Post–Intensive Care Unit Syndrome in a Cohort of COVID-19 Survivors in New York City

Abstract

Rationale

The coronavirus disease (COVID-19) pandemic has led to a dramatic increase in the number of survivors of critical illness. These survivors are at increased risk for physical, psychological, and cognitive impairments known collectively as post–intensive care syndrome (PICS). Little is known about the prevalence of PICS in COVID-19 survivors.

Objectives

To report the prevalence of physical, psychological, and cognitive impairment among COVID-19 intensive care unit (ICU) survivors receiving follow-up care in an ICU recovery clinic, to assess for associations between PICS and ICU-related factors, and to compare the cohort of ICU survivors who attended a post-ICU clinic with a cohort of ICU survivors who did not.

Methods

We performed a retrospective cohort study of COVID-19 ICU survivors admitted from March to May 2020 who were subsequently seen in a post-ICU recovery clinic in New York City. We abstracted medical chart data on available clinical screening instruments for physical, psychological, and cognitive impairment. Associations between these outcomes and care-related variables were tested. Baseline characteristics and in-hospital treatments of the post-ICU clinic cohort were compared with those of COVID-19 ICU survivors from the same institution who were not seen in the post-ICU clinic.

Results

Eighty-seven COVID-19 ICU survivors were seen in our post-ICU recovery clinic. The median age was 62 years, and 74% were male. The median length of hospitalization was 51 days, and the median length of ICU stay was 22 days. At the post-ICU follow-up visit, 29%, 21%, and 13% of patients reported clinically significant levels of depressive symptoms, anxiety, and post-traumatic stress disorder symptoms, respectively. Twenty-five percent had cognitive impairment. The overall prevalence of PICS was 90%. There were no associations between length of ICU stay, delirium, and exposure to benzodiazepines, steroids, or systemic paralytics with positive screening results for physical, psychological, or cognitive impairment. Baseline characteristics and ICU-related factors were similar in the cohort of COVID-19 ICU survivors who attended the ICU recovery clinic and those who did not.

Conclusions

PICS is common in COVID-19 survivors. We did not find any association with length of ICU stay or the use of benzodiazepines, steroids, or paralytics.

More than half a million Americans have died of coronavirus disease (COVID-19), but millions more have survived (1). Before the pandemic, it was known that those who survive critical illness are at increased risk for new or worsening depression, anxiety, post-traumatic stress disorder (PTSD), and cognitive and physical impairment, known collectively as post–intensive care syndrome (PICS) (2–5). The effects of PICS can be long-lasting and can impart remarkable physical, emotional, and financial stress on patients, their families, and society at large (2, 4, 6–10).

Known risk factors for the development of PICS include prolonged hypoxemia, acute respiratory distress syndrome (ARDS), delirium, long length of intensive care unit (ICU) stay, and severe sepsis (7, 11–15). These diagnoses are common in critically ill patients with COVID-19. Furthermore, treatment of COVID-19 ARDS often necessitates steroids and systemic paralytics, which are independent risk factors for critical illness neuropathy and physical impairments (16–19). Finally, because of the infectious nature of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), many patients with COVID-19 were isolated from their loved ones, communicated with doctors and nurses infrequently, and were cared for in understaffed and overstretched hospitals (20). This social isolation may predispose COVID-19 ICU survivors to higher rates of PICS than in pre-COVID-19 ICU populations.

Little has been published about COVID-19 and PICS, though leaders in the field of critical illness rehabilitation anticipate that rates could reach unprecedented levels (21). Few single-center descriptive cohort studies describing PICS in COVID-19 ICU survivors exist (22–27). Only one study (24) reported associations between a component of PICS (acute stress disorder) and care received or diagnoses made in the ICU. Furthermore, no studies to date have compared COVID-19 ICU survivors who followed up in an ICU recovery clinic with those who did not, introducing selection bias by reporting outcomes only among those able to attend a recovery clinic (28). This bias could lead to misestimation of the severity of disability in ICU survivors and could limit generalizability to those who do not attend a post-ICU clinic.

The first objective of this study is to describe a group of COVID-19 ICU survivors who attended a post-ICU recovery clinic and to determine the prevalence of physical, psychological, and cognitive impairment. The second objective is to assess for associations between the presence of PICS and ICU-related factors. We hypothesized that ICU factors such as length of ICU stay, systemic paralysis, use of steroids, benzodiazepine exposure, and the incidence of delirium may be associated with PICS. To explore the generalizability of our findings, we sought to compare the clinic cohort with contemporaneously admitted COVID-19 ICU survivors who did not attend the recovery clinic. We hypothesized that the ICU survivors who attended the clinic would be similar in baseline characteristics and ICU care received to those who did not.

Methods

Main Cohort

We performed a retrospective cohort study of ICU survivors with COVID-19 who were admitted to an ICU between March 2020 and May 2020 and subsequently attended an appointment at Weill Cornell Medicine’s (WCM) post-ICU recovery clinic.

