Psychiatric symptoms after acute respiratory distress syndrome: a 5-year longitudinal study

Abstract

Purpose

We aimed to characterize anxiety, depression, and posttraumatic stress disorder (PTSD) symptoms over 5-year follow-up after acute respiratory distress syndrome (ARDS) and determine risk factors for prolonged psychiatric morbidity.

Methods

This prospective cohort study enrolled patients from 13 medical and surgical intensive care units in four hospitals, with follow-up at 3, 6, 12, 24, 36, 48, and 60 months post-ARDS. Trained research staff administered the Hospital Anxiety and Depression Scale (HADS) (scores ≥8 on anxiety and depression subscales indicating substantial symptoms) and the Impact of Event Scale-Revised (IES-R, scores ≥1.6 indicating substantial PTSD symptoms) at each follow-up visit.

Results

Of 196 consenting survivors, 186 (95%) ever completed HADS and IES-R assessments; 96 (52%) had any continuous or recurring (prolonged) symptoms, and 71 (38%), 59 (32%), and 43 (23%) had prolonged anxiety, depression, and PTSD symptoms, respectively (median total durations 33–39 months, 71–100% of observed follow-up time). Prolonged psychiatric symptoms tended to co-occur across domains; the most common morbidity pattern involved substantial symptoms in all three domains. Worse pre-ARDS mental health, including prior depression and psychological distress in the period immediately preceding ARDS, was strongly associated with prolonged post-ARDS psychiatric morbidity, across symptom domains.

Conclusions

Clinically significant and long-lasting symptoms of anxiety, depression, and PTSD are common in the first 5 years after ARDS. In-hospital screening of psychiatric history, including recent anxiety and depression symptoms, may be useful for long-term mental health treatment planning after ARDS.

Introduction

Survivors of acute respiratory distress syndrome (ARDS) and other critical illnesses frequently have substantial anxiety, depression, and posttraumatic stress disorder (PTSD) symptoms [1-7]. However, based on recent systematic reviews of studies of critical illness survivors, we know little about the long-term course of psychiatric symptoms after critical illness, including risk factors for a prolonged course that may be ascertainable during the index hospitalization, given that most previous studies have followed patients ≤1 year [4-6]. Hence, we specifically investigated these issues in this report of a 5-year prospective longitudinal study of ARDS survivors.

Materials and Methods

Study population

In the Improving Care of Acute lung injury Patients (ICAP) prospective longitudinal cohort study, we enrolled consecutive mechanically ventilated patients with acute lung injury (ALI), as defined by American-European Consensus criteria [8] in effect at the time of study recruitment (2004-2007) (ClinicalTrials.gov Identifier: NCT00300248). To be consistent with the more recent Berlin criteria [9], we use the term ARDS, rather than ALI, throughout this manuscript. The ICAP study recruited patients from 13 medical and surgical intensive care units (ICUs) at four hospitals in Baltimore, Maryland, USA. To avoid inclusion of patients with primary neurologic disease or head trauma, we did not screen for eligible patients in neurologic specialty ICUs. Key exclusion criteria were 1) pre-existing illness with a life expectancy of <6 months; 2) pre-existing cognitive impairment (ascertained via medical records and/or proxy interview) or communication/language barriers; 3) no fixed address; 4) transfer to a study site ICU with pre-existing ARDS of >24 hours' duration; 5) >5 days of mechanical ventilation before ARDS; and 6) a physician order for no escalation of ICU care (e.g., no vasopressors or hemodialysis) at the time of study eligibility.

We obtained written informed consent after patients regained capacity, typically around the time of hospital discharge. Trained research assistants conducted follow-up assessments at 3, 6, 12, 24, 36, 48, and 60 months after ARDS onset. Institutional review boards at all participating study sites approved this study.

