Predictors of ICU Surrogates’ States of Concurrent Prolonged Grief, Posttraumatic Stress, and Depression Symptoms*

Abstract

OBJECTIVES:

Scarce research explores factors of concurrent psychologic distress (prolonged grief disorder [PGD], posttraumatic stress disorder [PTSD], and depression). This study models surrogates’ longitudinal, heterogenous grief-related reactions and multidimensional risk factors drawing from the integrative framework of predictors for bereavement outcomes (intrapersonal, interpersonal, bereavement-related, and death-circumstance factors), emphasizing clinical modifiability.

DESIGN:

Prospective cohort study.

SETTING:

Medical ICUs of two Taiwanese medical centers.

SUBJECTS:

Two hundred eighty-eight family surrogates.

INTERVENTIONS:

None.

MEASUREMENTS AND MAIN RESULTS:

Factors associated with four previously identified PGD-PTSD-depressive-symptom states (resilient, subthreshold depression-dominant, PGD-dominant, and PGD-PTSD-depression concurrent) were examined by multinomial logistic regression modeling (resilient state as reference). Intrapersonal: Prior use of mood medications correlated with the subthreshold depression-dominant state. Financial hardship and emergency department visits correlated with the PGD-PTSD-depression concurrent state. Higher anxiety symptoms correlated with the three more profound psychologic-distress states (adjusted odds ratio [95% CI] = 1.781 [1.562–2.031] to 2.768 [2.288–3.347]). Interpersonal: Better perceived social support was associated with the subthreshold depression-dominant state. Bereavement-related: Spousal loss correlated with the PGD-dominant state. Death circumstances: Provision of palliative care (8.750 [1.603–47.768]) was associated with the PGD-PTSD-depression concurrent state. Surrogate-perceived quality of patient dying and death as poor-to-uncertain (4.063 [1.531–10.784]) correlated with the subthreshold depression-dominant state, poor-to-uncertain (12.833 [1.231–133.775]), and worst (12.820 [1.806–91.013]) correlated with the PGD-PTSD-depression concurrent state. Modifiable social-worker involvement (0.004 [0.001–0.097]) and a do-not-resuscitate order issued before death (0.177 [0.032–0.978]) were negatively associated with the PGD-PTSD-depression concurrent and the subthreshold depression-dominant state, respectively. Apparent unmodifiable buffering factors included surrogates’ higher educational attainment, married status, and longer time since loss.

CONCLUSIONS:

Surrogates’ concurrent bereavement distress was positively associated with clinically modifiable factors: poor quality dying and death, higher surrogate anxiety, and palliative care—commonly provided late in the terminal-illness trajectory worldwide. Social-worker involvement and a do-not-resuscitate order appeared to mitigate risk.

KEY POINTS.

Question: What are the factors associated with ICU bereaved family surrogates’ membership in the prolonged grief disorder (PGD)-posttraumatic stress disorder (PTSD)-depressive-symptom states?

Findings: Surrogates’ concurrent bereavement distress was positively associated with clinically modifiable factors: poor quality dying and death, higher surrogate anxiety, and palliative care—commonly provided late in the terminal-illness trajectory, whereas social-worker involvement and a do-not-resuscitate order mitigated this risk.

Meaning: By improving quality of dying and death, providing early palliative care, facilitating social-worker involvement and a do-not-resuscitate order, and alleviating surrogates’ anxiety symptoms during bereavement, healthcare professionals may prevent concurrent symptoms of PGD, PTSD, and depression among bereaved surrogates of ICU decedents.

Death in an ICU is common (1), increasing (2), and costly (3), especially since COVID-19 (4). Meanwhile, quality of end-of-life (EOL) care in ICUs is improving but still poor (5, 6). Improving EOL ICU care is a priority (7), particularly for family surrogates (8, 9). ICU loss is traumatic and painful (10, 11). Although most adjust over time (12), bereaved surrogates may suffer prolonged grief disorder (PGD) (13), posttraumatic stress disorder (PTSD) (14), and depression (15). PGD, PTSD, and depression are distinguishable (16, 17) and co-occur more often than not (18), synergistically worsening physical/mental health and social functioning (19–22). We previously identified among ICU family surrogates four PGD-PTSD-depressive-symptom states: resilient, subthreshold depression-dominant, PGD-dominant, and PGD-PTSD-depression concurrent states (23).

