Lower Perceived Social Support During Hospitalization by Close Family Members may have Significant Associations with Psychological Distress 1 Month After Cardiac Arrest

Abstract

Background:

The perception of having poor social support is associated with worse symptoms of psychological distress in close family members of critically ill patients, yet this has never been tested after cardiac arrest.

Methods:

Close family members of consecutive patients with cardiac arrest hospitalized at an academic tertiary care center participated in a prospective study. The validated Multidimensional Scale of Perceived Social Support (MSPSS) cued to index hospitalization was administered before discharge. Multivariate linear regressions estimated the associations between the total MSPSS score and total scores on the Patient Health Questionnaire-8 (PHQ-8), Generalized Anxiety Disorder 2-item (GAD-2), and the Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5), assessed 1 month after cardiac arrest.

Results:

In 102 participants (mean age 52 ± 15 years, 70% female, 21% Black, 33% Hispanic) with complete data, the prevalence of depression, generalized anxiety, and probable posttraumatic stress disorder at a median duration of 28.5 days (interquartile range 10–63 days) from cardiac arrest was 61%, 34%, and 13%, respectively. A lower MSPSS score was significantly associated with higher PHQ-8 scores (β = − 0.11 [95% confidence interval − 0.04 to − 0.18]; p < 0.01), even after adjusting for family members’ age, sex, prior psychiatric condition, and witnessing of cardiopulmonary resuscitation and patient’s discharge disposition (β = − 0.11 [95% confidence interval − 0.02 to − 0.15]; p < 0.01). Similarly, significant inverse associations of total MSPSS scores were seen with 1-month GAD-2 and PCL-5 scores.

Conclusions:

Poor social support during hospitalization, as perceived by close family members of cardiac arrest survivors, is associated with worse depressive symptoms at 1 month. Temporal changes in social networks and psychological distress warrant further investigation.

Introduction

Close family members have unique needs due to the unexpected and life-threatening nature of sudden cardiac arrest [1]. Families of cardiac arrest survivors process initial trauma and navigate prognostic uncertainties prevalent throughout hospitalization [2] and beyond [3, 4]. The stressors of an intensive care unit (ICU) stay on caregivers have been thematically studied in the context of patients with severe acute brain injury transitioning out of neurocritical care units [5] and with resiliency-based interventions in various stages of development [6, 7]. Still, the uncertainty and distress perceived by the family members during the cardiac arrest of their loved ones may be unique, with additional applicability to all medical, surgical, and cardiac specialties necessitating an extension of this previous work. Studies assessing the prevalence and risk factors of close family members’ psychological burden are limited by methodological heterogeneity with convenience sampling, inclusion of bereaved families, and lack of racial and ethnic diversity. There is no consistency in the selection of assessment tools, timing of assessment, or key outcome measures [4]. This highlights the need for systematic longitudinal investigations of factors affecting family’s psychological distress to guide interventions reducing caregiving burden and improving overall health.

Low perceived social support correlates with higher distress levels in family members of patients after critical illness [8], traumatic brain injury [9], and spinal cord insult [10]. Despite clear qualitative evidence [4] after cardiac arrest, this relationship has never been quantified. To address this knowledge gap, we sampled a consecutive racially and ethnically diverse group of close family members of cardiac arrest survivors, estimating (1) the prevalence of psychological distress (i.e., depression, generalized anxiety, and posttraumatic stress disorder [PTSD]) 1 month after cardiac arrest and (2) their independent associations (depression as the primary outcome) with perceived social support levels during hospitalization. We further characterize the composition and strength of their social support networks.

Methods

Study Design

We recruited adult close family members of consecutive patients with cardiac arrest admitted between August 16, 2021, and June 28, 2023, to ICUs at Columbia University, a hospital located in the Washington Heights area of Northern Manhattan serving a high-risk, low-resource population comprising 68% Hispanic individuals, 20% White individuals, 33% with limited English, and 20% below the federal poverty line. Consented participants completed a questionnaire before hospital discharge reporting their sociodemographics, psychological risk factors, perceived social support, and network (Table 1). Psychological distress was assessed cross-sectionally via validated instruments during a telephone or in-person visit 1 month after cardiac arrest. Patient-related details, including demographics, hospital outcomes, and discharge dispositions, were obtained through electronic medical records. The study protocol was approved by the institutional review board. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines (see Supplemental Methods).

Table 1.