Definition of general anxiety, depression, and PTSD symptoms

At each follow-up, we assessed anxiety and depression symptoms using the Hospital Anxiety and Depression Scale (HADS) [10]. HADS anxiety and depression subscales both include seven items, each rated on a 4-point scale; we considered patients with a score ≥8, on either subscale, to have substantial anxiety or depression symptoms. This threshold is recommended and very commonly used with the HADS [4, 5]; it is associated with sensitivities and specificities of ∼0.8 for the detection of clinically important anxiety or depression syndromes in general medical patients [11]. We chose the HADS because of its utility in detecting psychiatric symptoms in patients with general medical problems. For example, the HADS depression items primarily assess anhedonia, rather than neurovegetative symptoms (e.g., diminished energy) that could be symptomatic of general medical conditions (e.g., muscular weakness) rather than depressive states.

We assessed PTSD symptoms using the Impact of Event Scale-Revised (IES-R) [12]. The IES-R consists of 22 items, each rated on a 5-point scale; we averaged item scores to generate a mean total score (range: 0–4). In a previous study, we examined the measurement properties of the IES-R vs. a semi-structured PTSD clinical interview; at a threshold of ≥1.6, the sensitivity for detecting PTSD was 100%, and the specificity was 85% [13]. Thus, we chose the ≥1.6 threshold to define substantial PTSD symptoms in the current study.

Definitions of remission and recurrence

We defined remission from anxiety or depression symptoms as having HADS anxiety or depression scores <8 at any follow-up time point after substantial symptoms were evident, along with a statistically reliable decrease in score using the Reliable Change Index (RCI) [14]. As in prior research, our RCI threshold was ≥4 [15], larger than the minimal important difference (2–2.5) for these subscales in acute respiratory failure survivors [16]. We defined recurrence as a HADS anxiety or depression score ≥8 at any follow-up assessment after remission, along with an increase in score of ≥4.

We defined remission from PTSD symptoms as having an IES-R score ≤1.6 at any follow-up time point after substantial symptoms were evident, along with a statistically reliable decrease in score using the RCI. As in prior research, our RCI threshold was ≥0.5 [15], larger than the minimal important difference (0.2) for this scale in acute respiratory failure survivors [16]. We defined recurrence as having an IES-R score ≥1.6 at any follow-up assessment after remission, along with an increase in score of ≥0.5.

When survivors had missing HADS anxiety, HADS depression, or IES-R scores during follow-up, we assumed that their prior status remained unchanged, as in prior work [15, 17, 18].

Definition of prolonged symptoms for analyses of potential risk factors

To examine risk factors for prolonged psychiatric symptoms in each domain of psychiatric morbidity, we categorized patients into one of two mutually exclusive groups: a prolonged symptoms group (symptoms were either continuous after onset or recurrent) and a comparison group (symptoms were absent throughout 5-year follow-up or remitting without recurrence).

Potential risk factors for prolonged psychiatric morbidity

We evaluated several candidate baseline (pre-ARDS) variables that could be associated with prolonged psychiatric symptoms after ARDS, based on systematic reviews of the literature [4-6]:

1. Demographics: age, sex, race, education, employment

2. Mental health: history of depression, cigarette smoking, heavy alcohol use, or illicit drug use – abstracted from patients' medical records from the index (ARDS-related) hospitalization; and retrospective reports of psychological distress in the period immediately preceding the index illness – measured using the mental health domain of the SF-36 [19], as previously validated in acute respiratory failure survivors [20, 21]

3. Physical functioning: retrospective reports of physical functioning in the period immediately preceding the index illness – measured using the physical function domain of the SF-36 [19]

We also evaluated potential critical illness- and intensive care-related risk factors. These included severity of illness (Acute Physiology and Chronic Health Evaluation II score) [22], ICU length of stay, maximum daily Sequential Organ Failure Assessment score [23], the proportion of ICU days patients were septic [24], the proportion of ICU days patients were delirious (assessed by trained research staff using the Confusion Assessment Method for the ICU) [25], and mean daily benzodiazepine, opiate, and systemic corticosteroid doses (presented as midazolam, morphine, and prednisone equivalents, respectively [26-28]).

Statistical methods

We used a Venn diagram and bivariable logistic regression to characterize cooccurrence of prolonged symptoms of anxiety, depression, and PTSD. To simplify comparisons across potential risk factors for prolonged psychiatric morbidity, we converted semi-continuous variables into binary variables using logical thresholds, thresholds near median values, or in the case of SF-36 variables, one standard deviation below matched population values. We examined each potential risk factor in unadjusted logistic regression models and, to avoid model overfitting, we selected variables with unadjusted p≤0.05 for inclusion in multivariable models. In multivariable models, we first adjusted for associated demographic factors, and then we additionally adjusted for independently associated baseline health variables. We conducted analyses using SAS® version 9.4 (2013, Cary, NC).

