Barriers to Surrogate Application of Patient Values in Medical Decisions in Acute Stroke: Qualitative Study in a Bi-ethnic Community

Abstract

Objective:

To identify barriers to surrogate decision maker application of patient values on life-sustaining treatments after stroke in Mexican Americans (MAs) and non-Hispanic Whites (NHWs).

Methods:

Qualitative analysis of semi-structured interviews with stroke patient surrogate decision makers completed approximately 6 months after hospitalization.

Results:

42 family surrogate decision makers participated (median age: 54.5; female: 83%; patients were MA:60%, NHW: 36%; and 50% were deceased at time of interview). We identified three primary barriers to surrogates’ application of patient values and preferences when making decisions on life-sustaining treatments: 1) A minority of surrogates had no prior discussion of what the patient would want in event of serious medical illness; 2) Surrogates struggled to apply prior known values and preferences to the actual decisions made; 3) Surrogates felt guilt or burden, often even in the setting of some knowledge of patient values or preferences. The first two barriers were seen to a similar degree in MAs and NHWs, though guilt or burden was reported more commonly among MAs (28%) than NHWs (13%). Maintaining patient independence (e.g. ability to live at home, avoid a nursing home, make their own decisions) was the most important priority for decision making for both MAs and NHWs, though MAs were more likely to list spending time with family as an important priority (24% vs. 7%).

Conclusion:

Stroke surrogate decision makers may benefit from: 1) Continued efforts to make advance care planning more common and more relevant; 2) Assistance in how to apply their knowledge of patient values to actual treatment decisions; and 3) Psychosocial support to reduce emotional burden. Barriers to surrogate application of patient values were generally similar in MAs and NHWs, though the possibility of greater guilt or burden among MA surrogates warrants further investigation and confirmation.

Introduction

In the setting of serious illness, decisions on life-sustaining treatments are often made by a surrogate decision maker, typically a family member. These decisions are not based solely on medical criteria but must incorporate individual factors such as how a person may view survival with disability. Traditional Western ethical frameworks emphasize that surrogate decision makers should attempt to make a decision that is rooted in the preferences and values of the patient whenever possible.² This concept is also emphasized in leading frameworks for medical decision making, which define high quality-decisions as those that are concordant with an individual’s values.³

In stroke, surrogate decision makers may face greater challenges due to time pressure on decisions such as emergency brain surgery and the lack of opportunity for patients to discuss wishes with surrogates in advance. Most post-stroke deaths occur after a decision to limit use of life sustaining treatments, commonly in the first several days of hospitalization.^4,5 Some under-represented groups such as Mexican Americans (MAs) are less likely than non-Hispanic Whites (NHWs) to have early limitations in life sustaining treatment,^6,7 though exact reasons for these differences are unclear.

This report addresses two primary research questions: 1) What barriers do surrogates experience in applying patient values and preferences when making decisions after stroke?; and 2) Do surrogate barriers or understanding of patient treatment priorities differ by ethnicity? This study includes findings from semi-structured surrogate interviews and is part of a larger mixed-methods study focused on surrogate decision makers of stroke patients in a predominantly bi-ethnic community in South Texas.^8,9

Methods

Study overview and eligibility

Participants came from the Outcomes Among Surrogate decision makers In Stroke (OASIS) study, a cohort study of surrogate decision makers for stroke patients in Nueces County, Texas. OASIS methods have been described previously.^8,9 This community is predominantly bi-ethnic, (about 65% MA and 30% NHW).¹ The MAs in the community are overwhelmingly non-immigrant, US citizens. Cases of ischemic stroke or intracerebral hemorrhage patients 45 years and older were identified from the Brain Attack Surveillance in Corpus Christi (BASIC) Project using active and passive surveillance as previously described.¹⁰ Surrogates were eligible if they identified as significantly involved in any of the following decisions: do-not resuscitate (DNR) orders, mechanical ventilation, brain surgery, long-term feeding tube, comfort care, palliative care, or hospice.

