 0.63-0.78). In a preplanned subgroup analysis, studies using resuscitation videos as decision aids compared with other interventions showed a stronger decrease in preference for life-sustaining treatment (RR, 0.56; 95% CI, 0.48-0.64 vs 1.03; 95% CI, 0.87-1.22; between-group heterogeneity P < .001). Also, a significant association was found between communication interventions and better patient knowledge (standardized mean difference, 0.55; 95% CI, 0.39-0.71).
Conclusions and Relevance
Communication interventions are associated with patient decisions regarding do-not-resuscitate code status and better patient knowledge and may thus improve code status discussions.
To inform patients about treatment options in case of a cardiac arrest and their involvement in the decision-making process regarding their code status is considered a cornerstone of patient-centered care.1 Physicians are encouraged to conduct such code status discussions to respect patient autonomy as an ethical principle.2-4 Also, it is important to ask hospitalized patients for their preference because cardiopulmonary arrest occurs in almost 1 per 1000 hospitalization days.5
However, the literature reports several shortcomings and challenges in conducting code status discussions. First, many patients have unrealistic expectations about cardiopulmonary resuscitation (CPR) and associated risks and benefits.6,7 Patients with in-hospital cardiac arrests generally have a poor prognosis, with a survival to hospital discharge rate less than 20%.8,9 Beyond, many survivors have substantial neurologic deficits, limiting the potential to live an independent life.10
However, physicians often omit code status discussions or do not describe resuscitation measures, such as chest compressions or mechanical ventilation.11 Although CPR is an invasive procedure with potential complications, risks and benefits are usually not communicated adequately to patients, contributing further to patient misconceptions.12,13
A recent study14 in patients with cancer found that physicians document a presumed code status rather than conduct a true discussion, leading to a high proportion of full code status. Almost one-third of patients who were documented as full code would have preferred a do-not-resuscitate (DNR) code status if adequately informed about the consequences of CPR.14
Moreover, code status discussions are often ineffective due to poor communication skills of physicians.15,16 This is particularly true for junior physicians, who conduct most of the code status discussions in clinical practice and often perceive themselves as unprepared to explain complex medical procedures.17 Furthermore, code status discussions are often conducted under time constraints in an impersonalized, procedure-focused way, missing the chance to focus on individual patient values and goals.18-21 A recent study22 from Switzerland found that treating physicians significantly altered patient choices, raising the question of patient autonomy.
To date, there is no consensus about the best approach to code status discussions to understand patient preference and choice regarding DNR code status. The objective of this systematic review and meta-analysis was to identify studies examining communication interventions designed to facilitate code status discussions. We were especially interested in the association of communication interventions with patient preference for CPR or DNR code status and knowledge regarding resuscitation and its outcome.
This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines.23 We included randomized clinical trials (RCTs) in which the association of communication interventions during code status discussions with patient-relevant outcomes was compared with a control group. Studies were eligible if they focused on the outcomes of patient preference for resuscitation or DNR or patient knowledge regarding life-sustaining treatment.
We performed a comprehensive search strategy consisting of a combination of Medical Subject Headings and free-text words. We searched PubMed, Embase, PsycINFO, and CINAHL.
We developed the search strategy in consultation with a medical librarian (H.E.) experienced in systematic reviews. Initial search terms were drawn from a small set of key articles. We used an iterative process of building a search strategy, running the search, scanning the relevant retrieved articles for additional terms, and then rebuilding the search strategy with the newly identified relevant terms and related Medical Subject Headings. Because we focused on RCTs, we also used a sensitivity and precision-maximizing RCT filter for our search.24 The final search strategy for PubMed, which was adapted for the other databases, is available in the Appendix (eTable 1 in the Supplement).
To identify additional published, unpublished, and ongoing studies, we (1) tracked relevant references through the cited reference search of Web of Science and PubMed, (2) applied the similar articles search of PubMed, and (3) screened all references of potentially eligible studies. The data search was performed between September 3 and November 19, 2018.
Two of us (C.B. and L.L.) screened the titles and abstracts of articles found by the systematic search strategy. Studies were selected according to the inclusion criteria. We read the full texts of studies considered eligible for inclusion, and disagreement was resolved by discussion and consensus. Studies with the same assessment of end points were selected for quantitative meta-analysis regarding the association of communication interventions with primary and secondary end points.