Characteristics of Family Members of Cardiac Arrest Survivors (N=102)

	% (N)
Family Member Attributes
Age, years (Mean ± Standard Deviation)	52 ± 15
Female Sex	70 (71)
Race/Ethnicity
- Non-Hispanic White	40 (41)
- Hispanic	33 (34)
- Black	21 (21)
- Other	6 (6)
Language
- English	88 (90)
- Spanish	12 (12)
Marital status
- Never married	25 (25)
- Domestic partnership	6 (6)
- Married	56 (57)
- Separated	5 (5)
- Divorced	7 (7)
- Prefer not to answer	2 (2)
Relationship with the patient
Partner/spouse	34 (35)
Child	29 (30)
Parent	12 (12)
Sibling	17 (17)
Extended family or other	8 (8)
Pre-morbid either lived with the patient or provided active assistance	69 (67)
Educational attainment
8th grade or less	1 (1)
Some high school	6 (6)
High school diploma or GED	11 (11)
Trade school/vocational school	1 (1)
Some college, no degree	25 (25)
College degree	31 (32)
Some graduate school, no degree	4 (4)
Graduate degree	20 (20)
Prefer not to answer	2 (2)
Occupational status
Working now, full-time	42 (43)
Working now, part-time	14 (14)
Disabled, permanent or temporary	4 (4)
Temporarily laid off, sick leave, or maternity leave	3 (3)
Keeping house	4 (4)
Student	2 (2)
Unemployed, looking for work	13 (13)
Retired	18 (19)
Types of Co-morbidities
- Heart disease	6 (6)
- Hypertension	29 (29)
- Lung disease	4 (4)
- Diabetes	11 (11)
- Ulcer or Stomach disease	5 (5)
- Kidney disease	3 (3)
- Liver disease	0 (0)
- Anemia or other blood disorders	6 (6)
- Cancer	5 (5)
- Musculoskeletal disorders	34 (32)
- Overweight/Obesity	26 (25)
- Prior Psychiatric condition (e.g., depression, anxiety, post-traumatic stress)	22 (22)
Number of comorbidities
None	23 (23)
1-2	48 (48)
≥3	29 (29)
Witnessed CPR	19 (19)
Positive psychological wellbeing
Optimism (Mean ± Standard Deviation)	16 ± 4
Purpose in Life (Mean ± Standard Deviation)	4.7 ± 0.9
Use spirituality as a coping mechanism
Strongly agree	10 (10)
Agree	1 (1)
Neutral	38 (38)
Disagree	5 (5)
Strongly disagree	46 (46)
Perceived social support
Total score on MSPSS (Mean ± Standard Deviation)	69 ± 15
- Family sub-score	23 ± 5
- Friends sub-score	22 ± 6
- Significant other sub-score	24 ± 6
Low to medium perceived social support (MSPSS total score <61)	24 (24)
Number of supporters reported by family members (n=325)
1	9 (8)
2	14 (12)
3	19 (17)
4	15 (13)
5	43 (38)
Composition of Social Network (N=325)
Friend	22 (73)
Sibling	18 (58)
Child	17 (56)
Parent	9 (28)
Spouse/partner	8 (27)
Other (unspecified relation)	6 (18)
Sister/brother-in-law	5 (16)
Cousin	4 (13)
Aunt/uncle	2 (8)
Niece/nephew	2 (8)
Daughter/son-in-law	2 (5)
Mother/father-in-law	1 (4)
Grandchild	1 (3)
Ex-partner	1 (3)
Congregation or pastor	1 (2)
Business partner	0 (1)
Grandparent	0 (1)
Landlord	0 (1)
Where supporters live (n=319)
With family members	24 (75)
Close to family members	51 (164)
Far from family members	25 (80)
How often supporters get together with family members (n=309)
Every day	34 (104)
A few times a week	21 (66)
Once a week	8 (25)
A few times a month	19 (59)
Less than once a month	18 (55)
How often supporters talk with family members (n=313)
More than once a day	45 (142)
Once a day	17 (54)
A few times a week	24 (75)
Once a week	7 (22)
A few times a month	4 (13)
Less than once a month	2 (7)
Days between cardiac arrest and assessment of social support (Median, Interquartile Range)	15 (7 – 32)
Days between cardiac arrest and assessment of psychological distress (Median, Interquartile Range)	29 (10 – 66)
Patient Attributes
Age, years (Mean ± Standard Deviation)	61 ± 16
Female sex	36 (37)
Independence in daily activities prior to cardiac arrest
Home, unassisted	69 (70)
Home, assisted with activities of daily living	23 (23)
Institutionalized	9 (9)
Out-of-Hospital Cardiac Arrest	22 (22)
Length of ICU stay, days (Median, Interquartile Range)	14 (8 - 26)
Tracheostomy
- Yes	33 (34)
- No	67 (68)
Health insurance
- Uninsured or Medicaid	45 (46)
- Medicare	27 (25)
- Private	28 (28)
Functional Outcomes, modified Rankin score (mRS) at discharge
- mRS 0	1 (1)
- mRS 1	17 (17)
- mRS 2	2 (2)
- mRS 3	23 (24)
- mRS 4	24 (25)
- mRS 5	18 (18)
- mRS 6	15 (15)
Discharge disposition after cardiac arrest
- Home	41 (42)
- Facility (Acute rehab/Sub-acute rehab/LTAC)	44 (45)
- Died during index hospitalization after study completion	15 (15)

CPR cardiopulmonary resuscitation, GED general education diploma, ICU intensive care unit, LTAC long-term acute care, mRS modified Rankin score, MPSS Multidimensional Scale of Perceived Social Support

Study Participants

Inclusion criteria included close family members of adult (≥ 18 years) patients with either in-hospital or out-of-hospital cardiac arrest. To be classified as a “close family member”, participants were required to be either the patient’s designated health care proxy or someone with a close immediate relationship to the patient who was present at the bedside during the consent process. Simply being present at the bedside was not sufficient; in cases in which multiple individuals were present, they were asked to specify who the patient’s primary contact or caregiver was. This procedure helped identify those individuals who were most closely connected to the patient and most affected psychologically by the traumatic event. Exclusion criteria were (1) the patient died within 30 days, (2) unable to visit during working hours, and (3) neither English nor Spanish speaking.

Study Measures

Indicators of Psychological Distress as Outcomes

For the primary outcome, the Patient Health Questionnaire-8 (PHQ-8), a validated eight-item self-report instrument, measured current depressive symptoms. A total score of 5–9 indicated mild depressive symptom burden, and a score ≥ 10 indicated major depressive symptom burden. For secondary outcomes, clinically significant generalized anxiety symptoms and probable PTSD symptoms were measured using the validated Generalized Anxiety Disorder 2-item (GAD-2 score ≥ 3) and the 20-item PTSD Checklist for DSM-5 (PCL-5 score ≥ 31), respectively. Primary and secondary outcomes were measured at a median of 29 days (interquartile range [IQR] 10–66) after cardiac arrest (Table 1).

Perceived Social Support as a Primary Predictor

The 12-item Multidimensional Scale of Perceived Social Support (MSPSS) [11] assessed participants’ perceived levels of social support during cardiac arrest hospitalization on a Likert scale with three subscales for support from family, friends, and significant others. MSPSS was completed at a median of 15 days (IQR 7–32) after cardiac arrest (Table 1). Continuous measures of MSPSS were used for primary analyses. To further characterize the composition and strength of their social support networks, participants were asked to list up to five supporters and their relationships, living proximity, and typical frequency of meeting and talking since the cardiac arrest (Supplemental Methods).

Statistical Analysis

Continuous variables were summarized as mean (SD) or median (IQR), whereas categorical variables were presented as frequency (percentage). Fisher’s exact test and the Wilcoxon rank-sum test were used to compare participants in each group.

Selection of Covariates and Models

We followed the recommendation that covariates be selected for inclusion a priori [12]. Based on published findings of factors that might confound the association between perceived social support and depression, demographic variables (age, sex) and family members’ psychiatric history were treated as a priori covariates. Younger age [8, 13–15], female sex [13, 14, 16], witnessing cardiac arrest [14, 16–18], and preexisting psychiatric disorders [8, 14, 16, 19, 20] (i.e., predisposition to anxiety/depression/PTSD to the traumatic events in the past) are known risk factors of development of future depression to traumatic events.

To explore other independent associations between potential predictors and family’s symptoms of depression, univariate analyses were performed with several items, and a p value of 0.01 was considered to be significant and included in the multivariate models. The factors considered were among (1) family member attributes (e.g., sociodemographic variables, including marital status, relationship to the patient, number of current medical comorbidities), (2) psychological risk factors (e.g., whether family members had witnessed the cardiac arrest; factors associated with positive psychological well-being, including optimism and purpose in life; preexisting diagnosis of psychiatric disorders, such as depression, anxiety, or PTSD; or current use of spirituality as a coping mechanism), (3) patient-related characteristics (e.g., age, sex, health insurance, location of arrest, length of ICU stay), and (4) hospital-related outcomes (e.g., was patient discharged with a tracheostomy, their discharge functional status, and dispositions).