Results

Cohort description

Of the 520 patients enrolled in our study (74% of whom had sepsis as the primary ARDS risk factor [29]), 274 (53%) died before the initial 3-month follow-up assessment, and 50 (9.6%) were not available for informed consent or declined participation (Figure 1). Risk factors for mortality included older age and greater organ dysfunction [29]. Of the 196 consenting 3-month survivors, 186 (95%) completed at least one HADS and IES-R assessment during 5-year follow-up; 65 (33%) died during follow-up, and 7 (3.6%) were not consented to continue in Years 3 to 5 of the study (Figure 1). Of the 124 five-year survivors, 99 (80%) completed at least 6 of the 7 longitudinal assessments. Table 1 shows pre-ARDS demographic, health, and critical illness/intensive care characteristics of the 186 participants with mental health assessments.

Figure 1. Flow diagram of study participants.

^a Some patients had a follow-up visit but did not have complete Hospital Anxiety and Depression Scale (HADS) or Impact of Event Scale-Revised (IES-R) data for the following reasons, respectively, physically incapable/cognitively incapable/other:

At 3 months, 22 did not have complete HADS data (9/5/8), and 23 did not have complete IES-R data (7/5/11).

At 6 months, 12 did not have complete HADS data (4/2/6), and 13 did not have complete IES-R data (4/2/7).

At 12 months, 14 did not have complete HADS data (5/4/5), and 15 did not have complete IES-R data (5/4/6).

At 24 months, 10 did not have complete HADS data (0/6/4), and 11 did not have complete IES-R data (1/5/5).

At 36 months, 8 did not have complete HADS data (0/3/5), and 9 did not have complete IES-R data (0/3/6).

At 48 months, 7 did not have complete HADS or IES-R data (0/2/5). At 60 months, 7 did not have complete HADS or IES-R data (0/1/6).

^bThe study initially consented patients for 3, 6, 12 and 24-month follow-up. Upon receipt of a new grant for follow-up until 60 months, patients were required to provide new informed consent for the extended follow-up duration. The start of funding for this extended follow-up was delayed, resulting in 15 patients missing their 36-month follow-up

Table 1. Pre-ARDS baseline and critical illness/intensive care variables in survivors (n=186).

	Threshold	N (%)	Median (IQR) a
Demographics
Age, years	<50:	99 (53%)	49 (40-57)
Female sex		81 (44%)
Caucasian b		107 (58%)
Education, years b	≤12:	112 (63%)	12 (11-14)
Unemployed b		74 (40%)
Baseline health status
History of depression		39 (21%)
History of cigarette smoking b		139 (75%)
History of heavy alcohol use		46 (25%)
History of illicit drug use		59 (32%)
SF-36 mental health domain normalized score b,c	<40:	51 (34%)	50 (36-56)
SF-36 physical function domain normalized score b,c	<40:	52 (34%)	49 (32-57)
Critical illness/intensive care factors
APACHE II score	≥20:	131 (70%)	23 (19-28)
ICU length of stay, days	≥14:	105 (56%)	14 (10-23)
Maximum SOFA score	>10:	57 (31%)	9 (7-11)
Proportion of ICU days with sepsis	≥0.75:	91 (49%)	0.73 (0.11-0.96)
Proportion of ICU days with delirium	≥0.75:	75 (40%)	0.69 (0.49-0.88)
Average daily midazolam equivalents, mg	≥75:	55 (30%)	34 (9-85)
Average daily morphine equivalents, mg	≥100:	92 (50%)	97 (40-199)
Average daily prednisone equivalents, mg	<50:	158 (85%)	4 (0-31)

ARDS=acute respiratory distress syndrome; APACHE=Acute Physiology and Chronic Health Evaluation; ICU=intensive care unit; IQR=interquartile range; N=number of patients; SF-36=Medical Outcomes Short-Form 36-item instrument; SOFA=Sequential Organ Failure Assessment