Two of us (C.B. and L.L.) independently extracted the data of the included studies. Relevant outcomes for our systematic review and meta-analysis were patient preference for resuscitation or DNR code status and knowledge regarding CPR.
The RCTs were assessed for methodological quality using the Cochrane Risk of Bias Tool to rate the risk of bias in random sequence generation, allocation concealment, selective reporting, masking, completeness of outcome data, and other possible bias25 (eTable 2 in the Supplement). If at least 1 of the domains was rated as high risk, the trial was considered at high risk of bias. If all domains were judged as low, the trial was considered to be at low risk of bias. Otherwise, the trial was considered at unclear risk of bias. Two of us (C.B. and L.L.) performed data extraction and risk of bias assessment independently; disagreement was resolved by involvement of a third author (S. Hunziker).
We express dichotomous data risk ratios (RRs) with 95% CIs and report continuous data as the mean differences with 95% CIs. Data were pooled using a fixed-effects model. We identified heterogeneity (inconsistency) through visual inspection of the forest plots. We used the I2 statistic, which quantifies inconsistency across studies, to assess the consequences of heterogeneity on the meta-analysis. An I2 statistic of 50% or more indicates a considerable level of heterogeneity. If data were not suitable for direct comparison, we applied narrative synthesis.
For the primary end point, we performed several predefined subgroup analyses that stratified the results based on the following: type of intervention (video intervention vs no video intervention), age (<75 vs ≥75 years), risk of bias according to the Cochrane Risk of Bias Tool, study setting (outpatients vs hospitalized patients), marital status (≤65% vs >65% of patients married), education of the population (>30% vs ≤30% with a college degree or higher), and sex (≤55% vs >55% male). These cutoffs for stratification were chosen post hoc based on the distribution among trials to achieve a balanced number of patients per group. For the secondary end point, we performed several predefined subgroup analyses stratifying the results based on age (<75 vs ≥75 years) and risk of bias.
Statistical analyses were performed using the METAN package in Stata (Stata MP, version 15.1; StataCorp LP). Two-sided P < .05 was considered statistically significant.
A total of 7001 records were identified through our database searches. We removed duplicates (n = 1203) and discarded 5206 studies after examining titles and 559 studies after screening abstracts. Of the remaining 33 full-text articles, 15 studies26-40 were eligible for inclusion (eFigure in the Supplement). Six studies were judged to be at low risk of bias, 4 studies at high risk of bias, and 5 studies at unclear risk of bias.
Table 1 lists characteristics of the 15 included RCTs. Publication dates ranged from 1999 to 2018, and studies were conducted mostly in the United States (14 trials26-35,37-40), with 1 trial36 from Australia. Across all studies, a total of 2405 participants were included, with study sample sizes ranging from 50 to 313 per trial. In 8 studies,26,27,31-33,35,36,39 participants were recruited among hospitalized patients, and a further 5 studies28,29,34,38,40 recruited outpatients, whereas 1 study37 investigated residents of a nursing facility and 1 study30 recruited outpatients and hospitalized patients.
Eight studies26-30,34,36,38 used advanced diseases with a life expectancy less than 1 year, such as metastatic cancer, end-stage congestive heart, or renal failure, as the inclusion criteria, while 7 studies31-33,35,37,39,40 had no exclusion criteria based on illness. The mean age of the study population was 60 years or older in 12 studies. Six studies26,27,31,35,37,40 only recruited patients older than 60 or 65 years. Eleven studies assessed the outcome of preference for DNR of intervention vs control groups, and 8 studies assessed knowledge regarding CPR.
All studies used a dichotomous format (yes or no) to investigate the association of communication interventions with patient preference for CPR. Patient knowledge was assessed through questionnaires; 5 studies used the same questionnaire as in a previous study.41
Eleven included studies26-29,31,34,35,37-40 applied a video-based intervention. Ten videos showed simulated cardiac arrests and medical procedures undertaken during CPR, such as chest compressions and intubation. Some videos also contained images of real patients being treated on intensive care units, and other videos also provided information regarding end-of-life care or advance directives.37,40 Other studies used designed advance care planning interviews,30 standardized scripted explanations,32,33 or written information36 as interventions.
All studies used either structured questionnaires or interviews for data collection. One study30 did not specify assessment of preference for CPR.