All OASIS participants completed a series of quantitative surveys extending out to 12 months post-stroke. Selected participants were invited to participate in a single qualitative semi-structured interview conducted at approximately 6 months (one surrogate per patient). Qualitative participants were purposively sampled to achieve balance on surrogate ethnicity and patient survival, as well as individuals who would likely be willing to share their experiences in an in-depth interview. Qualitative interviews were completed from 2/2018–9/2019.

Surrogate and patient race-ethnicity were classified by self-report, typically done by the surrogate as most patients were unable to speak for themselves. Race-ethnic categories were classified as MA, NHW, or other based on the category of the patient (one dyad had a MA patient and NHW surrogate; no quotes from this dyad are presented and other dyads had concordant race-ethnicity).

Definitions

The term “values” in health literature can have different meanings in different contexts. We follow a framework from the medical decision-making literature that defines values as an “…umbrella term referring to what matters to an individual relevant to a health decision.”¹¹ The term “preferences” is used to describe the “…extent to which a decision option or health state is desirable or acceptable…”.¹¹ During the interview, we used general terms such as “wishes” to describe what the patient would want or what was most important for decision making to limit confusion on the distinction between values and preferences.

Qualitative Data Collection

The qualitative study team consisted of two interviewers (CO and RL), a project manager (EC), and two stroke neurologists (LM and DZ). Interviews were conducted by CO and RL after training with DZ. Study physicians were not involved in participants’ medical care. Participants knew the interviewers through the ongoing OASIS quantitative study and were told that the purpose of the research was to improve surrogate decision-making for people in the future.

An interview guide (see Supplemental Appendix) was developed based on study aims, with portions adapted from a prior study of stroke surrogate decision makers.¹² The precise questions may have varied due to interview flow and timing or minor modifications to the guide over time to improve clarity; the target interview length was approximately 1 hour. After discussion of the patient background and initial presentation with stroke, the interviewer focused a set of questions on one decision, either the decision that the surrogate had previously selected as the hardest (of the options DNR, feeding tube, mechanical ventilation, brain surgery, or comfort measures only) or another decision of the surrogate’s choice

To explore ethnic differences in end-of-life treatments in this community,^6,7 we investigated whether MAs and NHWs had different priorities when considering quality of life or facing serious illness. Surrogates were read a list of priorities adapted from other work (e.g. maintaining independence, avoiding dependence on machines, living as long as possible).^13–15 Surrogates were then asked which of these items was most important to the patient, with responses categorized by consensus review of the coding team and summarized by race-ethnicity.

Interviews were completed in the participant’s home or in the OASIS office (except for one telephone interview) and were audio recorded with an encrypted device prior to professional transcription. Interviews were offered in either English or Spanish, though all interviews were completed predominantly in English. Occasional Spanish phrases were translated for analysis by interviewer CO (certified in Spanish to English translation). Study staff uploaded de-identified transcripts to MAXQDA 2020 software (VERBI Software, Berlin). Surrogates were given $10 for their time. All participants were also offered a list of local community resources (including mental health) at the conclusion of the interview.

Qualitative Analysis

We used descriptive qualitative methodology, accepting the data as representing our participants’ perceptions while also recognizing our role as researchers in describing the phenomenon based on our interpretations of participant perceptions.^16,17 Qualitative content analysis was used to identify unifying and recurrent themes in the interview data using an iterative and reflexive approach.^18,19 The coding process followed a previously described hybrid method that combined deductive and inductive approaches,²⁰ where data were initially organized using an a priori template set of codes based on our research questions. Next, template codes were revised while inductive codes were applied to new concepts in an iterative process. The first 4 interviews were coded independently by CO, RL, and DZ and then discussed to reach consensus on each coded segment. Subsequent interviews were coded by CO or RL with every 3^rd interview double coded and reviewed for consensus to maintain reliability. After the initial coding, all interviews were again reviewed by DZ and RL to identify unifying themes related to barriers in application of patient values and preferences to medical decision making. Presented quotes may have been edited for clarity to remove non-contributory phrases.

Statistical analysis and regulatory approval

Descriptive statistics are presented overall and by ethnicity, though due to the small sample size, formal hypothesis testing for ethnic differences was not performed. Comparison of the frequency and proportions of selected codes and themes are provided by ethnicity as a general description of the data and to generate hypotheses for future work. This study has been approved by the University of Michigan IRB and the IRBs of both participating hospital systems in Corpus Christi, Texas. Informed consent was obtained from all participants.

Results

Participant Characteristics

42 surrogates were interviewed between 2/2018 and 9/2019. Participant demographics are summarized in Table 1. Patient race-ethnicity was MA in 25/42 (60%), NHW in 15/42 (36%), and other in 2/42 (5%), with 21/42 (50%) patients deceased at the time of the qualitative interview. Surrogates were mostly women (83%) and predominantly the child of the patient (57%) or the spouse (26%). Qualitative participants were not different from the overall OASIS sample on key demographics, language preference, and severity measures (see Supplemental Appendix).

Table 1.

Participant characteristics

	All Participants (N=42)*	Patient Race-Ethnicity Mexican American (N=25)	Patient Race-Ethnicity non-Hispanic White (N=15)
		N (%) or Median (Q1, Q3)	N (%) or Median (Q1, Q3)
Patients
Patient Age	71.5 (65, 82)	72 (63, 80)	69 (65, 82)
Patient Sex Female	23 (54.8)	12 (48.0)	10 (66.7)
Patient Race-ethnicity
Mexican American	25 (59.5)	25 (100.0)	0 (0.0)
Non-Hispanic White	15 (35.7)	0 (0.0)	15 (100.0)
Other	2 (4.8)	0 (0.0)	0 (0.0)
Stroke type
Ischemic	28 (66.7)	17 (68.0)	9 (60.0)
Intracerebral hemorrhage	14 (33.3)	8 (32.0)	6 (40.0)
National Institutes of Health stroke scale	15.5 (6, 23)	13 (6, 20)	16 (4, 30)
Pre-stroke Rankin
0–2	25 (59.5)	14 (56.0)	10 (66.7)
3–5	17 (40.5)	11 (44.0)	5 (33.3)
Intensive care unit admission	27 (64.3)	13 (52.0)	13 (86.7)
Ventilator	11 (26.2)	4 (16.0)	7 (46.7)
Feeding tube	10 (23.8)	4 (16.0)	5 (33.3)
Tracheostomy	1 (2.4)	1 (4.0)	0 (0.0)
Brain surgery	1 (2.4)	0 (0.0)	1 (6.7)
Ventriculostomy	1 (2.4)	1 (4.0)	0 (0.0)
Do-not-resuscitate order
At any time	18 (42.9)	10 (40.0)	7 (46.7)
Within 24 hours	12 (28.6)	7 (28.0)	4 (26.7)
Comfort care	18 (42.9)	10 (40.0)	7 (46.7)
Died in hospital or discharged to hospice	18 (42.9)	10 (40.0)	7 (46.7)
Deceased at time of qualitative interview	21 (50.0)	14 (56.0)	7 (46.7)
Patient Relation to Surrogate
Parent	24 (57.1)	15 (60.0)	8 (53.3)
Spouse	11 (26.2)	5 (20.0)	6 (40.0)
Other Family Member	7 (16.7)	5 (20.0)	1 (6.7)
Surrogate
Surrogate Age	54.5 (46, 64)	55 (46, 62)	57 (48, 68)
Surrogate Sex Female	35 (83.3)	24 (96.0)	10 (66.7)
Surrogate Race-ethnicity
Mexican American	24 (57.1)	24 (96.0)	0 (0.0)
Non-Hispanic White	16 (38.1)	1 (4.0)	15 (100.0)
Other	2 (4.8)	0 (0.0)	0 (0.0)
Surrogate Education
Some High School	4 (9.5)	3 (12.0)	1 (6.7)
Completed High School	7 (16.7)	6 (24.0)	1 (6.7)
Some college or more	31 (73.8)	16 (64.0)	13 (86.7)
Surrogate reported that patient had a written advance directive	13 (31.7)	8 (33.3)	5 (33.3)
Decisions discussed*
Do-not-resuscitate order	36 (90.0)	21 (91.3)	14 (93.3)
Ventilator	25 (59.5)	16 (64.0)	8 (53.3)
Feeding Tube	23 (56.1)	15 (62.5)	7 (46.7)
Brain surgery	23 (56.1)	15 (60.0)	8 (57.1)
Comfort Measures	21 (50.0)	13 (52.0)	7 (46.7)
Hardest Decision*
Do-not-resuscitate order	12 (28.6)	8 (32.0)	4 (26.7)
Mechanical Ventilation	6 (14.3)	2 (8.0)	4 (26.7)
Feeding Tube	2 (4.8)	2 (8.0)	0 (0.0)
Brain surgery	11 (26.2)	8 (32.0)	3 (20.0)
Comfort Care	5 (11.9)	2 (8.0)	2 (13.3)
Other/None of the above	6 (14.3)	3 (12.0)	2 (13.3)
Interview location
Participant Home	28 (66.7)	16 (64.0)	10 (66.7)
Staff office or other	14 (33.3)	9 (36.0)	5 (33.3)

^*The scope of decisions discussed and selection of the hardest decision was assessed at the post-discharge survey. Note that discussion of a topic with the healthcare team does not mean that the patient had the intervention.

Assessment of most important patient priority

When surrogates were asked about the most important patient priority when considering quality of life or facing serious illness, the most common response (15/42, 36%) was maintaining patient independence (e.g., ability to live at home, avoid a nursing home, or make their own decisions). Patient independence was most important in 8/25 (32%) MAs and 7/15 (47%) NHWs. The next most common response was spending time with family, though this response was more common among MAs than NHWs (all surrogates:7/42, 17%; MA: 6/25, 24%; NHW: 1/15, 7%). Only one surrogate (MA) indicated that living for a long time was the most important patient priority.

Overview of barriers to surrogate application of patient values and preferences

We identified three major themes of barriers to surrogates’ application of patient values and preferences to decisions after stroke (Figure). The first barrier was that some surrogates had not had any prior discussions with the patients about what they would want in the setting of serious illness. The second barrier was that some surrogates had knowledge of the patient’s values and preferences, but still struggled with the application of this prior knowledge to the current situation. The final barrier occurred when surrogates reported experiencing guilt or burden around decision making, even when they knew what the patient would want.

Figure: Overview of barriers to surrogate application of patient values and preferences in medical decision making.

See text for additional details of subthemes for barrier 2 and Table 2 for examples of barrier 3 stratified by ethnicity. MA=Mexican American; NHW=non-Hispanic White

Barrier 1: No prior discussion of patient values or preferences

When asked about prior advance directives or conversations with the patient, 9/42 (21%) indicated that they had never had any discussion with the patient about what they would want in the event of serious illness. If anything, this lack of any discussion was more common in NHWs (5/15, 33%) than in MAs (3/25, 12%). Among those who never had specific discussions with the patient, about half of these surrogates still felt that they were able to make decisions based on their understanding of patient values or general knowledge about the patient. One representative surrogate stated, “…I knew from not like a specific conversation, but from knowing him he wouldn’t wanna live like that…” (Interview 14, NHW).

