Medical Decision-Making and Bereavement Experiences After Cardiac Arrest: Qualitative Insights From Surrogates

This was a cross-sectional analysis from a convergent mixed methods⁹  longitudinal study involving surrogates of adult (≥18 years) ICU patients who received post–cardiac arrest care at a high-volume, urban academic medical center in the northeastern United States. The longitudinal study aimed to explore both end-of-life decision-making experiences and symptoms during the first 6 months of bereavement. Here, we present baseline (ie, approximately 1 month after loss) qualitative findings by decisional conflict survey scores. Surrogates were eligible if (1) the patient, although successfully resuscitated, did not survive hospitalization and (2) the surrogate preferred to be interviewed in English. Surrogates were excluded if they reported that they were not involved in medical decision-making in the ICU. We recruited consecutively by phone 1 month after the patient’s death from February 22, 2022, through February 28, 2023. All study participants provided informed consent. The study protocol was approved by the Columbia University Irving Medical Center Institutional Review Board and was followed in accordance with the Helsinki Declaration of 1975.

Surrogates completed surveys assessing end-of-life decision-making and symptoms at 4 time points: approximately 1, 2, 3, and 6 months after the patient’s death. Surveys were completed either online (REDCap)¹⁰  or by telephone (1 researcher [C.E.D.]). The baseline (ie, about 1 month after death) surveys included the low-literacy version of the Decisional Conflict Scale (DCS-LL; Flesch-Kincaid grade level 3.1)¹¹  to retrospectively measure decisional conflict experienced during end-of-life decision-making. The DCS-LL includes 10 items (with responses of yes, no, and unsure) and 4 subscales (informed, values clarity, support, and uncertainty) with both total and subscale scores ranging from 0 (no conflict) to 100 (extremely high conflict). A generally accepted cut point of greater than 25 indicates moderate-to-high decisional conflict.^6,11  We used total DCS-LL scores to identify surrogates with higher decisional conflict (ie, scores above the median) for analysis of qualitative data. The original version of the DCS has been used in prior studies of ICU surrogates^6,12,13  given its established psychometric properties¹⁴ ; however, we opted to administer the low-literacy version given the anticipated needs of our cohort and because it omits the Effective Decision subscale/domain from the original DCS, which includes items inappropriate for bereaved ICU surrogates (eg, “I expect to stick with my decision”). The DCS-LL has demonstrated reliability (Cronbach α 0.83-0.88) among non-ICU surrogate groups.^11,15,16 

Patient electronic health record data were abstracted to collect patient demographics and details pertaining to the hospitalization (eg, life-support measures) including surrogate decisions around withholding (eg, do-not-resuscitate [DNR] order) or withdrawing treatments.

Individual interviews were conducted approximately 1 and 3 months after the patient’s death. The first interview focused on decisional conflict; the second interview (findings not reported here) focused on the evolution of surrogate symptom experiences. We conducted the interviews either by telephone or by video conference. Guided by qualitative descriptive methodology,^17,18  the interviews aimed to gain insider knowledge from bereaved surrogates around decisional conflict associated with following a patient’s cardiac arrest that preceded their death. An iterative, semi-structured interview guide¹⁹  (Table 1) guided each interview. Before the guide was used, an interdisciplinary team of content experts provided input, and the guide was then pilot tested with 2 surrogates. One researcher (C.E.D.; a doctoral student with a critical care nursing background) conducted all interviews. A distress protocol was in place during interviews, if needed, and all participants received a list of bereavement resources. Interviews were audio recorded, transcribed using RevAI (Rev.com, Inc), and cleaned before analysis. Data were managed using NVivo12Plus (QSR International) software.

Baseline (ie, approximately 1 month) data were included in the analyses. Figure 1 depicts the mixed methods analytic approach. We performed both directed (ie, responses to interview guide questions) and conventional (ie, inductive) content analysis²⁰  using an iterative codebook²¹  prior to knowledge of DCS-LL scores. Team members (C.E.D., A.S., M.G.) independently read the first 3 transcripts line by line to develop the initial codebook to guide coding of the remaining transcripts. Codes were collapsed into categories, and from the categories, we identified crosscutting themes. Discrepancies were resolved through team discussion. A data saturation table was created to track data adequacy. Descriptive statistics were then computed for demographic characteristics and DCS-LL scores. Finally, we compared qualitative findings based on level of decisional conflict (ie, above vs at or below the median DCS-LL score).