Of the 15 eligible trials, 4 did not report data regarding patient preference for resuscitation and were excluded from the quantitative analysis. The remaining 11 trials26-31,35-38,40 (1463 patients) were pooled for the meta-analysis (Figure 1).
Five of these 11 studies reported no significant association of interventions with patient preference for CPR, and 6 trials reported a significant decrease in preference for CPR. Compared with usual care, the pooled results showed a significant association between the communication interventions and a lower preference for CPR (390 of 727 [53.6%] vs 284 of 736 [38.6%]; RR, 0.70; 95% CI, 0.63-0.78). There was high heterogeneity among trials (I2 = 81.2%; P < .001).
To assess the association of communication interventions with patient preference for CPR in predefined subgroups, we stratified our results by type of intervention, age, risk of bias, study setting, marital status of participants, education, and sex (Table 2). When stratified by type of intervention, trials that used videos showing resuscitation as a decision aid in their intervention group compared with other types of interventions demonstrated a stronger decrease in preference for CPR (RR, 0.56; 95% CI, 0.48-0.64 vs 1.03; 95% CI, 0.87-1.22; between-group heterogeneity P < .001). Studies with low risk of bias had a stronger association with lower preference for CPR compared with trials with higher risk of bias (RR, 0.52; 95% CI, 0.43-0.63 vs 0.87; 95% CI, 0.76-0.99; between-group heterogeneity P < .001). Stratification by study setting also showed no difference between outpatients and hospitalized patients (RR, 0.64; 95% CI, 0.51-0.79 vs 0.71; 95% CI, 0.60-0.85; between-group heterogeneity P = .82). When stratified by marital status, the intervention had a stronger association with lower preference for CPR in trials with no more than 65% vs greater than 65% of patients being married (RR, 0.47; 95% CI, 0.38-0.58 vs 0.84; 95% CI, 0.50-1.39; between-group heterogeneity P = .02).
Also, interventions had a stronger association with decreased preference for CPR in trials of patients with low education level (ie, ≤30% with college degree or higher) (RR, 0.48; 95% CI, 0.39-0.59 vs 0.94; 95% CI, 0.74-1.18; between-group heterogeneity P < .001). Regarding demographics, interventions had stronger association with reduced preference for CPR in trials that included older patients (≥75 years) compared with younger patients (RR, 0.58; 95% CI, 0.50-0.68 vs 0.86; 95% CI, 0.73-1.01; between-group heterogeneity P = .003) and in trials that had larger proportions of male patients (>55% vs ≤55% male) (RR, 0.49; 95% CI, 0.40-0.59 vs 0.68; 95% CI, 0.54-0.85; between-group heterogeneity P < .001).
Patient knowledge regarding CPR was assessed in 10 studies. Five trials used varying instruments to measure knowledge, which could not be standardized. We pooled the remaining 5 trials26-29,37 (including 652 patients) that used the exact same questionnaire for meta-analysis. In the pooled analysis, we found a significant association between communication interventions and higher patient knowledge (overall standardized mean difference [SMD], 0.55; 95% CI, 0.39-0.71). There was some heterogeneity among trials (I2 = 53.9%; P = .07) (Figure 2).
We then stratified the analysis by age and risk of bias. In low-risk trials, there was a stronger association between communication interventions and higher knowledge compared with higher-risk trials (SMD, 0.60; 95% CI, 0.43-0.77 vs 0.28; 95% CI, −0.10 to 0.67; between-group heterogeneity P = .14). Stratification by age did not show a significant difference between older and younger patients (SMD, 0.59; 95% CI, 0.39-0.80 vs 0.48; 95% CI, 0.23-0.73; between-group heterogeneity P = .48).
Three studies29,33,40 evaluated the associations of communication interventions with completion or presence of advance directives; however, they had too much heterogeneity to be included in a meta-analysis. Nicolasora et al33 assessed new completion rates of advance directives at hospital discharge and found that the intervention led to a significantly higher proportion of completed advance directives (0.8% vs 12.7%; P < .001). Yamada et al40 investigated the same topic 4 weeks after hospital discharge but also included patients who intended to fill out an advance directive, without reporting specific numbers. According to the authors, their results showed no significant findings. Epstein et al29 looked at advance care planning documentation overall, which included advance directives. It was not reported whether advance directives were completed in relation to the video intervention or whether they had already been in place before the study. The study found no statistical difference between the video intervention and control groups.