We used hierarchical linear regressions to analyze the associations between total PHQ-8 and MSPSS scores. In model 1, we adjusted for demographics (age, sex); model 2 included other family attributes (presence of prior psychiatric history, witnessed cardiopulmonary resuscitation [CPR]); model 3 incorporated discharge disposition (home vs. inpatient rehabilitation facility) added after showing significant (p < 0.01) bivariate differences (Supplementary Table 3). After estimating the β coefficients and their significance, to provide a comprehensive understanding of the associations, we calculated (1) the R² value to indicate the overall strength of the relationship and a measure of the model’s goodness of fit, (2) the effect size for individual predictors using η², and (3) confidence intervals for effect sizes to convey the precision of estimates.

Our models fulfilled assumptions for linear regressions. The variables did not present with multicollinearity; the variance inflation factors of the multivariate models were < 1.5. To ensure the robustness of our regression model, we employed both the Linktest and the Ovtest. The Linktest assessed potential specification errors by examining the significance of the squared predicted values. The test revealed a coefficient of 0.02 with a standard error of 0.03 for the squared predicted values, with a p value of 0.6 indicating that the model specification was well specified. Additionally, the Ovtest, which evaluates whether higher-order terms of the fitted values improve the model, yielded an F-statistic of 1.2 and a p value of 0.3. This result supports the adequacy of our model specification.

Sample Size and Power Analysis

The finding that MSPSS was significantly associated with PHQ-8 in the fully adjusted model suggests that limited power was not an issue. Nevertheless, we performed post hoc power analysis using an R² test in a multiple linear regression to test the significance of all coefficients included in the final model. With α = 0.05, R² = 0.33, and 6 tested covariates, the estimated power was greater than 95%. All statistical analyses were performed using STATA 18.

Results

Study Population

Of 438 cardiac arrest admissions, 247 met exclusions. Of 191 eligible participants approached, 137 who consented and 102 with complete 1-month data were analyzed (consolidated standards of reporting trials diagram, Supplementary Fig. 1).

Close Family Member Characteristics

The average age of participants was 52 ± 15 years. The majority identified as female (n = 71, 70%) and from minority racial and ethnic backgrounds (n = 61, 60%), with Hispanic (n = 34, 33%) and Black (n = 21, 21%) being the most represented groups. The most common relationship to the patient was spouse or partner (n = 35, 34%). Additionally, 22 participants (22%) had psychiatric history, and 19 (19%) had witnessed CPR (see Table 1).

Notable characteristics among the close family members included having an education level of high school or general education diploma or less (n = 18, 18%), reporting no current employment (n = 41, 40%), and managing uninsured patients or on Medicaid (n = 45, 45%). Fewer participants reported speaking only Spanish (n = 12, 12%), and those finding spirituality to be a significant coping mechanism were also less common (n = 11, 11%).

Patient Characteristics

Almost all patients lived at home before cardiac arrest (n = 93, 92%). After a median ICU length of stay of 14 days (IQR 8–26), 34 (33%) required tracheostomy, 58 (57%) had poor functional status (modified Rankin score > 3) at hospital discharge, and 42 (41%) went home.

Prevalence of Depressive, Generalized Anxiety, and PTSD Symptoms 1 Month After Cardiac Arrest

Of 100 participants with complete PHQ-8 scale data assessed at a median duration of 28.5 (IQR 10–63) days after cardiac arrest, 61 (61%) screened positive for current depression (35 [35%] with mild and 26 [26%] with major depression). The prevalence of generalized anxiety and probable PTSD was 34 (34%) and 13 (13%), respectively (Fig. 1).

Figure 1.

Prevalence Estimates of Depression, Generalized Anxiety, and Post-traumatic Stress Disorder at 1 month in Close Family Members of Cardiac Arrest Survivors

Associations of MSPSS Total Score and Subscale Scores with PHQ-8 Total Score

Lower total MSPSS scores were significantly associated with higher PHQ-8 scores (Table 2), both in bivariate analysis (β = − 0.1, p ≤ 0.01) and after adjusting for age, female sex, witnessed CPR, psychiatric history, and discharge disposition (β = − 0.1, p = 0.01; R² = 0.33; medium-to-large effect size, η² = 0.1, 95% confidence interval 0.002–0.2). Significant inverse associations with depressive symptoms existed in a multivariate model for MSPSS family (β = − 0.3, p ≤ 0.01) and significant other subscales (β = − 0.2, p = 0.02) but not for MSPSS friends (β = − 0.1, p = 0.2).