^aFor continuous or semi-continuous variables

^bmissing data (race, n=1; education, 9; employment, 2; smoking, 1; SF-36, 35)

^cage- and sex-matched general population values: mean 50, standard deviation 10

Prevalence and course of psychiatric morbidity after ARDS

Of the 186 participants, 71 (38%), 59 (32%), and 43 (23%) had a prolonged (i.e., continuous or recurring) course of substantial anxiety, depression, and PTSD symptoms, respectively; 96 (52%) had any prolonged symptoms. Median total duration of symptoms varied from 33–39 months, representing 71–100% of the observed follow-up time. Of those with prolonged anxiety symptoms, 56 and 15 had continuous and recurring symptoms, respectively (median duration 39 months in each case). Of those with prolonged depression symptoms, 46 and 13 had continuous and recurring symptoms, respectively (median durations 36 and 27 months, respectively). Of those with prolonged PTSD symptoms, 24 and 19 had continuous and recurring symptoms, respectively (median durations 26 and 33 months, respectively). An additional 41, 39, and 30 participants had remitting symptoms of anxiety, depression, and PTSD, respectively (median durations 12 months in each case). The Supplemental Figure shows individual and mean trajectories for anxiety, depression, and PTSD symptoms in patients with and without prolonged symptoms.

Notably, prolonged symptoms of anxiety, depression, and PTSD tended to co-occur across domains; the most common pattern of prolonged psychiatric morbidity involved all three domains (Figure 2). Odds ratios (95% confidence intervals) relating pairs of prolonged symptom domains were as follows: anxiety and depression 7.4 (3.7–15), anxiety and PTSD 9.1 (4.1–20), and depression and PTSD 5.8 (2.8– 12) (all p<0.001).

Figure 2. Venn diagram illustrating co-occurrence of prolonged anxiety, depression, and PTSD symptoms.

Risk factors for prolonged psychiatric morbidity after ARDS

Demographic factors associated with prolonged anxiety symptoms after ARDS included female sex, Caucasian race, and education ≤12 years (Table 2). Education ≤12 years also was associated with prolonged depressive and PTSD symptoms. Another marker of lower socioeconomic status, baseline unemployment, was associated with prolonged PTSD symptoms. Odds ratios (ORs) for these demographic risk factors ranged from 2.0-3.0.

Table 2. Unadjusted associations between baseline and critical illness/intensive care factors and prolonged psychiatric morbidity during 5-year follow-up in ARDS survivors.

	Prolonged anxiety symptoms			Prolonged depression symptoms			Prolonged PTSD symptoms
	Yes	No		Yes	No		Yes	No
	(n=71)	(n=115)	OR (95% CI)	(n=59)	(n=127)	OR (95% CI)	(n=43)	(n=143)	OR (95% CI)
Baseline demographic factors
Age <50 years	55%	52%	1.1 (0.6-2.0)	46%	57%	0.6 (0.3-1.2)	65%	50%	1.9 (0.9-3.8)
Female sex	58%	35%	2.6 (1.4, 4.7)**	53%	39%	1.7 (0.9-3.2)	54%	41%	1.7 (0.8-3.3)
Caucasian	69%	51%	2.2 (1.2, 4.0)*	68%	53%	1.9 (1.0-3.5)	56%	59%	0.9 (0.5-1.8)
Education ≤12 years	73%	57%	2.0 (1.0-3.9)*	79%	56%	3.0 (1.4-6.2)**	76%	59%	2.2 (1.0-4.8)*
Unemployed	44%	38%	1.3 (0.7, 2.4)	48%	37%	1.6 (0.9-3.4)	58%	35%	2.6 (1.3-5.2)**
Baseline health status
History of depression	34%	13%	3.4 (1.6-7.1)**	34%	15%	2.9 (1.4-6.0)**	40%	15%	3.6 (1.7-7.7)***
History of cigarette smoking	80%	72%	1.5 (0.8-3.1)	81%	72%	1.6 (0.8-3.5)	88%	71%	3.0 (1.1-8.1)*
History of heavy alcohol use	30%	22%	1.5 (0.8-3.0)	20%	27%	0.7 (0.3-1.5)	28%	24%	1.2 (0.6-2.7)
History of illicit drug use	34%	30%	1.2 (0.6-2.2)	44%	26%	2.2 (1.2-4.3)*	49%	27%	2.6 (1.3-5.3)**
Normed SF-36 mental health domain score <40	54%	23%	4.0 (1.9-8.1)***	57%	24%	4.1 (2.0-8.6)***	61%	27%	4.4 (1.9-10)***
Normed SF-36 physical function domain score <40	46%	28%	2.2 (1.1-4.5)*	52%	27%	2.9 (1.4-6.1)**	32%	35%	0.9 (0.4-2.1)
Critical illness/intensive care factors
APACHE II score ≥20	65%	74%	0.7 (0.3-1.2)	63%	74%	0.6 (0.3-1.1)	63%	73%	0.6 (0.3-1.3)
Intensive care unit length of stay ≥14 days	55%	57%	0.9 (0.5-1.6)	59%	55%	1.2 (0.6-2.2)	63%	55%	1.4 (0.7-2.8)
Maximum SOFA score >10	27%	33%	0.7 (0.4-1.4)	29%	32%	0.9 (0.5-1.7)	40%	28%	1.7 (0.8-3.4)
Proportion of ICU days with confirmed sepsis ≥0.75	52%	47%	1.2 (0.7-2.2)	54%	47%	1.4 (0.7-2.5)	51%	48%	1.1 (0.6-2.2)
Proportion of ICU days with delirium ≥0.75	41%	40%	1.0 (0.6-1.9)	34%	43%	0.7 (0.4-1.3)	44%	39%	1.2 (0.6-2.5)
Average daily midazolam equivalents ≥75 mg	27%	31%	0.8 (0.4-1.5)	34%	28%	1.3 (0.7-2.6)	30%	29%	1.0 (0.5-2.2)
Average daily morphine equivalents ≥100 mg	45%	52%	0.8 (0.4-1.4)	51%	49%	1.1 (0.6-2.0)	44%	51%	0.8 (0.4-1.5)
Average daily prednisone equivalents <50 mg	90%	82%	2.0 (0.8-5.1)	90%	83%	1.9 (0.7-4.8)	93%	83%	2.8 (0.8-9.9)

ARDS=acute respiratory distress syndrome; APACHE=Acute Physiology and Chronic Health Evaluation; CI=confidence interval; ICU=intensive care unit; OR=odds ratio; SF-36=Medical Outcomes Short-Form 36-item instrument; SOFA=Sequential Organ Failure Assessment

Bold type highlights statistically significant differences between groups

^*p≤0.05,

^**p≤0.01,

^***p≤0.001

Having worse pre-ARDS mental health was strongly and consistently associated with prolonged psychiatric morbidity after ARDS (Table 2). For example, having a history of depression was associated with ORs of approximately 3 across all three outcome domains. Similarly, having an SF-36 mental health score >1 standard deviation below the matched population norm (i.e., <40) was associated with ORs ≥4.0 across all three outcome domains. Other potentially relevant factors include a history of illicit drug use, which was associated with ORs >2.0 for prolonged depression and PTSD symptoms, and a history of cigarette smoking, which was associated with an OR of 3.0 for prolonged PTSD symptoms.

Having an SF-36 physical function normed score <40 was associated with ORs ranging from 2.2-2.9 for prolonged anxiety and depression symptoms (Table 2).

In contrast to results for demographic and baseline health variables, no critical illness/intensive care variable was associated with prolonged psychiatric morbidity.

Results of regression models adjusted for associated demographic variables demonstrated that a history of depression prior to ARDS was independently associated with prolonged post-ARDS psychiatric morbidity (ORs: 2.4–2.9), as was having an SF-36 mental health normed score <40 (ORs: 3.5–4.2) (Table 3). Similarly, having an SF-36 physical function normed score <40 was associated with prolonged post-ARDS depression symptoms independently of associated demographic variables (OR 2.8) (Table 3).

Table 3. Adjusted associations between baseline health variables and prolonged psychiatric morbidity during 5-year follow-up in ARDS survivors^a.

	Prolonged anxiety symptoms AORb (95% CI)	Prolonged depression symptoms AORc (95% CI)	Prolonged PTSD symptoms AORd (95% CI)
History of depression	2.5 (1.1-5.4)*	2.4 (1.1-5.2)*	2.9 (1.3-6.5)**
History of cigarette smoking			2.1 (0.7-6.0)
History of illicit drug use		1.7 (0.9-3.5)	1.6 (0.7-3.5)
Normed SF-36 mental health domain score <40	3.5 (1.6-7.5)***	4.2 (1.9-9.0)***	3.7 (1.5-9.0)**
Normed SF-36 physical function domain score <40	1.9 (0.9-4.1)	2.8 (1.3-6.1)**

AOR=adjusted odds ratio; ARDS=acute respiratory distress syndrome; CI=confidence interval; SF-36=Medical Outcomes Short-Form 36-item instrument

^aEach cell in this table is a result from a separate regression model, adjusted for associated demographic variables (Table 2), specifically:

^bsex (except for SF-36 variables, which are sex-normed), race, and education;

^ceducation; and

^deducation and employment

Bold type highlights statistically significant differences between groups

^*p≤0.05,

^**p≤0.01,

^***p≤0.001

In a final set of regression models, we simultaneously adjusted for associated demographic variables and a history of depression prior to ARDS. These models demonstrated, once again, that an SF-36 mental health normed score <40 was independently associated with prolonged anxiety symptoms (OR=2.9, 95% confidence interval 1.3–6.5), prolonged depressive symptoms (OR=4.0, 95% CI 1.8–9.0), and prolonged PTSD symptoms (OR=2.8, 95% CI 1.7–7.2). Similarly, an SF-36 physical function normed score <40 was independently associated with prolonged post-ARDS depression symptoms (OR=2.7, 95% CI 1.3–5.9).

Discussion

In this multi-site, 5-year prospective longitudinal cohort study of 186 ARDS survivors, we found that prolonged symptoms of anxiety, depression, and PTSD were very common, affecting 38%, 32%, and 23% of survivors, respectively, with 52% of survivors having prolonged psychiatric morbidity in at least one of the three evaluated domains. Median total durations of symptoms were long, 33–39 months, representing 71–100% of observed follow-up time, and co-occurrence across symptom domains was substantial. Prior depression and psychological distress immediately preceding the ARDS-related hospitalization were associated with substantially increased risk for prolonged post-ARDS psychiatric morbidity, across all three psychiatric symptom domains.

The identified risk factors for prolonged psychiatric morbidity in these ARDS survivors are consistent with existing literature on psychiatric morbidity in the context of other stressors. Specifically, lower socioeconomic status and prior psychiatric morbidity are associated with anxiety and depressive disorders in contexts such as trauma and natural disasters [30-32]. In our study, pre-ARDS depression and psychological distress immediately preceding the ARDS-related illness had the strongest associations with prolonged psychiatric disturbances. Furthermore, psychological distress immediately preceding the ARDS-related illness was associated with prolonged post-ARDS psychiatric morbidity independently of a history of depression prior to ARDS. Thus, while determining whether a patient has a history of depression is helpful, inquiring about symptoms of anxiety and depression just before the critical illness provides additional prognostic information regarding long-term psychiatric morbidity. In addition, physical morbidity also increases risk for depressive disorders [33], consistent with our finding that worse baseline physical functioning was associated with prolonged depression symptoms. Finally, specific critical illness and intensive care characteristics were not associated with prolonged psychiatric morbidity. Putting our findings together, we hypothesize that the severe stresses of ARDS and intensive care precipitate prolonged psychiatric morbidity, especially in those most vulnerable to stress, and that the common critical illness and critical care factors (e.g., respiratory failure and mechanical ventilation) likely outweigh differences across illnesses and care practices. Our findings regarding risk factors for long-term psychiatric morbidity are analogous to findings regarding risk factors for poor functional recovery after critical illness (e.g., lower baseline body mass index or functional self-efficacy, or pre-ICU hearing or vision impairment [34]).

There are important clinical implications of these findings. Specifically, patients with prior depression-related diagnoses or symptoms of psychological distress deserve particular attention, to help recognize and address the risk for long-term psychiatric morbidity after ARDS. Relevant interventions may include early resumption of prior psychiatric treatments stopped during critical illness, such as antidepressant medications and cognitive-behavioral therapy, and assessment of effectiveness of prior treatments to determine if changes are indicated. ARDS survivors with lower socioeconomic status and/or impaired physical functioning also merit attention. We hope that universal preventive efforts, such as ICU diary programs and in-ICU psychological interventions, help reduce the prolonged suffering that so many survivors of critical illness endure [35, 36]. Though post-ICU physical rehabilitation efforts have not been effective in reducing psychiatric morbidity [37], it remains unclear whether in-ICU rehabilitation or later regular exercise-related interventions (e.g., as part of a behavioral activation program) might also be beneficial. Unfortunately, at the time our patients were critically ill, our ICUs did not have delirium prevention/management, sleep promotion, or diary protocols, so we could not address effects of these measures on long-term mental health.

Our findings also add to the literature on the validity of the SF-36 mental health domain. Most prior studies have examined convergent validity (e.g., cross-sectional associations between mental health scores and HADS anxiety scores [20, 21]). In the current study, we found that low pre-ARDS SF-36 mental health domain scores had predictive validity in anticipating prolonged anxiety, depression, and PTSD symptoms in survivors.

In the current study, we could not address whether the marked stresses of ARDS and intensive care caused prolonged anxiety, depression, or PTSD symptoms during follow-up. Notably, in a very large population-based longitudinal cohort study, having a psychiatric diagnosis at baseline was associated with a hazard ratio of 2–3 of hospitalization (with or without critical illness), but having a critical illness was markedly associated with new psychiatric diagnoses and medications, with hazard ratios >20 [38]. Thus, we consider critical illness to be an important risk factor for psychiatric morbidity.

There are additional potential limitations of this study. First, we measured anxiety, depression, and PTSD symptoms using validated and recommended questionnaires [11, 13, 39], rather than clinical diagnostic interviews. The latter are not feasible given the response burden – especially during early recovery after ARDS. Second, we used medical records and retrospective patient interviews to identify baseline health status, since directly assessing patients' history prior to the unexpected onset of ARDS is not feasible. Third, we did not account for the possible salutary effect of treatment on psychiatric morbidity during follow-up, and we may have missed symptom onset, remission, and recurrence in between follow-up visits. We know from prior work that symptoms of anxiety, depression, and PTSD are strong predictors of psychiatric treatment [15]; in fact, patients from this cohort sought care from psychiatrists more than any other medical specialists after hospital discharge [40]. Fourth, although we controlled for potential confounders in our analysis of risk factors, residual confounding may have influenced our findings. Nevertheless, it is not possible to randomize patients to the risk factors of interest (e.g., pre-ARDS depression); thus, observational research provides essential information regarding such associations. Fifth, our findings in ARDS survivors may not generalize to other populations of critical illness survivors. Finally, though we chose widely used and validated symptom thresholds along with rigorous statistical methods (with RCIs) to ensure valid course information, future researchers might consider alternative analytic methods to describe longitudinal psychiatric symptom trajectories (e.g., latent growth curve modeling).

Conclusions

Clinically significant and long-lasting symptoms of anxiety, depression, and PTSD are common in the first 5 years after ARDS. Pre-ARDS depression and psychological distress were strongly associated with prolonged post-ARDS psychiatric morbidity. In-hospital screening of psychiatric history may be useful for counselling patients and family members regarding the risk of prolonged post-ARDS psychiatric morbidity, and for long-term mental health treatment planning after ARDS.

Supplementary Material

Supplemental Figure: Individual and group mean (thick line) symptom trajectories of anxiety (top), depression (middle), and PTSD (bottom), by group (symptoms continuous or recurring vs. absent or remitting)

HADS-A=Hospital Anxiety and Depression Scale-Anxiety subscale; HADS-D=Hospital Anxiety and Depression Scale-Depression subscale; IES-R=Impact of Event Scale-Revised