Several studies that used videos as decision aids assessed patient perception regarding the video intervention by ratings on a Likert-type scale.26-29,37,39 According to the results of those studies, patients generally were more comfortable watching a video, rating its content as useful or helpful in the process of decision making. One study28 used the Decisional Conflict Scale as a validated questionnaire to assess patient decision-making ability. In that study, the mean uncertainty score was significantly higher in the video group compared with the control group (13.7; 95% CI, 12.8-14.6 vs 11.5; 95% CI, 10.5-12.6; P = .002), indicating less uncertainty among patients who had seen the video in choosing between their treatment options.
One study31 investigating the effect of a video as a decision aid among 119 patients hospitalized on a general medical ward asked them about trust in their treating health care team as a secondary outcome. Trust was assessed on a 5-point Likert-type scale ranging from “agree” to “disagree.” There was no significant difference between groups (76% vs 93%; P = .08).
Rhondali et al34 investigated the extent to which patients perceived their physician as compassionate. Patients saw videos showing simulated code status discussions. Videos ended either with the physician making a recommendation or asking about patient preference. Independent of their allocated group, patients who opted for full code rated their physician as less compassionate than patients who opted for comfort care.
The findings of this systematic review and meta-analysis investigating associations between communication interventions to discuss code status and patient preference for resuscitation and patient knowledge regarding life-sustaining measures and outcome are 3-fold. First, we found a strong association between communication interventions and patient decisions regarding DNR code status, with lower preference for life-sustaining therapies if patients received a communication intervention compared with usual care. This association was more pronounced in studies with lower risk of bias. Second, associations between communication interventions and patient preference for a DNR code status were stronger when video-assisted decision aids were used, in trials that included older patients, in men, and among patients with lower healthy literacy. However, it is important to note that only a limited number of video interventions were tested in different settings. Third, communication interventions were also associated with better knowledge regarding resuscitation measures and the outcome of cardiac arrests. Again, trials with lower risk of bias had a stronger association with patient knowledge.
In line with our results demonstrating that more information delivered by communication interventions is associated with a higher probability for patients to choose a DNR code status, a previous trial38 found that health literacy (the ability to comprehend medical consequences) is a predictor of patient choice of DNR status. Therefore, more information may help patients make individualized informed decisions regarding resuscitation measures. Today, shared end-of-life decision making is considered an ethical obligation of patient-centered care to discuss equivalent treatment options, emphasizing patient autonomy and self-determination.1,42-44 However, decision making during code status discussions is often challenged by uncertainty surrounding interventions and therapies that might be available but whose outcomes remain uncertain.45,46 The results of the present systematic review and meta-analysis suggest that communication interventions, including video-assisted ones, enable patients to actively participate in the decision-making process by increasing their knowledge. This assumption is supported by a study28 using video that found a simultaneous increase in patient knowledge and decrease in decisional conflict regarding choice of care.
Previous studies11,47,48 reported variable quality of health care providers’ communication skills with hospitalized patients regarding code status. Herein, videos had the potential to inform patients in a standardized way and thereby promote shared decision making. However, interventions using visual components (eg, chest compressions, intubation, and ventilation) have been criticized because they may influence patients and lead them to a particular treatment choice. Furthermore, video tools as decision aids might not be applicable in some clinical settings due to limited accessibility and may be not suitable for elderly patients. Yet, some studies26-29,37,38 using Likert-type scales to assess patient comfort reported that patients were comfortable with watching a resuscitation video. However, there is also concern that videos as a decision aid might impair the patient-physician relationship. In 1 study,31 patients receiving a video intervention reported less trust in their treating health care team. In a study4 of advance care planning interventions, patients who opted for life-sustaining treatment perceived their physician as less compassionate, suggesting that these patients might not have approved of the video approach. Hence, video-assisted interventions may be useful adjuvants for code status discussions but should not be a substitute for direct patient-physician communication. A more flexible approach that can be adapted to individual patient needs might be more favorable and easier to implement in busy clinical environments.