Table 2.

Associations Between Perceived Social Support by Close Family Members during Cardiac Arrest hospitalization and Depressive symptoms at 1 Month

Covariates	Bivariate Model	Coefficient (95% CI)	P value	Multivariate Model 1	Coefficient (95% CI)	P value	Multivariate Model 2	Coefficient (95% CI)	P value	Multivariate Model 3	Coefficient (95% CI)	P value
Family: MSPSS total score	− 0.1	(−0.18, −0.04)	<0.01	−0.1	(−0.19, −0.05)	<0.01	−0.1	(−0.17, −0.04)	<0.01	−0.1	(−0.15, −0.02)	0.01
Family: Age	−0.1	(−0.14, 0.01)	0.07	−0.1	(−0.15, −0.01)	0.03	−0.1	(−0.13, −0.002)	0.04	−0.1	(−0.14, −0.01)	0.02
Family: Female sex	−2.2	(−4.56, 0.17)	0.07	−2.9	(−5.19, −0.71)	0.01	−2.6	(−4.70, −0.45)	0.01	−1.9	(−4.08, 0.26)	0.08
Family Prior Psychiatric History	4.4	(2.0, 7.0)	<0.01				3.7	(1.36, 6.02)	<0.01	3.4	(1.06, 5.68)	<0.01
Family Witnessed CPR	3.3	(0.6 – 6.0)	0.01				3.2	(0.76, 5.61)	0.01	3.3	(0.96, 5.74)	<0.01
Patient: Discharge Disposition
Home	reference									reference
Facility	−3.3	(−5.64, −1.04)	<0.01							−2.3	(−4.49, −0.17	0.03
Died	−0.1	(−3.25, 3.13)	0.9							−0.03	(−2.93, 2.98)	0.9

Associations of MSPSS Total Scores with GAD-2 and PCL-5 Total Scores

Lower total MSPSS scores were significantly associated with higher GAD-2 (β = − 0.03, p = 0.01) and PCL-5 total scores (β = − 0.30, p < 0.01) in an adjusted model (Supplementary Tables 1 and 2).

Characterization of Social Support Networks of Close Family Members of Hospitalized Patients with Cardiac Arrest

Of 325 total supporters (a median of 4 [IQR 3–5] supporters per participant [n = 88]), the most frequently mentioned were friends (n = 73, 22%), siblings (n = 58, 18%), and children (n = 56, 17%). Supporters most frequently lived close by (n = 164, 51%), met with participants daily to several times weekly (n = 170, 55%), and talked at least once/day (n = 196, 62%) (Table 1).

Sensitivity Analysis

We found similar significant associations between the MSPSS total score and PHQ-8 as a dichotomous outcome (Supplemental Table 4). Associations of PHQ-8 scores with dichotomized MSPSS were also significant (Supplemental Results).

Comparison of Family and Patient-Related Characteristics by Cardiac Arrest Location

Our subgroup analysis revealed no significant differences between in-hospital (78%) and out-of-hospital (22%) cardiac arrests regarding the total score on the MSPSS, its subdomains, the PHQ-8, the PCL-5, the GAD-2, length of ICU stay, or the proportion of patients discharged with a poor functional outcome (modified Rankin score ≥ 4) (see Supplemental Table 5).

Discussion

Close family members of cardiac arrest survivors with lower perceived social support during hospitalization report poor psychological outcomes 1 month after cardiac arrest. This finding is consistent with other studies on families of critically ill patients [8]. The long-term stress-related consequences of hospitalization and beyond include increased risk for caregiving burden [20] and cardiovascular morbidities, such as coronary heart disease or stroke [21]. Social supports are potentially modifiable resources that can be leveraged to promote family health. Interventions targeting qualitative aspects of social relationships, such as network interventions, social prescribing, and support groups, present a promising strategy [22]. Future work should also test whether interventions of informational support reducing uncertainty among families about prognosis and recovery [3, 23] or cognitive behavioral therapies counteracting maladaptive social cognitions around loneliness [10] are effective in improving the perception of support and psychological well-being. Additionally, early screening of close family members during hospitalization may help identify vulnerable individuals who may benefit from such targeted interventions. Such a holistic patient–family approach holds wide-ranging individual and societal benefits, including the efficiency of health care spending resulting from the long-term care provided by close family members.