We incorporated multiple methods to enhance the trustworthiness of the qualitative data. As findings emerged, they were shared in subsequent interviews to ascertain whether the themes resonated with other surrogates (ie, member checking).²²  The interviewer engaged in reflexivity and used peer debriefing to reduce the risk of bias^23–25  during data collection and analysis. An audit trail was established that included codebook iterations, coded transcripts, field notes, and team decisions. Survey and electronic health record data were used to triangulate the interview findings. The report of this study adheres to the Standards for Reporting Qualitative Research.²⁶ 

Of 53 surrogates approached, 16 participated (including 2 surrogates for 1 patient). Nine opted to be interviewed by phone and 7 by video conference. On average, interviews took place 36 days (median; range, 32-47 days) after the patient’s death and lasted 50 minutes (median; range, 36-88 minutes). Table 2 presents characteristics of the surrogates (median age, 59.4 years; 50% female; 31% Black or African American; 25% Hispanic or Latino; 50% spouse or partner). Of the 15 patients (median age, 59.4 years; 47% female; 33% Black or African American; 13% Hispanic or Latino), most (n = 12) had experienced an in-hospital cardiac arrest with ICU and hospital lengths of stay averaging around 1 and 2 weeks, respectively. Roughly two-thirds of surrogates (n = 11) opted to withdraw life-sustaining treatments. The remaining surrogates either made the decision to withhold resuscitative efforts (ie, made the patient DNR; n = 2), pursued treatments that were ultimately unsuccessful (n = 1), or were agreeable to discontinuing treatments after brain death determination (n = 2). Additionally, 4 surrogates (for 3 patients) decided to donate the patient’s organs.

Figure 2 depicts DCS-LL scores reported by the group. Overall, 9 surrogates scored 0; those who scored higher than 0 all scored below the moderate-to-high cut point. Subscale scores ranged highest for the values clarity and uncertainty subscales. Given a median total DCS-LL score of 0, scores were dichotomized as “some” versus “no” decisional conflict for analysis of the interview data.

Two themes reflecting decision-making experiences emerged from the directed analysis: Theme 1 consisted of 4 categories and theme 2 consisted of 2 categories. A third theme (4 categories) from the inductive analysis reflected broader experiences during bereavement. Table 3 presents supporting exemplar quotes with identifiers linked to the article text. Figure 3 provides an example of our concordance/discordance analysis approach. Table 4 presents additional exemplar quotes and concordance/discordance by decisional conflict level. Data saturation was reached for all categories but one (from the inductive analysis; Table 5).

The significance of the surrogate’s role was emphasized by one as “performing the ultimate act” (participant 8). Although the role often fell to a person who generally knew the patient very well (eg, legal next of kin), one notable concept emerged around becoming the decision maker: last man standing. Regardless of decisional conflict level, some participants described assuming their surrogate role because they were the only one available and/or willing to make decisions (Table 3, quote 1).

Decisional conflict was reported across types of treatment decisions. Surrogates from both groups described the weight of responsibility of withholding or withdrawing treatments (Table 3, quote 2). A notable difference was that surrogates who reported decisional conflict included those who opted to pursue all possible life-saving treatments (ie, full code) during initial discussions with staff, before the in-hospital cardiac arrest, and then later withheld or withdrew treatments.

Surrogates cited many factors that influenced their decision-making experiences, including family involvement. While most surrogates who did not report decisional conflict felt supported by their family, the surrogates still carried the onus of responsibility (Table 3, quote 3). Although some surrogates reported knowing the patient’s wishes regarding medical treatments, others (with decisional conflict) were unclear. One, whose loved one had had a chronic, life-limiting condition since birth, experienced uncertainty during her decision to withdraw treatments (Table 3, quote 4). Surrogates who understood treatments and long-term prognosis commonly reported no decisional conflict. Others, who reported decisional conflict, described poorer medical understanding (Table 3, quote 5). In both decisional conflict groups, surrogates cited cultural or patient-specific characteristics influencing decision-making (Table 3, quote 6).

Regardless of decisional conflict, surrogates generally reflected positively on their decisions (Table 3, quote 7). They described an absence of guilt and felt thankful for the time they were allowed to make their decisions (Table 3, quote 8). Four surrogates who discussed their decisions to pursue organ donation were glad to have done so (Table 3, quote 9).

Most surrogates, representing both decisional conflict groups, felt that staff were available to talk with and were grateful for the empathy of staff during conversations. One, however, who reported decisional conflict, felt less supported (Table 3, quote 10). Surrogates from both groups interacted with many clinicians in the ICU, which sometimes resulted in their receiving conflicting information or information that seemed inconsistent with their own assessment of the patient (Table 3, quote 11). Regardless of decisional conflict, surrogates emphasized a need for honesty from clinicians as they processed the situation and prepared for decision-making (Table 3, quote 12).

Surrogates from both decisional conflict groups recalled an impaired ability to process medical information because of high distress in the ICU. One surrogate cited difficulty in understanding that interfered with withdrawal of treatment after brain death (Table 3, quote 13).

All surrogates talked in detail about how their lives had changed since their loss. A feeling shared by many was summarized by one: “I felt sort of lost of what to do after he passed. There was no guidance of, like, these are the steps that’ll happen next … I wish there was a handbook” (participant 6). They described efforts to navigate life after the loss and difficulties they encountered in various aspects of their lives.

Surrogates disclosed financial hardship, whether due to funeral or burial costs or to the sudden loss of the patient’s income. One faced unique challenges after his partner lost her job during her hospitalization (Table 3, quote 14).

Surrogates reported experiencing physical symptoms, including fatigue and disturbed sleep (Table 3, quote 15). Others described how their diet had changed (Table 3, quote 16).

Regardless of decisional conflict, the circumstances of the loss often weighed heavily on surrogates, who described persistent feelings of shock. Three were actively searching for a medical cause or attributed the arrest to medical mismanagement. Three others had a clear understanding of the cause and were ruminating over it (Table 3, quote 17).

Surrogates described challenges facing disrupted routines, resuming professional responsibilities, or taking on new household duties (eg, managing finances). They experienced changes in their caregiving roles: one was left to parent young children independently and 2 others assumed co-parenting roles for a grandchild. Surrogates also described mourning the loss of routines (Table 3, quote 18).

This mixed methods study provides insight into surrogate medical decision-making in the ICU and the experiences of bereaved surrogates after the cardiac arrest of a loved one. Interview data from surrogates included rich descriptions of decision-making experiences which, for many, aligned with the DCS-LL subscale and subconstruct scores (eg, values clarity subscale). An important finding was that surrogates who reported decisional conflict included those who had opted for all possible life-saving treatments during initial goals-of-care conversations, before the patient’s in-hospital cardiac arrest, and then later withheld or withdrew treatments. We were unable to discern exactly which decision they felt conflicted about (ie, pursuing vs withholding or withdrawing treatments) through the survey, but qualitative data showed that those individuals encountered suboptimal support from clinicians during treatment conversations and were left with a poor understanding of what had caused the cardiac arrest. Fostering a supportive environment during discussions, offering information around cardiac arrest etiology, and allowing for questions may help to alleviate decisional conflict as surrogates contemplate decisions to withhold or withdraw treatments, especially after successful resuscitation following in-hospital cardiac arrest.

The surrogate experiences described here highlight opportunities for improved family-centered care. Our participants described assuming the surrogate role when others declined or were unavailable; as the “last man standing,” surrogates may benefit from recognition and enhanced emotional support. Communication with clinicians left lasting impressions, with surrogates from both groups recalling conflicting messages. They emphasized the importance of honesty, specifically around prognosis, as being essential to their decisions to withhold or withdraw treatments. Surrogates also described how they navigated life after the loss. Many were in shock and described persistent, unresolved feelings around the cause of the cardiac arrest. These findings are similar to those of studies involving family members bereaved after sudden cardiac arrest,^27–29  which recommend follow-up care to allow families the opportunity to debrief.³⁰  Our surrogates also described financial hardship, disruptions in health behaviors and physical symptoms, and adjustments in roles or responsibilities in the weeks after the ICU stay. Together, the findings highlight opportunities for future interventions, including enhanced care coordination to connect surrogates with community resources as they navigate challenges during bereavement. Post-ICU follow-up may be a novel way to optimize the well-being of bereaved surrogates.

Future research and practice recommendations to support ICU surrogates bereaved after cardiac arrest are especially important because 3 in 4 make life-limiting treatment decisions.²  We learned that there are limitations to measuring decisional conflict around end-of-life decisions. Whereas the low DCS-LL scores our surrogates reported aligned with other studies that administered the original DCS to surrogates,^12,31  our surrogates’ low scores may reflect recall bias because they completed the survey more than 1 month after having made their decisions. The feelings they experienced during decision-making may have abated. Their memory of decisional conflict may also have lessened as a result of social support received during the first month of bereavement. Despite there being precedent for evaluating decisional conflict retrospectively (as much as 6 months after the decision),³²  results of retrospective measurement must be interpreted with caution. Future studies should limit the time between decision-making and measurement of the construct. Researchers must also carefully consider how survey language is framed. The DCS-LL questions may have seemed ambiguous despite our intention for them to be specific about decisions preceding the death. Because surrogates face multiple treatment decisions in the ICU, the language preceding the survey questions should be considered carefully so that the construct is appropriately measured.

We had the additional benefit of comparing DCS-LL scores with qualitative data. The higher-scoring subscales were values clarity and uncertainty. It was apparent through interviews that certainty during decision-making overlapped with values clarity; those who knew the patient’s wishes for treatment felt certain about their decisions. However, surrogates who reported uncertainty (ie, uncertainty subscale > 0) were not unclear about the patient’s values (ie, values clarity subscale = 0), which suggests that something else may have been driving uncertainty around end-of-life decisions. Although total DCS-LL scores were generally low, many surrogates with a total score greater than 0 had a score greater than 0 on the informed subscale (eg, “Do you know the risks and side effects of each option?”). The qualitative data suggest that uncertainty about neurologic prognosis after cardiac arrest may be an important contributing factor to decisional conflict. Prognostic uncertainty is distressing for surrogates making life-limiting treatment decisions³³  but may not be fully captured by the DCS-LL.

To our knowledge, no study has explored the long-term effects of decisional conflict among bereaved ICU surrogates; future work should assess relationships between decisional conflict and important surrogate outcomes (eg, symptoms) and seek an enhanced understanding of regret around treatment decisions. Because bereaved ICU surrogates are at risk for adverse outcomes such as prolonged grief disorder,^34–36  the influence of decision-making experiences on surrogate outcomes warrants further exploration, especially given the prognostic uncertainty that can accompany critical illness (eg, cardiac arrest). Future studies should also strive to include all types of decision makers, as this study involved mostly surrogates who elected to withdraw treatments.

Our study had important limitations. We administered the DCS-LL retrospectively. In addition, although the original DCS has established validity and reliability among ICU surrogates,¹⁴  we used the DCS-LL, which has not been psychometrically tested in this group. Interviews were conducted at least 1 month after patient death to respect early bereavement, but the timing may have contributed to recall bias and self-selection bias. As a result, the findings may not be representative of those who opted not to participate in this study. Of note, most participants withdrew treatments (ie, very few elected to pursue treatments that were unsuccessful), which further limits the findings. Although we reached data saturation for themes 1 and 2, which reflected decision-making experiences, the data reflected in theme 3 describing bereavement experiences may not have been fully explored. Our data also do not describe later bereavement experiences relating to ICU decision-making (eg, regret). Furthermore, our study was limited to surrogates of patients who experienced cardiac arrest at a single, urban academic medical center in the United States; decision-making experiences may vary in the context of other critical illnesses or diagnoses, hospital settings and locations, and countries with sociolegal differences relating to surrogate involvement in medical decision-making. These limitations not-withstanding, the study reflected the experiences of surrogates from diverse racial and ethnic backgrounds and used a rigorous mixed methods approach.

These findings provide novel insight into the experiences of bereaved ICU surrogates who made medical decisions for patients after a cardiac arrest. Two themes, “the ultimate act” and “the legacy of clinician communication,” highlight salient aspects of surrogate decision-making near an ICU patient’s end of life, and a third theme, “I wish there was a handbook,” captures difficulties navigating life after loss. We also identified similarities and differences in experiences between those who reported decisional conflict and those who reported none. Future research should explore how medical decision-making influences ICU surrogate outcomes and should consider other decision experiences such as regret. Further, as we aim to improve outcomes for bereaved surrogates of patients who have experienced cardiac arrest, consideration must be given to the challenges the surrogates encountered after their loss. Novel approaches to interventions anticipating these challenges are warranted.

This work took place at Columbia University School of Nursing. We thank our participants who generously contributed their time and shared their experiences; we honor their loved ones. We also thank the Cardiac Arrest Neuropsychosocial Outcomes Evaluate-Family research program (CANOE-F; R01HL153311) for recruitment support, including Danielle Rojas, Sabine Abukhadra, and Isabella Tincher. We thank the members of the Qualitative Debriefing Group at Columbia University School of Nursing for their support of the researchers throughout the conduct of this study.