Co-occurrence of PGD, PTSD, and depression symptoms manifests as distinct latent patterns among bereaved adults (19–27). In prior studies, bereaved adults survived anthropogenic, traumatic, violent, or suicidal losses (19, 20, 22, 27) or lost their loved one predominantly (80–91.3%) from illness not specifically in ICUs (21, 24–26), although ICU literature suggests investigating factors of psychologic distress from a broader stress-response perspective (11). Furthermore, most prior studies (19–22, 24–27) identified psychologic-distress patterns and predictors cross-sectionally, which cannot determine directional relationships between factors and psychologic-distress states nor consistency of factor associations over time.

Predictors of psychologic-distress classes have rarely been comprehensively explored. The integrative framework of predictors for bereavement outcomes (28) includes: 1) intrapersonal risk factors, 2) interpersonal risk factors, 3) appraisal and coping, 4) bereavement-related stressors and 5) death circumstances. All existing studies explored intrapersonal risk factors including demographics (gender [19, 21, 22, 24, 25], preexisting mental health conditions [20], age [22, 24–26], education [20, 24, 26, 27], and financial adequacy [20]) and bereavement-related factors (kinship [20–22, 24–27], time since loss [19, 21, 22, 24–26], and cause of loss [natural vs. unnatural] [21, 24–26]). Four studies explored appraisal and coping factors (e.g., sense of unrealness, meaning-making, and anxious or depressive avoidance) (19, 21, 25, 26). Only one examined interpersonal risk factors like social support (20).

Lastly, modifiable clinical predictors of concurrent symptoms of PGD, PTSD, and depression have never been studied among bereaved ICU surrogates. Modifiable factors present the greatest potential for improving clinical EOL care practice. Indeed, co-occurrence of complicated grief and PTSD-related symptoms were associated with whether the patient died while intubated, family had said goodbye to the patient, and family was present at time of death (13). Therefore, we drew from the integrative framework of predictors for bereavement outcomes (28), emphasizing modifiable factors of PGD-PTSD-depressive-symptom states among bereaved ICU family surrogates. This study augments previous analyses of the same dataset: examinations of associations of severe anxiety and depressive symptoms (29) and clinically significant PTSD symptoms (30) in the first bereavement year with objective process-based indicators of high-quality EOL ICU care and subjective family-assessed satisfaction with ICU care, as well as the reflexive, intertwined nature of psychologic bereavement outcomes (31, 32).

MATERIALS AND METHODS

Study Design, Setting, and Study Participants

This study extends our previous identification of four PGD-PTSD-depressive-symptom states for bereaved ICU surrogates (23) using data from our longitudinal, observational study on effects of quality of EOL ICU care on family surrogates’ bereavement adjustment (29–32). Briefly, we consecutively recruited the primary family-surrogate decision maker of critically ill patients (Acute Physiology and Chronic Health Evaluation II scores > 20) from level III medical ICUs at two academically affiliated Taiwanese hospitals from January 2018 to March 2020 and followed them through July 2022. Each enrolled surrogate signed informed consent for participation. This study was approved by the Chang Gung Medical Foundation Institutional Review Board (201700343B0; March 20, 2017; Impact of end-of-life care quality in intensive care units on adjustment of bereaved family members). Study procedures followed ethical standards of the responsible committee on human experimentation (institutional) and of the Helsinki Declaration of 1975.

Measures

Outcome Variable: PGD-PTSD-Depressive Symptoms States

Symptoms of PGD, PTSD, and depression were measured by the Prolonged Grief (PG)-13 scale (33), Impact of Event Scale-Revised (IES-R) (34), and Hospital Anxiety and Depression Scale (HADS) (35), respectively. Instrument scoring details are in Online Data Supplement 1 (http://links.lww.com/CCM/H585). Identification and description of symptom states (resilient, subthreshold depression-dominant, PGD-dominant, and PGD-PTSD-depression concurrent states) (23), 6–24 months post-loss are detailed in Online Data Supplement 2 (http://links.lww.com/CCM/H585), and significant differences in symptoms across the four states were reported (23) (Fig. E1, http://links.lww.com/CCM/H585).

Independent Variables

Factors associated with symptom states were examined based on the integrative framework of predictors for bereavement outcomes (28), except for appraisal and coping. Measurement details are in Online Data Supplement 1 (http://links.lww.com/CCM/H585).

Intrapersonal risk factors included surrogates’ sociodemographics and personal vulnerabilities, for example, financial hardship, preexisting physical-mental health problems (36), and anxiety symptoms.

Interpersonal risk factors were indicated by perceived social support measured by the Medical Outcomes Study Social Support Survey (MOS-SSS) (37).

Bereavement-related stressors included type of loss, patient demographics and clinical characteristics, and time since loss.

Death circumstances were indicated by process-based indicators of high-quality EOL care in ICUs (38) (hereafter as care-quality indicators; Online Data Supplement 1, http://links.lww.com/CCM/H585) and surrogates’ perceived quality of the patient’s dying and death in ICUs by the quality of dying and death (QODD) questionnaire in ICU (39). Latent class analysis was used to identify QODD latent classes (40): high, moderate, poor to uncertain, and worst QODD class (Online Data Supplement 3, http://links.lww.com/CCM/H585). QODD classes differ by physical symptom control, emotional preparedness for death, and amount of life-sustaining treatments (LSTs) received (40) (Table E1, http://links.lww.com/CCM/H585).

Data Collection

Time-invariant intrapersonal risk factors and bereavement-related stressors were collected at enrollment. Patient clinical characteristics and care-quality indicators were abstracted from medical records throughout admission. QODD was assessed at 1 month post-loss. Surrogates’ grief-related psychologic distress (including anxiety symptoms) and perceived social support were assessed by phone interviews at 1, 3, 6, 13, 18, and 24 months post-loss to comply with the greater than or equal to 1 month duration criterion for PTSD (33) and to avoid measuring the anniversary effect.

Data Analysis

Factors associated with PGD-PTSD-depression-symptom states were identified by a multinomial logistic regression model in Latent GOLD 5.1 (Statistical Innovations Inc. Belmont, MA) using resilient state as reference. Lagged time-varying variables of anxiety symptoms and perceived social support were measured in the prior wave of assessment to establish temporal precedent to the outcome variable. The effect of each independent variable was represented as adjusted odds ratio (AOR) with 95% CI.

RESULTS

Participant Characteristics

A total of 288 bereaved family surrogates provided sufficient data on independent and outcome variables to constitute the sample. Detailed characteristics of the whole sample and across the four PGD-PTSD-depression-symptom states are in Table E2 (http://links.lww.com/CCM/H585). Most surrogates were female (58.0%), married (74.0%), the patient’s adult child (53.5%), and on average (sd) 49.6 years old (12.5 yr old). Few had preexisting physical-mental problems that required emergency department (ED) visits (6.6%) or hospitalization (4.5%). Overall anxiety symptoms were low (3.9 [3.7]/21) over the first two bereavement years.

Factors Associated With Membership in PGD-PTSD-Depression-Symptom States

Surrogate demographics were generally not associated with membership in PGD-PTSD-depression-symptom states (Table E3, http://links.lww.com/CCM/H585). Educational attainment greater than or equal to senior high school and married status decreased surrogates’ membership in the PGD-PTSD-depression concurrent (AOR [95% CI] = 0.114 [0.016–0.803]) and PGD-dominant (0.113 [0.035–0.365]) state, respectively. In contrast, personal vulnerability was significantly associated with PGD-PTSD-depression-symptom state membership. Financial hardship (4.983 [1.332–18.632]) and ED visits in the year before the patient’s critical illness (98.790 [6.607–1477.047]) increased membership in the PGD-PTSD-depression concurrent state. Prior use of mood medications increased membership in the subthreshold depression-dominant state (6.234 [1.162–33.439]). Higher anxiety symptoms increased subsequent membership in the three more profound psychologic-distress states (ranged 1.781 [1.562–2.031] to 2.768 [2.288–3.347]).

Surrogates in the PGD-dominant and the PGD-PTSD-depression concurrent states perceived substantially lower social support than those in the resilient state (mean [sd] = 67.9 [20.2] and 57.2 [15.0] vs. 73.7 [12.2]) (Table E2, http://links.lww.com/CCM/H585), but differences were not statistically significant primarily due to small states (i.e., small numbers of participants in the PGD-dominant and the PGD-PTSD-depression concurrent states [Table E3, http://links.lww.com/CCM/H585]). However, better perceived social support increased surrogates’ membership in the subthreshold depression-dominant state (1.053 [1.019–1.089] per unit increase in MOS-SSS score).

Bereavement-related stressors were not associated with surrogates’ membership in PGD-PTSD-depression-symptom states, except for type of loss and time since loss (Table E3, http://links.lww.com/CCM/H585). Spousal surrogates were more likely than other relationships (e.g., surrogates other than spouse or adult child) to be in the PGD-dominant (11.235 [1.833–68.855]) state. Membership decreased in the three more profound psychologic-distress states over time, reaching statistical significance for the PGD-dominant (0.328 [0.131–0.820]) state between 6 and 18 months post-loss.

Death-circumstances evident as care-quality indicators were primarily not associated with membership in PGD-PTSD-depression-symptom states (Table E3, http://links.lww.com/CCM/H585). However, social-worker involvement (0.004 [0.001–0.097]) and a do-not-resuscitate (DNR) order issued before death (0.177 [0.032–0.978]) decreased membership in the PGD-PTSD-depression concurrent and subthreshold depression-dominant state, respectively. In contrast, palliative care increased membership in the PGD-PTSD-depression concurrent state (8.750 [1.603–47.768]).

QODD classes were significantly associated with PGD-PTSD-depression-symptom states (Table E3, http://links.lww.com/CCM/H585). Poor-to-uncertain or worst QODD classes generally led to the three more profound psychologic-distress states than did the high QODD class. Surrogates whose loved one was evaluated as in the poor-to-uncertain QODD class were significantly more likely than those of the high QODD class to be in the subthreshold depression-dominant (4.063 [1.531–10.784]) and the PGD-PTSD-depression concurrent (12.833 [1.231–133.775]) states. Furthermore, the worst QODD class increased surrogate membership in the PGD-PTSD-depression concurrent state (12.820 [1.806–91.013]).

DISCUSSION

We confirmed the utility of the integrative framework of predictors for bereavement outcomes (28) in determining bereaved ICU surrogate membership in PGD-PTSD-depression-symptom states. Factors of the three more profound psychologic-distress states included surrogate financial hardship, ED visits and use of mood medications in the year before the patient’s critical illness, higher anxiety symptoms and better perceived social support in the prior wave of assessment, spousal loss, provision of palliative care, and poor-to-uncertain or worst QODD class (Table E3, http://links.lww.com/CCM/H585). Buffering factors included surrogates’ higher educational attainment, married status, time since loss, social-worker involvement in EOL care, and a DNR order issued before the patient’s death.

From the intrapersonal risk-factor domain, we confirmed vulnerability of those with lower education (20, 24, 26) and financial inadequacy (20) for more profound/persistent psychologic distress as shown by increased membership in the PGD-PTSD-depression concurrent state in our study. Lower educational attainment is closely associated with financial hardship (41). Financial toxicity/distress from critical illness is common for family surrogates (42) and increases uncertainty while making LST decisions (42). Echoing a recent report of increased suicidal ideation among bereaved cancer caregivers with financial hardship (43), our finding highlights the negative effect of financial hardship on suffering PGD-PTSD-depression concurrent state during bereavement.

We also confirmed preexisting mental health problems pose vulnerability by increased membership in the three more profound psychologic-distress states. Specifically, use of mood medications in the year before the patient’s critical illness and higher anxiety symptoms in the prior wave of assessment increased membership in the subthreshold depression-dominant state and the three more profound psychologic-distress states, respectively. Our findings contrast with no observed relationship between preexisting mental-health status and group membership for bereaved 9/11 family members (20) and an association between use of mood medication and higher PTSD symptoms for bereaved ICU family members (13). Our inability to find associations between use of mood medications and membership in the PGD-dominant state may be attributable to insufficient power, despite higher use among surrogates in this more profound psychologic-distress state than in the resilient state (5.6% vs. 2.2%) (Table E2, http://links.lww.com/CCM/H585). However, we confirmed the important role of preexisting anxiety symptoms in positive associations with depression (44), PTSD symptoms (14), and prolonged-grief symptoms (18) using a lagged approach to establish temporal precedence, acknowledging that temporal associations of anxiety symptoms with subsequent depression and PTSD symptoms are complex (31). Finally, we observed that ED visits in the year before the patient’s critical illness increased membership in the PGD-PTSD-depression concurrent state. Whether ED visits indicate preexisting mental vulnerability or whether loss in an ICU reactivates feelings of fear and powerlessness from prior ED visits to produce profound PGD, PTSD, and depressive symptoms during bereavement warrants further investigation preferably by qualitative research.

Better perceived social support increased surrogate membership in the subthreshold depression-dominant state, which seems counterintuitive and contrasts with the finding that greater satisfaction with social support predicted lower likelihood in the comorbid anxiety/depression/PTSD with grief-related functional impairment group among bereaved individuals who survived the September 11, 2001, terrorist attacks in New York City (20). We speculate that social support facilitates natural grieving (elevated depression symptoms below the clinical threshold) (Fig. E1C, http://links.lww.com/CCM/H585) to prevent severe PGD or PTSD symptoms and acknowledge potentially insufficient power to detect lower membership in the two more profound psychologic-distress states than in the resilient state.

Regarding bereavement-related stressors, we confirmed kinship and time since loss were associated with surrogates’ membership in PGD-PTSD-depressive-symptom states. Spousal loss was associated with increased membership in the PGD-dominant state, consistent with the observation that losing a partner was associated with increased membership in the PGD class (24–27) or the high-PGD moderate-depression/PTSD class (21), although no association (22) and decreased membership in comorbid anxiety/depression/PTSD with grief-related functional impairment (20) were also reported in prior studies for bereaved individuals. Loss of the long-lasting, more intimate spousal relationship than other relationships may bring family surrogates painful grief reactions of yearning, longing for and/or a persistent preoccupation with thoughts and memories of the deceased; marked sense of disbelief; difficulties with acceptance; and anger-characterized as PG symptoms (33).

In addition, we found PGD-dominant state membership declined over time consistent with reports that less time since loss brought increased membership in the PGD class (22, 25, 26) or the high-PGD moderate-depression/PTSD class (21), despite no significant difference in time since loss across classes reported in prior studies (19, 24). Our finding supports gradual adjustment to bereavement (12).

Regarding death circumstances, palliative care, a DNR order, social-worker involvement, and QODD classes were associated with PGD-PTSD-depressive-symptom states. A DNR order decreased surrogates’ membership in the subthreshold depression-dominant state partially consistent with a report that a DNR order predicts improved mental health after loss (45). A DNR order may indicate surrogates’ preparedness for loss and prevent unnecessary patient suffering from potentially ineffective but frightening cardiopulmonary resuscitation, thereby facilitating bereavement adjustment to decrease membership in the subthreshold depression-dominant state. Social-worker involvement decreased membership in the PGD-PTSD-depression concurrent state. Social workers in Taiwan, beyond providing psychologic support, primarily seek financial resources for patients who need a subsidy, thereby mitigating financial hardship in providing EOL care, which in turn decreased surrogates’ membership in the PGD-PTSD-depression concurrent state.

Counter to our expectation, palliative care increased membership in the PGD-PTSD-depression concurrent state. We speculated that the failure of palliative care to relieve surrogates’ psychologic distress may be related to late referral, limiting realization of the full benefits of palliative care for which an optimal duration of at least 3–4 months was suggested (46). However, in international routine practice, the median duration from initiation of palliative care to death was 18.9 days (interquartile range [IQR], 0.09 d), substantially shorter for noncancer diseases (6 vs. 15 d for cancer) (47). Indeed, in our study, palliative care was provided a median of 7.0 days (IQR, 4–16 d) before death commonly in response to high physical- and psychologic-symptom distress, primarily among patients and sometimes among family members, as well as heavy caregiving burden while the patient was still in ICU. Furthermore, bereavement care (e.g., sending a sympathy card, calling families, providing a support group) is rarely provided in Taiwan to support bereaved surrogates’ psychologic needs under hospice care relative to worldwide practices, but bereavement care is still globally recognized as limited and poorly resourced. Given that surrogates with potentially higher psychologic distress and heavier caregiving burden were referred late to palliative care without adequate bereavement care to meet their post-loss psychologic needs, their likelihood of being in the PGD-PTSD-depression concurrent state might increase. Our finding warrants cross-national validation.

Poor-to-uncertain or worst QODD classes generally increased membership in the three more profound psychologic-distress states, reaching a statistically significant level for the subthreshold depression-dominant and the PGD-PTSD-depression concurrent states. ICU patients in the poor-to-uncertain and worst QODD latent classes were perceived to suffer pain and dyspnea most frequently, had moderate/insufficient/uncertain emotional preparedness for death, less religious and family support, and received more LSTs than those in the high QODD latent class (Table E1, http://links.lww.com/CCM/H585). Painful memories of unrelieved pain/dyspnea, uncertain or insufficient patient emotional preparedness for their own death, and patient suffering from frightening LSTs may preclude bereaved surrogate acceptance and sense-making of the patient’s death, leading to PGD symptoms like yearning, longing for, and/or a persistent preoccupation with thoughts and memories of the deceased. These memories of the death circumstances can also be distressing, intrusive, and challenging to manage, thereby subsequently eliciting excessive cognitive or behavioral avoidance of loss reminders to increase surrogates’ vulnerability to PTSD symptoms (48). Those who feel excessive guilt and self-blame (especially after EOL care decisions like initiating LSTs) may progress to endorse negative beliefs about themselves, the world, and the future, leading to profound depressive symptoms (48).

Several limitations should be acknowledged. Our study should be replicated with other bereaved family samples internationally to support generalizability given the worldwide cultural differences in manifestation of emotional symptoms during bereavement (12, 30). Our results cannot be generalized to unnatural causes of loss. We used symptom severity scales (PG-11, IES-R, and HADS), not structured diagnostic interviews, to assess symptom severity of PGD, PTSD, depression, and anxiety. In this observational study, a causal relationship cannot be inferred between factors and PGD-PTSD-depressive-symptom states nor can we exclude the potential impact of unmeasured covariates, especially considering that predictors from the appraisal and coping domain were excluded in this study. Future research should explore the role of surrogates’ appraisal of the loss (19, 21, 25) and their coping strategies (25, 26).

CONCLUSIONS AND CLINICAL IMPLICATIONS

Membership in more profound PGD-PTSD-depressive-symptom states for bereaved surrogates of ICU decedents was associated with both unmodifiable pre-traumatic intrapersonal risk factors and bereavement-related stressors as well as more powerful (49) but modifiable peri-/posttraumatic interpersonal risk factors and death circumstances. Special efforts should target vulnerable family surrogates who report financial hardship or have low educational attainment, preexisting mental illness requiring mood medications or ED visits, or high anxiety symptoms during bereavement. Focused EOL ICU care tailored to modifiable factors is urgently warranted to prevent subclinical depression, PGD, and concurrent PGD-PTSD-depressive symptoms among bereaved ICU family surrogates. Specifically, healthcare professionals should provide high-quality EOL ICU care by improving symptom management, facilitating patients’ emotional preparedness for their death, leveraging family support for the dying patient and their surrogate, avoiding potentially inappropriate LSTs, and facilitating a DNR order, social worker involvement in EOL care, and early referral to palliative care to actualize full benefits. Doing so may facilitate a good QODD and minimal concurrent PGD, PTSD, and depressive symptoms during bereavement.

ACKNOWLEDGMENTS

We thank Erica Light (Language Editor).

Supplementary Material

ccm-52-1885-s001.docx ^{(98KB, docx)}