Some of the surrogates who did not have prior specific knowledge of what the patient would want defaulted to continuing intensive treatments. In two cases, this decision was later felt to be problematic by either the surrogate or the patient. In one example, a surrogate who was asked how much her husband’s wishes influenced her decision on DNR status responded, “He didn’t make those known to me” (Interview 15, NHW). This surrogate based her decision on what the adult children would want: “…you gotta resuscitate him … cuz the kids will be upset if I just let him die”. This surrogate later indicated that she had some regrets about this decision over time, stating “…he’s alive, but he’s not living. He’s miserable.” In another example, a surrogate who had not discussed prior wishes with her mother, proceeded with full code status and brain surgery for a hemorrhage, stating, “It was an easy decision cuz I’ve…for me, it was like, fix my mom” (Interview 41, NHW). However, later the surrogate learned that her mother was not happy with this decision: “…she was mad. She was very angry that she was resuscitated …she was not happy to be alive. She was very angry with us… cuz she didn’t wanna live this way, so that was quite shocking, you know.” While these concerns that arose after defaulting to continuing intensive treatments were not necessarily common, they are still mentioned due to potential for contributing to subsequent surrogate regret or conflict between patient and surrogate.

Barrier 2: Difficulty applying known values and preferences to the situation

Most surrogates indicated that they had at least some minimal knowledge of the patient’s values and preferences from sources such as prior conversations, understanding the general values of the patient, prior experiences with family illness or formal advance care planning. Despite this knowledge, surrogates still reported challenges in applying this knowledge to the patient’s particular medical situation. This scenario was the most common barrier to surrogate application of values and preferences, occurring in 11/42 (26%) cases overall and in 6/25, (24%) MAs and 5/15 (33%) NHWs. There were four sub-themes related to this barrier.

One subtheme was when the knowledge was not specific enough to guide current clinical decision making, for example: “we didn’t … discuss this specific scenario (Interview 07, MA). In another example, a surrogate questioned how a patient’s previously stated preference against “life support” would apply to different treatments, such as a long-term feeding tube:

… they didn’t want to have life support…you see, there are so many different degrees of life support. We were doing life support really with the feeding tube and stuff, so you know, you don’t think of every scenario

(Interview 33, NHW).

The second sub-theme around applying known preferences occurred when there was concern that patient’s preferences could lead to pain or suffering, generally when the patient had stated a desire to continue intensive treatment.

I know what my dad wanted. He wanted to be resuscitated but then when the doctor came in and told me the process of that, he would probably end up breaking some of his ribs and after that, he’d be in a lot of pain and stuff. I didn’t want Dad to suffer, so … I’m gonna do the do-not-resuscitate, because I don’t want Dad to go through all that.

(Interview 23, MA)

A third sub-theme around barriers in applying known preferences occurred when prior discussions had occurred a long time ago: “So, it’s hard to say what she would want. We just know what she wanted ten, fifteen years ago” (Interview 11, NHW). This concern was generally minor, as surrogates tended to still feel confident that they were making decisions in line with what was most important to the patient.

The final barrier related to application of known preferences occurred when the surrogate perceived that prior patient written preferences to limit treatments were not followed by the healthcare team, leading to some frustration or distress. This situation was rare, with only two surrogates reporting this occurrence. However, these scenarios highlight frustrations that surrogates may experience if they perceive that advance directives are not effectively enacted in practice.

It’s like they didn’t care. They were doing their…what they considered to be their standard emergency room procedure with no regard for any advance directive we provided…it was like they were doing all these tests that were not gonna do anything for my mother, and she had already told them by paper don’t do any of it. So, we were frustrated with that…

(Interview 11, NHW)

As we did not have access to the actual written advance care planning documents, we do not know if the surrogate frustration arose due to their own misunderstanding of scenarios in which the advance care planning documents would be applied, or if existing documents were not available or not correctly interpreted by the medical team.

Barrier 3: Feeling guilt or burden over decision making

Overall, 9/42 (21%) of surrogates expressed either feeling guilty or burdened with the process of making decisions for their loved one, often even in the setting of some knowledge of patient values or preferences. Note that this code was applied only in the setting of clear or overt expressions of guilt or burden by the surrogate; there may have been other cases where surrogates were feeling guilt that were not captured. These expressions of guilt or burden were more prevalent in MAs (7/25, 28%) than NHWs (2/15, 13%) and among cases where the patient died (6/21, 29%) than survived (3/21, 14%). Examples of these expressions of guilt or burden are shown in Table 2. This feeing generally was related to not wanting to make a decision that could result in the death of a family member. Others reported difficulty reconciling their own desire to keep their loved one alive with the patient’s preference to limit intensive treatments.

Table 2: Example quotes demonstrating surrogate expressions of guilt or burden surrounding decision making.

MA examples

NHW examples

You are like now making a decision for somebody else. We do not resuscitate you, that means nobody’s gonna try to help you… they already told me there’s no recovery, but it’s still a hard decision to make, say do not bring this person back…It’s like I’m trying to be God…I don’t want that role. I was feeling hurt. I was feeling why do I have to make this horrible decision, you know. I was feeling like I was a bad person… I was feeling like I was killing my own wife. I think if…as much as I knew my husband’s wills and my mom wills, it doesn’t make any easier that you did it…but I think if they had it in paper and the doctor would say, mira [Eng. Look] sorry, but your husband’s wish was this.…So, we don’t need your signature here. We don’t need your okay. We got the okay from your husband five years ago … So, I think I will be okay with that because it wouldn’t have been me carrying that decision. …my heart was hurting cuz I didn’t wanna do what I knew my dad wanted me to do…but the most important thing was thinking of what he wanted…and honoring his wishes. And that was the hard part because what he wanted wasn’t what I wanted. I wanted to keep him on. I wanted to see… if he can beat this… …who are we really to make that decision for someone else, you know?…And I know that maybe he’s not capable right now to think things out, but the burden falling on just one person is you know, kinda bad…You know, cuz I don’t know…that’s…it’s a big decision.

…when I found out there was no hope of him ever coming out of it, then…because I didn’t want to go against his wishes…So it was an easy decision on that, but it was hard. …I already knew what a DNR was …and I knew what it did and I felt like at the time when I said, you gotta resuscitate him at all means you know, cuz the kids will be upset if I just let him die.

Discussion

In this qualitative interview study, we identified three primary barriers to surrogate application of patient values and preferences after acute stroke: 1) Some surrogates had no prior discussions to determine what the patient would want; 2) Surrogates had some relevant knowledge of what the patient values and preferences, but still struggled with the application to actual decisions for several reasons including lack of specificity; and 3) Surrogates experienced emotional burden or guilt surrounding making important decisions for a loved one. These findings are important for clinicians to be aware of when guiding surrogates through these important decisions, and important for researchers interested in designing interventions to improve patient-centered care, surrogate decision making, and surrogate outcomes.

Our work adds to the existing body of literature of qualitative studies focused on surrogate decision making for adults who lack decision capacity,^21–24 with more specific focus on acute neurological disease. A systematic review of qualitative studies on surrogate decision makers in and out of hospital settings had identified 6 key themes including “struggle and reluctance, seeking reassurance, communication with healthcare professionals, family support, older adults’ wishes, and negative impact.”²² Although we did not formally analyze our results using this framework, we did identify several similar themes including the considerable burden felt by surrogates and concerns that following patient’s previously stated preferences could lead to pain.

Many efforts to improve decisions on life-sustaining treatments to date have focused on improving advance care planning. Our data provide some support for continuing this work while also suggesting opportunities for improvement. While a minority of our participants had completed formal advance care planning documents, some family members were comforted by having patient wishes documented in writing, often even if these discussions occurred a long time ago. Others found that the written advance care planning documents or conversations did not adequately cover the actual decisions they had to make in the hospital. Even in the absence of formal advance care planning, many surrogates still felt that they had sufficient information to guide decision making based on prior discussions or general knowledge of the patient. Only a small number of surrogates felt that they had no relevant knowledge to guide decision making, with some of these opting to continue full intensive treatments. As universal uptake of advance care planning seems unlikely and would still be fraught with challenges of specificity and interpretation, other approaches will be needed. Disease-specific advance care planning might be an option for conditions with a predicable progressive course and known treatment options, but this approach would be more challenging in complex acute-onset diseases such as stroke. Other options to consider could include more research into which values and treatment preferences tend to cluster together,^25,26 tools to predict preferences based on past behavior,²⁷ or other tools or clinician training to help surrogates apply prior general patient statements to the specific decisions to be made.