A 2012 British multicenter cohort study49 investigated medical records of patients who had undergone resuscitation after an in-hospital cardiac arrest. In more than 75% of patients who received CPR, the code status was unknown, and 67% of patients who were resuscitated had an underlying preexisting fatal disease. An independent post hoc assessment of all cases found that a DNR status would have been appropriate in 85% because the risk-benefit ratio was unfavorable for these patients. As in patients with diseases for which they are receiving palliative care, CPR is not beneficial and may even prolong the dying process.3 In addition, we found that the interventions of our studies herein were associated with a greater reduction in patient preference for CPR in patients 75 years or older compared with younger patients. Also, in patients with a poor prognosis, we observed that the interventions had a stronger association with patient choice of a DNR code status. Therefore, such patients may receive the most benefit from communication interventions.
In general, a patient decision regarding DNR code status is a legal order to withhold CPR or advanced cardiac life support in case of cardiac arrest or respiratory failure and has important medical and socioeconomic consequences.50,51 Those 2 systematic reviews found variability in DNR decision making and implementation of DNR code status, leading to suboptimal care with undesired withdrawal of treatment in case of clinical deterioration. A standardized decision-making and documentation process of code status discussions may thus help improve quality of care and enable physicians to make decisions in the best interest of their patients. Today, an increasing number of hospitals and care centers use medical decision systems, such as Physician Orders for Scope of Treatment (POLST), Medical Orders for Scope of Treatment (MOLST), or Recommended Summary Plan for Emergency Care and Treatment (ReSPECT), which embed treatment plans in case of clinical deterioration. Based on our findings, it would be relevant to integrate communication interventions into such decision systems to further improve the uniformity of clinical care and strengthen patient involvement in the decision process.
In 1995, the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT),52 a landmark trial to investigate different approaches to improve care for seriously ill patients, reported shortcomings in communication during code status discussions. Despite all research efforts over more than 20 years, there is still need for large and high-quality RCTs focusing on interventions to facilitate code status discussions. In our systematic review and meta-analysis, we found only 3 studies30,32,33 that investigated interventions other than videos on patient preference for care and knowledge regarding resuscitation. There is clearly need for further trials regarding this important topic.
We are aware of several limitations to this systematic review and meta-analysis. The meta-analysis is based on a small number of trials and patients that could be considered for the quantitative analysis, and additional research is needed to confirm these results. A large proportion of trials targeted a population of terminally ill patients with a life expectancy less than 1 year, and generalizability to other patient populations is thus limited. In addition, the study populations were similar regarding ethnicity (mostly white) and age group (most were aged ≥60 years), again limiting generalizability of our results. Furthermore, most trials were performed in the United States, limiting transferability to other populations due to differences in medical and socioeconomic systems. Also, 5 of our 15 RCTs were performed by the same 2 groups of investigators (ie, by El-Jawahri et al26-28 and by Volandes et al37,38), and the findings from their trials had stronger effects compared with trials from other groups regarding patient preference for a DNR code status. However, those 5 studies had low risk of bias, and trials were performed in different settings (ie, outpatients vs hospitalized patients) and with different patient populations (ie, those with palliative vs curative diseases). Therefore, validation of our results by independent research groups is warranted. The number of trials and patients was small, also limiting interpretation of our subgroup analyses and increasing the risk for type II error.
Communication interventions may be an effective decision aid for code status discussions, potentially altering patient preference and increasing patient knowledge. More informed patients may be better able to participate in the decision-making process, which might prevent unwanted excessive medical procedures. There is still urgent need for large-scale RCTs to investigate further approaches to facilitate code status discussions.
Accepted for Publication: April 16, 2019.
Published: June 7, 2019. doi:10.1001/jamanetworkopen.2019.5033
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Becker C et al. JAMA Network Open.
Corresponding Author: Sabina Hunziker, MD, MPH, Medical Communication, Department of Psychosomatic Medicine, University Hospital Basel, Petersgraben 4, CH-4031 Basel, Switzerland (email@example.com).
Author Contributions: Drs Becker and Hunziker had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Becker, Lecheler, Rueter, Schaefert, Bassetti, Hunziker.
Acquisition, analysis, or interpretation of data: Becker, Lecheler, Hochstrasser, Metzger, Widmer, Thommen, Nienhaus, Ewald, Meier, Schaefert, Hunziker.
Drafting of the manuscript: Becker, Lecheler.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Becker, Lecheler, Hunziker.
Administrative, technical, or material support: Metzger, Nienhaus, Ewald, Meier, Rueter, Schaefert, Hunziker.
Supervision: Rueter, Schaefert, Hunziker.
Conflict of Interest Disclosures: None reported.
Create a personal account or sign in to: