The Association between Patient Health Status and Surrogate Decision Maker Post-Traumatic Stress Disorder Symptoms in Chronic Critical Illness

Abstract

Rationale: Surrogate decision-makers of patients with chronic critical illness (CCI) are at high risk for symptoms of post-traumatic stress disorder (PTSD). Whether patient health status after hospital discharge is a risk factor for surrogate PTSD symptoms is not known.

Objectives: To determine the association between patient health status 90 days after the onset of CCI and surrogate symptoms of PTSD.

Methods: We performed a secondary analysis of the data from a multicenter randomized trial of a communication intervention for adult patients with CCI and their surrogate decision-makers.

Results: Surrogate PTSD symptoms were measured at 90 days using the Impact of Events Scale–Revised. For patients who were alive at 90 days, location was used as a marker of health status and included the following categories: 1) home (relatively good health and low acuity), 2) acute rehabilitation (moderate care needs and impairments, generally expected to improve), 3) skilled nursing facility (moderate care needs and impairments, generally not expected to improve significantly or quickly), 4) long-term acute care facility (persistently high acute care needs and functional impairment), and 5) readmission to an acute care hospital (suggesting the highest acuity of illness and care needs of the cohort). Patients who died before 90 days were categorized as deceased. In the analyses, 365 surrogates and 256 patients were included. Among patients, 49% were female, and the mean age was 59 years. Among surrogates, 71% were female, and the mean age was 51 years. A directed acyclic graph was constructed to identify covariates to be included in the model. Compared with symptoms seen among surrogates of patients living at home, heightened PTSD symptoms were seen among surrogates of patients who were readmitted to an acute care hospital (β coefficient, 15.9; 95% confidence interval [CI], 4.5 to 27.3) or had died (β coefficient, 14.8; 95% CI, 8.8 to 20.9) at 90 days.

Conclusions: Surrogates of patients with CCI who have died or have been readmitted to an acute care hospital at 90 days experience increased PTSD symptoms as compared with surrogates of patients who are living at home. These patients and surrogates represent a readily identifiable group who may benefit from enhanced emotional support.

Chronic critical illness (CCI) results from persistent multisystem organ dysfunction after resolution of acute critical illness (1). One-year mortality ranges from 50% to 60%, and survivors are often left with profound physical and cognitive dysfunction (2). One year after intensive care unit (ICU) discharge, fewer than 10% of survivors are functionally independent (3). Most CCI survivors live with chronic serious illness and require extensive healthcare services and supports (3). During 1-year follow-up, patients with CCI may spend up to 74% of all days alive in a hospital, in a postacute facility, or receiving home health care (4).

Patients with CCI often depend on family members to act as caregivers and surrogate decision-makers. Although informal caregiving is essential and valued, CCI surrogates also experience negative effects on physical and mental health as well as financial toxicity (5–7). Psychological distress is common, with approximately one-third of CCI surrogates displaying clinically significant symptoms of post-traumatic stress disorder (PTSD) 90 days after a loved one’s ICU admission (8). Prior work has shown that surrogates of ICU patients who die and patients who survive are at risk for significant PTSD symptoms (9, 10). Caregiving extends beyond the ICU stay, yet no prior research has examined the impact of patient postacute health status on surrogate PTSD symptoms. Our objective was to measure the association between patient health status 90 days after hospital discharge and surrogate PTSD symptoms.

Methods

We performed a secondary analysis of the data from a multicenter, randomized controlled trial designed to determine whether a structured communication intervention improved emotional outcomes for surrogates of patients with CCI (11) (clinicaltrials.gov identifier NCT 01230099). Trial participation was offered to adult patients in medical ICUs at three tertiary care centers and one community hospital in the United States. Patients were included who had received at least 7 days of mechanical ventilation and were expected neither to be liberated from the ventilator nor die in the subsequent 72 hours. These criteria, which typically are considered in decision-making about tracheotomy and other options, represent a practical and widely accepted point of demarcation between acute critical illness and CCI. Study participation was offered to two categories of surrogate decision-makers: the primary surrogate with responsibility for healthcare decision-making if the patient lacked capacity and additional surrogate decision-makers if they also participated in healthcare decision-making. Patients who had chronic neuromuscular disease, trauma, or burns were excluded because they have a different clinical course and prognosis (2). All patients and surrogates from both the intervention group and control group in the trial with complete 90-day follow-up data were pooled for this secondary analysis. This work was approved by the University of North Carolina Institutional Review Board (#10-1692). The work for this study was performed at The University of North Carolina at Chapel Hill.

Primary Outcome Measure

The presence and severity of surrogate PTSD symptoms at 90-day follow-up were defined using the Impact of Events Scale–Revised (IES-R), a validated instrument that has been used to evaluate PTSD symptoms in families of ICU survivors and nonsurvivors (12, 13). The IES-R includes 22 items in three subscales: thought intrusion, avoidance, and hyperarousal. Item responses for how distressing each difficulty has been over the past 7 days range from 0 (not at all) to 4 (extremely). Summed subscale scores for the IES-R range from 0 to 88, with higher scores indicating worse symptoms, and a minimal clinically important difference of approximately 4.0 (14). All available surrogates were asked to complete the IES-R at 90 days after randomization.

Predictor Variable

To provide a comprehensive evaluation of patient health status as a risk factor for surrogate PTSD, we assessed patient health status at 90 days using a six-level categorical variable ranging from death to residing at home (indicating relatively low care needs). We grouped surviving patients into categories of patient health status on the basis of the Centers for Medicare and Medicaid Services level-of-care determinations, which take into account the severity of illness, care needs, and degree of functional impairment (15):

• 1.Home: Living at home is consistent with being in relatively good health and having lower care needs and less functional impairment.
• 2.Acute rehabilitation facility: Being in such a facility signifies being in fair health and having moderate care needs and moderate functional impairments that were generally expected to improve.
• 3.Skilled nursing facility or nursing home: Being in such a facility signifies being in fair health and having moderate care needs and moderate functional impairment that were generally not expected to improve significantly or quickly.
• 4.Long-term acute care facility: Being in such a facility signifies being in poor health and having persistently high acute care needs and high functional impairment.
• 5.Short-term acute care hospital: All patients in this category had been readmitted to an acute care hospital after previous discharge. This signified being in very poor and potentially worsening health, with these patients having the highest acuity of illness and care needs of the cohort.
• 6.Death: Patients who died between trial enrollment and the 90-day assessment were categorized as deceased.

Analyses

Descriptive statistics were analyzed using the mean and 95% confidence interval (CI) for continuous variables and the frequency and percentage for categorical variables. A multiple linear regression model was performed using the patient health status at 90 days as the exposure and the surrogate IES-R score as the outcome. A directed acyclic graph (DAG) was constructed to identify potential confounders and minimize the impact of confounding on the measured association between patient health status and surrogate PTSD symptoms (16, 17). Use of a DAG allows for a hypothesis-driven approach to defining confounders and is recommended over using traditional P value–based selection methods for causal inference in critical care research studies (18). Potential confounders were selected by investigators (B.W., S.S.C.) a priori on the basis of prior literature and a plausible mechanistic role in the development of PTSD symptoms. The model was adjusted for multiple respondents, using a mixed model, considering the patients as a random factor. All tests were two-sided and had a significance level of 0.05. Analysis was performed using SAS 9.4 (SAS Institute, Inc.).

Potential patient-related confounding factors included age, insurance status, prehospital functional status, presence of select chronic comorbidities, level of responsiveness measured at the time of trial enrollment, severity of illness at the time of trial enrollment, presence of acute kidney injury requiring new renal replacement therapy during ICU admission, length of hospital stay, and tracheostomy. Potential surrogate-related confounding factors were age, employment status, relationship to the patient, and psychological distress measured at the time of trial enrollment. Functional status before admission was scored as a summary of the total number of activities of daily living the patient was able to perform as reported by their primary surrogate (19, 20). Severity of illness at trial enrollment was defined by generating an estimate of 1-year mortality using the ProVent 14 score (21). Select chronic comorbidities used in the analysis included chronic cerebrovascular disease or hemiplegia, liver disease, a history of end-stage renal disease, or cancer. On the basis of a prior analysis of these data showing that surrogate anxiety and depression (measured by the Hospital Anxiety and Depression Scale [HADS]) and patient responsiveness (measured by the Richmond Agitation and Sedation Scale [RASS]) on or near Day 10 of mechanical ventilation are risk factors for surrogate PTSD symptoms (22), these variables were also included in the model. The RASS score (23) was measured for each patient by a trained research assistant to assess the patient’s level of alertness at the time of enrollment in the clinical trial. For analyses, the RASS score was categorized as −5 or −4 (unresponsive), −3 to −1 (arousable), or ⩾0 (awake). Surrogate symptoms of anxiety and depression were assessed at the time of enrollment using the HADS (24, 25). The HADS is a validated instrument designed to measure symptoms of anxiety and depression in hospitalized patients and has been used for family participants in the ICU setting (9). The HADS consists of two subscales (anxiety and depression), each containing seven items with a score ranging from 0 (lowest degree of symptoms) to 3 (highest degree of symptoms). A score ⩾11 on either subscale suggests the presence of anxiety and/or depression disorder(s); scores from 8 to 10 may represent “borderline” symptoms (22). Additional information on available variables can be found in descriptions of the original clinical trial (11).

Results

A total of 256 patients with CCI and their 365 surrogates were enrolled in the original clinical trial; the cohort for this analysis consisted of 224 patients with 306 surrogates with complete data at 90 days. Baseline variables were measured on Day 10 ± 3.4 of mechanical ventilation. A total of 218 (71%) surrogates were female, and the mean surrogate age was 51 years old. The majority of surrogates were either the spouse or partner or adult child of the patient. The mean baseline HADS score was 16.0 (95% CI, 15.1–16.9). The mean IES-R score at 90 days was 23.3 (95% CI, 21.3–25.4). A total of 91 (30%) of 306 surrogates had an IES-R score ⩾33, consistent with a probable diagnosis of PTSD. Complete surrogate characteristics are shown in Table 1.

Table 1.

Characteristics of surrogate decision-makers

Characteristic	Surrogates (n = 306)
Age, mean (95% CI), yr	51 (49.3–52.4)
Female sex, n (%)	218 (71)
Ethnicity, n (%)
Hispanic or Latino	38 (12)
Not Hispanic or Latino	267 (88)
Race, n (%)
Black	72 (24)
White	193 (63)
Other	41 (13)
Marital status, n (%)
Married or live with partner	207 (68)
Separated or divorced	38 (12)
Single or widowed	61 (20)
Education, n (%)
Advanced degree	53 (17)
College graduate	82 (27)
Some college	84 (28)
High school or less	86 (28)
Employment, n (%)
Disabled from employment	30 (10)
Employed or student	168 (55)
Homemaker	22 (7)
Retired	58 (19)
Unemployed	27 (9)
Religion, n (%)
Catholic	56 (18)
Jewish	17 (6)
Protestant	200 (66)
Other	16 (5)
None	16 (5)
Relationship, n (%)
Child	109 (36)
Parent	41 (13)
Sibling	37 (12)
Spouse or partner	100 (33)
Other	19 (6)
Number of decision-makers per patient, n (%)
1	152 (50)
2 or more	154 (50)
HADS score at baseline, mean (95% CI)
Total	16 (15.1–16.9)
Anxiety subscale	10 (9–10.1)
Depression subscale	7 (6–7)
Primary surrogate income, n (%)
<$15,000	31 (14)
$15,000–$39,999	41 (19)
$40,000–$100,000	71 (33)
>$100,000	25 (12)
Prefer not to answer	49 (23)

Definition of abbreviations: CI = confidence interval; HADS = Hospital Anxiety and Depression Scale.

Patients had a mean age of 59 years, and about half (n = 109, 49%) were female. The mean score for activities of daily living in the 2 weeks before hospital admission was 5 (95% CI, 4.6–5.2) out of a possible score of 6, indicating high functionality. The mean total hospital length of stay was 47.7 days. Over half of the patients (54%) received a tracheostomy. At the time of the 90-day assessment, 43% had died, 39% were living at home, 6% were in a skilled nursing facility or nursing home, 5% were readmitted to an acute care hospital, 4% were in a long-term acute care hospital, and 3% were in acute rehabilitation. Complete patient characteristics are shown in Table 2.

Table 2.

Characteristics of patients

Characteristic	Patients (n = 224)
Age, mean (95% CI), yr	59 (56.4–60.7)
Female sex, n (%)	109 (49)
Race, n (%)
Black	53 (24)
White	138 (63)
Unavailable	17 (8)
Other	10 (5)
Insurance, n (%)
Medicare	107 (48)
Medicaid	20 (9)
Commercial	76 (34)
None	21 (9)
Language, n (%)
English	197 (90)
Spanish	12 (6)
Other	9 (4)
Score for activities of daily living, mean (95% CI)*	5 (4.6–5.2)
Instrumental score for activities of daily living, mean (95% CI)†	16 (14.8–16.9)
History of liver disease, n (%)	29 (13)
History of cancer, n (%)	53 (24)
History of stroke, n (%)	20 (9)
History of end-stage renal disease, n (%)	9 (4)
Presence of advance directive at enrollment, n (%)	31 (14)
One-year mortality as predicted by ProVent 14 score, mean (95% CI), %	62 (58.6–64.8)
RASS score at enrollment, n (%)
−5 or −4 (unresponsive)	95 (44)
−3 to −1 (arousable)	81 (38)
0 and above (awake)	38 (18)
Acute renal failure requiring renal replacement therapy, n (%)	48 (19)
Tracheostomy, n (%)	139 (54)
Duration of mechanical ventilation after randomization, mean (95% CI), d	20.2 (15.1–25.2)
Hospital length of stay, mean (95% CI), d	47.7 (42.1–53.3)
Health status by level of care at 90 d, n (%)
Died	94 (43)
Readmitted to an acute care hospital	10 (5)
Long-term acute care hospital	9 (4)
Skilled nursing facility	14 (6)
Acute rehabilitation facility	6 (3)
Living at home	86 (39)

Definition of abbreviations: CI = confidence interval; RASS = Richmond Agitation Sedation Scale.

^*The range is 0 (dependent) to 6 (independent) in 6 activities.

^† The range is 8 (dependent) to 31 (independent) in 8 activities.

Patient health status at 90 days was then assessed for association with the surrogate IES-R score. The DAG created to identify the minimally sufficient set of variables for the multivariate model is shown in Figure 1. All variables included in the final model are shown in Table 3, together with the associated mean change in the IES-R score as represented by a β coefficient and 95% CI. After adjusting for potential confounders (patient and surrogate demographics, patient severity of illness at trial enrollment, and baseline surrogate psychological distress), surrogates of patients currently readmitted to an acute care hospital had greater PTSD symptoms than surrogates of patients who were living at home (β coefficient, 15.9; 95% CI, 4.5–27.3). A similar comparatively higher burden of PTSD symptoms was observed among surrogates of patients who had died by 90 days compared with surrogates of patients living at home (β coefficient, 14.8; 95% CI, 8.8–20.9). The adjusted mean IES-R score was 28.9 (95% CI, 16.8–41.0) for surrogates of patients who were readmitted to an acute care hospital, 27.8 (95% CI, 20.3–35.4) for surrogates of patients who had died, and 13.0 (95% CI, 5.6–20.4) for surrogates of patients who were living at home. To assess the impact of the communication intervention from the original trial on surrogate PTSD symptoms, we ran an additional model that also contained randomization arm as a covariate, which did not significantly change our results (see the online supplement).

Figure 1.

Directed acyclic graph. AKI = acute kidney injury; ESRD = end stage renal disease; HADS = Hospital Anxiety and Depression Scale; PTSD = post-traumatic stress disorder; RASS = Richmond Agitation and Sedation Scale.

Table 3.

Results of multivariate model

Variable Name	Coefficient (95% CI)*	P Value
Patient-related variables
Age	−0.09 (−0.48 to 0.29)	0.63
Insurance status
Medicare	−10.5 (−27.4 to 6.31)	0.41
Medicaid	−2.08 (−20.6 to 16.4)
Commercial	−7.46 (−24.4 to 9.50)
None	−6.30 (−24.7 to 12.1)
Other	—
Score for activities of daily living score,  mean (95% CI)†	−0.82 (−2.21 to 0.58)	0.25
History of cancer	−6.25 (−12.0 to -0.53)	0.03
History of liver disease	−2.48 (−10.1 to 5.09)	0.51
ProVent 14 score	0.02 (−0.14 to 0.17)	0.85
RASS score
−5 or −4 (unresponsive)	4.92 (−2.89 to 12.7)	0.38
−3 to −1 (arousable)	2.37 (−5.56 to 10.3)
0 and above (awake)	—
Hospital length of stay	0.03 (−0.01 to 0.07)	0.15
Acute renal failure requiring renal  replacement therapy	−3.22 (−9.77 to 3.33)	0.33
Tracheostomy performed	−1.98 (−7.44 to 3.48)	0.47
Health status by level of care at 90 d
Died	14.8 (8.8 to 20.9)	0.0002
Readmitted to an acute care hospital	15.9 (4.5 to 27.2)
Long-term acute care hospital	3.4 (−9.5 to 16.2)
Skilled nursing facility or nursing home	3.4 (−6.6 to 13.4)
Acute rehabilitation facility	13.9 (−1.6 to 29.3)
Living at home	—
Surrogate-related variables
Age	−0.18 (−0.54 to 0.18)	0.32
Employment
Disabled from employment	2.20 (−7.74 to 12.1)	0.25
Employed	5.25 (−2.19 to 12.7)
Homemaker	−4.61 (−16.0 to 6.79)
Retired	5.60 (−3.93 to 15.1)
Student	8.12 (−9.07 to 25.3)
Unemployed (but not disabled)	—
Relationship to patient
Child	0.56 (−13.0 to 14.1)	0.66
Parent	3.49 (−11.6 to 18.6)
Sibling	1.98 (−10.0 to 14.0)
Spouse or partner	5.36 (−5.55 to 16.3)
Other	—
HADS score	0.94 (0.67 to 1.21)	<0.0001

Definition of abbreviations: CI = confidence interval; HADS = Hospital Anxiety and Depression Scale; IES-R = Impact of Events Scale–Revised; RASS = Richmond Agitation Sedation Scale.

All models were adjusted for multiple respondents.

^*For continuous variables, the coefficient represents the mean change in IES-R score per 1-unit increase. For categorical variables, the coefficient represents mean difference in IES-R score as compared with the reference group.

^† The range is 0 (dependent) to 6 (independent) in 6 activities.

Discussion

In this secondary analysis of the data from a randomized trial of a communication intervention for surrogates of patients with CCI, we found that patient death or current readmission to an acute care hospital are associated with increased PTSD symptoms in surrogates. Other less severe health conditions did not increase PTSD symptom risk in this population.

This is the first study to demonstrate an association between worse patient health status after hospital discharge and greater surrogate PTSD symptoms in CCI. Prior work has shown an association between poor patient health and increased surrogate depression, but this work was limited by smaller sample sizes (26–29). Douglas and colleagues (30) performed a secondary analysis of 270 CCI surrogates and found that hospital readmission and patient location in an institution were risk factors for increased depression symptoms 2 months after hospital discharge but did not examine differences in surrogate distress by specific postdischarge health status. Our work provides new evidence for differences in surrogate experience based on patient health status. Because patients with CCI experience a broad range of outcomes after hospital discharge (4), analysis of surrogate PTSD symptoms by health status along this full spectrum allows identification of surrogates at highest risk for adverse outcomes (4). Our results could facilitate targeting of enhanced practical, emotional, or financial support for those surrogates at highest risk for PTSD symptoms.

To date, interventions to reduce distress for surrogates of critically ill patients have been largely unsuccessful (11, 13, 31, 32). Interventions have been deployed either while the patient was still in the ICU or in the immediate postdischarge period, with follow-up assessments of surrogate outcomes occurring 3–6 months later. One possible reason that these interventions were not successful is that there are stressors that emerge later in the course of critical illness, such as patient health status at 90 days, that are difficult to predict. Enhanced surrogate support across the care continuum has recently been identified as a key component of care for ICU survivors, and surrogates of patients who die or who remain unstable and experience acute care readmissions may be most in need of longitudinal support (33–35).

However, before supportive interventions are designed, further research is also needed to understand the caregiving experience in CCI. Our finding that surrogates of patients who are readmitted to an acute care hospital are more distressed than surrogates of patients who are at home is interesting, given that the daily burden of physical caregiving is lower when patients are hospitalized. There are several possible reasons that patient health status at 90 days is associated with increased surrogate PTSD. Prior investigations have shown that hope is a protective factor against surrogate PTSD, and it is possible that when patients are so ill that they return to acute care or die, surrogates experience trauma from loss of hope for patient recovery (36). Surrogates may experience stress due to persistent caregiving demands, multiple transitions and hand-offs, and lack of certainty about what to expect, and all of these may contribute to PTSD (37, 38). Recurrent hospitalization is associated with higher out-of-pocket costs, and it is possible that financial stress is also contributing to PTSD symptoms (4, 7, 39). Rehospitalization may also recapitulate any trauma symptoms associated with the first hospitalization. Patients with complex CCI are often managed in larger regional medical centers, and so longer distance from a loved one may worsen surrogate distress (40). Further work is needed to understand which aspects of the patient being in an acute care hospital contribute to heightened PTSD symptoms so that targeted interventions can be developed.

Limitations

Our study has several limitations. First, although we used preexisting definitions of levels of care to derive health status, we did not directly measure it. Second, we do not know the reason for patient hospital readmission by 90 days, which limits our understanding of how this contributes to caregiver distress. Future work should contain detailed descriptions of health status after hospital discharge, including a direct measure of functional status and assessment of care dependency as well as characteristics of hospital readmission. In addition, because PTSD symptoms were assessed just once at 90 days, it is not known whether high degree of surrogate PTSD symptoms represent a change in distress from baseline. However, we attempted to control for high baseline psychological distress by including in-hospital anxiety and depression scores in our multivariable model. We did not measure potentially important mediators of the association between patient health status and surrogate distress, such as surrogate lifestyle restriction, surrogate fatigue, or the nature of patient care needs (26, 27). Future work in this area should focus on determining which specific aspects of the patient being in an acute care hospital contributed to heightened surrogate PTSD symptoms.

Conclusions

In conclusion, surrogates of patients with persistently poor health requiring acute hospital readmission or resulting in death after the onset of CCI experience worse PTSD symptoms at 90 days than surrogates of patients who are healthy enough to remain at home. These results provide novel detail about the importance of health status to the patient–surrogate dyad’s well-being and highlight a potential target for improving the experience of one of the most vulnerable groups of patients and surrogates in the United States.