We also explored ethnic differences in surrogate barriers after stroke. Even though MAs in this community tend to be less likely to have a formal advance care plan than NHWs,⁹ we found that the barriers in application of patient values and preferences were largely similar when comparing MAs and NHWs. The one area of possible difference by ethnicity was that MAs were more likely to express feeling guilt or burden over the process of making a decision that could result in the death of a loved one. Surrogates struggling with the emotional burden of these decisions has been described in several other studies.^22–24 One study specifically focused on diverse populations reported the surrogate sense of burden was common among all race-ethnic groups studied.²⁴ The greater burden reported by MA surrogates in our study warrants further study in other larger populations and in different communities. If these feelings of guilt or burden led to delays or deferral of decisions to limit treatment in the setting of serious illness, this finding might contribute to the differences in end-of-life treatment patterns we have observed in this community.^6,7 Given the multitude of patient, health system, and surrogate factors that go into decisions on timing of treatment limitations, we do not have sufficient sample size to test this hypothesis. We have collected quantitative data on psychological outcomes including depression, anxiety, and post-traumatic stress symptoms in this population; analyses of these outcomes are ongoing and will be reported separately.

To understand reasons for the ethnic differences in treatment patterns, we also explored differences in treatment priorities. Both MAs and NHWs mentioned maintaining patient independence as the most important priority. However, MAs did seem to mention spending time with family as the top priority more commonly than NHWs. While we cannot assess in this analysis whether this prioritization of family contributed to larger-scale differences in treatment patterns, the importance of a strong family orientation is well-described in MAs and other Latinx cultures. The concept of “familism” emphasizes strong family relationships, loyalty, support and mutual obligations including an expectation to provide care for older adults and sick relatives.^28,29 This sense of familism may run counter to the traditional European-American framework prioritizing individual autonomy.^30–32 Further study of the influence of familism on decision making after severe illness may be warranted, and providers caring for MA and other cultures should be aware of this priority.

This work has limitations. Interviews were conducted about 6 months after the time of the stroke, and therefore we were unable to assess how their initial recovery expectations may have influenced this process; recall bias may also be a concern. However, this timing was intentional as to allow surrogates time to reflect on events and be beyond initial bereavement or adaptation to new caregiving roles. Due to multiple overlapping decisions and our modest sample size, we were not able to analyze our findings by type of decision. We were also unable to investigate the impact of factors such as timing of death (early v. late), healthcare costs, or mixed cultural elements within a family and propose that these factors could be addressed in future work. Our interviews were limited to surrogate reports only, and we did not have data from the physicians or copies of any advance care plans. Even though our qualitative population was demographically similar to the overall OASIS cohort, there may be selection bias in those interviewed. This study was done in a single community with a population of mostly female MA surrogates with at least some college education, and results may not be applicable to other populations such as those with a higher proportion of recent immigrants or lower levels of English proficiency. However, this focus on a community with a large MA population and without an academic medical center helps to enhance the diversity of existing literature on surrogate decision making in critical illness.

Conclusion

Surrogates making decisions for a loved one after a stroke may benefit from: 1) Continued efforts to make advance care planning more common and more relevant; 2) Assistance in how to apply their knowledge of patient values to actual treatment decisions; and 3) support to reduce emotional burden. The possibility of greater burden of decision making among MA surrogates warrants further study in larger populations.