Additional family-related attributes contributing to distress were younger age, psychiatric history, and witnessed CPR. They constituted our a priori covariates derived from the previous literature on patient [14] and family [8, 17] distress. Notably, compared to rehabilitation facilities, discharge disposition being home was the only patient-related characteristic demonstrating a significant association with family distress. This may be attributed to the family’s lack of formal preparation for caregiving and expectations at home after cardiac arrest. In this cohort, 23% of patients with poor functional status were discharged home. Although functional status lacked a significant relationship with the family’s depression, a dyadic approach accounting for the patient’s cognitive and psychological deficits is needed to understand the protective effect of rehabilitation on a family’s mental health.

Despite friends being the most frequent supporters listed, the MSPSS subscale analysis revealed that support from family and significant others drove the overall association between perceived social support and distress. Studies employing a quantitative social network analysis approach [22] are warranted to understand the underpinnings of network structure and composition on family distress.

Our findings are particularly urgent, with 61% of close family members exhibiting symptoms of depression and 29% screening positive for major depression 1 month after cardiac arrest. These surpass previously reported rates among survivors (8–18%) [24] and their close family members (14–17%) at 3 months and 12 months after cardiac arrest [17, 20]. The differences in prevalence are most likely explained by the variability in the timing of the assessments and the instruments used to measure depressive symptoms. However, our findings are comparable to depression rates within 3 months of other ICU-based conditions (34–43%) [8]. Notably, 16% of these family members had persistent depressive symptom burden over time, highlighting the importance of early identification and intervention for this vulnerable group [20].

Study Strengths

This study has several notable strengths. To our knowledge, this study recruited the largest sample size of family members of patients with cardiac arrest to date. Our sample included a diverse group of family members, with ages ranging from 18 to 80 years, a wide race and ethnicity distribution, and both in-hospital and out-of-hospital cardiac arrests. This speaks to the high generalizability of our findings.

This is also the first study, to our knowledge, to examine the role of perceived social support in depressive outcomes with a focus on family members of patients with cardiac arrest. These findings help fill the gap in the literature, which predominantly features cardiac arrest patient outcomes.

Study Limitations and Next Steps

There are also limitations to this study. First, given that the psychological assessments were conducted at a single time point, it is important to measure changes in perceived social support and psychological distress over time to gain a clearer understanding of their relationships and inform intervention development. It is important to acknowledge that baseline assessments of psychological distress in family members prior to the cardiac arrest were not available, which limits our ability to measure changes in emotional states during hospitalization and beyond. Because of the acute nature of ICU conditions, this limitation is inherent. However, previous studies investigating the psychological impact of cardiac arrest on survivors have addressed this issue by using psychiatric history as a proxy marker. This approach has been shown to be strongly associated with the development of psychological distress following traumatic events [14]. Similarly, we incorporated psychiatric history as an a priori covariate in our study. Future research could benefit from assessing longer-term outcomes by using 1-month scores as the baseline, which would facilitate understanding of temporal associations and inform the development of future interventions.

There was also a lack of separation between participants whose close family members experienced an in-hospital cardiac arrest and those who experienced an out-of-hospital cardiac arrest. Although further analysis indicated no significant differences between the two groups, there may be some unmeasured confounding that could be investigated in a larger cohort.

Regarding model selection, the consistency between the significant unadjusted and the fully adjusted association between the perceived social support and family members’ psychological distress scores further supports our decision to control for theoretically important covariates. Concerns about model overfitting led us not to model other variables that may influence this association (e.g., survivor’s comorbidities, length of ICU stay, or family’s social determinants, including race and ethnicity or educational, occupational, or insurance status). However, these factors were nonsignificant in univariate analysis, which diminished our concern about omitting them from the models.

This study identified a notable correlation between perceived social support and psychological distress among close family members. However, as an observational study, it does not establish causality. Although we accounted for several confounding variables in our analysis, there remains a possibility of residual confounding that could affect the observed relationship.

Additionally, our characterization of family members’ social support networks is limited by only allowing participants to report up to five people who have supported them since the cardiac arrest. To fully map social support networks, future studies should prompt participants to provide an exhaustive list of all supporters, the nature of their relationships, and the strength of their relationships. A more complete understanding of the key individuals composing family members’ support networks is essential for building future interventions.

Conclusions

Overall, our study demonstrated the feasibility of recruiting a diverse group and comprehensively assessing key family- and patient-related factors in the short-term development of psychological symptoms. These findings need to be confirmed in a larger funded, longitudinal study with in-depth quantitative and qualitative assessments of family members’ social support networks and the patient’s cognitive and psychological status at multiple time points following cardiac arrest.