Comparison Cohort

A second cohort of ICU survivors with COVID-19 who did not attend the post-ICU clinic were used as a comparison cohort. These patients were admitted to WCM between March 2020 and May 2020 during the peak of the pandemic in New York City. Baseline characteristics, comorbidities, and in-hospital treatments received for COVID-19 ARDS were compared between the two cohorts.

Clinic Procedures

Those who survived ICU admission and were discharged from the hospital before May 2020 were contacted via telephone by a member of the post-ICU team. The clinic and its mission were described, and an appointment was subsequently offered. After May 2020, patients were referred to the post-ICU recovery clinic by inpatient providers, outpatient providers, and self-referral. Between May 2020 and July 2020, patients were seen in the post-ICU recovery clinic by video visits exclusively. After July 2020, patients were offered video or in-person visits.

All patients seen in the post-ICU recovery clinic, whether they were seen in person or via telemedicine, were offered screening for PICS components using validated tools (29, 30). Patients seen in person completed the surveys independently, if able, and patients seen virtually were administered surveys by the treating physicians. Screening results were reviewed with each patient, and referrals were made if the patient had an impairment in any of the domains tested (physical, psychological, and cognitive). Patients with weakness or neuropathy were referred to physical or occupational therapy, patients with focal nerve injuries were referred to rehabilitation medicine or neurology, patients with psychological symptoms were referred to psychology, and patients with cognitive dysfunction were referred to a neuropsychologist or an occupational therapist for cognitive rehabilitation. Physicians also performed a detailed review of the hospital admission, a history of present illness (since hospital discharge), physical exam, and medication reconciliation. The emphasis of the visit was on patient recovery, with special attention to often overlooked aspects of health: sleep, diet, ability to swallow and speak, weight loss or gain, and ability to work, in addition to a more formal evaluation of functional independence. The clinic was staffed by pulmonary and critical care fellows and attending physicians, with support from a critical care nutritionist, a social worker, and an ICU pharmacologist.

Clinic Screening Instruments

The Hospital Anxiety and Depression Scale (HADS) was used to assess anxiety and depressive symptoms. Total scores were recorded, and a score greater than 8 in either domain was considered a positive result (31, 32).

The Post-Traumatic Symptom Scale–10 (PTSS-10) (33–36) was used to assess PTSD symptoms in video visits, and the PTSD Checklist for DSM-5 (PCL-5) (37, 38) was used to assess PTSD symptoms for in-person visits. Total scores were recorded. A cutoff of 35 was considered a positive result on the PTSS-10, and a cutoff of 31 was considered a positive result on the PCL-5 (33, 38).

Scores on the EuroQol-5D-3L (EQ-5D-3L) were used to assess quality of life in video visits, and the EuroQol-5D-5L (EQ-5D-5L) was used to assess quality of life for in-person visits. The EQ-5D-3L gives three answer choices for each question (none, some, and extreme), and the EQ-5D-5L allows five answer choices to the same questions (none, slight, moderate, severe, and extreme). A positive result for any of the EQ-5D subscores was defined as any response that was not “none” (39–41).

The Montreal Cognitive Assessment (MoCA) was used to assess cognitive function for in-person visits, and the MoCA-Blind was used to assess cognitive function for video visits. The MoCA is scored out of 30 points; a score of 25 or less was considered a positive result for cognitive dysfunction (42). The MoCA-Blind is scored out of 22 points, and a score of 18 or less was considered to indicate cognitive dysfunction (43). MoCA and MoCA-Blind scores were not routinely adjusted for level of education, because of lack of education data in the medical record.

Using the available data on psychological, cognitive, and physical impairments above, we created a dichotomous variable indexing presence versus absence of PICS. Presence of PICS was defined as a positive result on any of the following tests: MoCA, PTSS-10 or PCL-5, HADS, or an EQ-5D subscore.

Data Collection

Patient characteristics, inpatient treatment, and outcomes were abstracted retrospectively by physician team members using hospital admission data as well as the outpatient medical record. For the post-ICU recovery clinic cohort, recovery clinic notes and results of the physical, psychological, and cognitive screens were also reviewed. All data with the exception of Sequential Organ Failure Assessment (SOFA) scores were manually abstracted through chart review and managed in the secure, web-based software REDCap (Research Electronic Data Capture) (44, 45). The Weill Cornell-Critical Care Database for Advanced Research automated SOFA score calculation was used to determine the worst daily SOFA score throughout each patient’s hospitalization course (46). The institutional review board at WCM approved this study with a waiver of the requirement to obtain informed consent (20-08022560, 20-04021909).

Statistical Analyses

Cohort characteristics were described using medians and interquartile ranges for continuous variables and sample sizes and percentages for categorical variables. Differences in baseline presentation and in-hospital treatments in ICU survivors who did and did not present to the post-ICU recovery clinic were assessed using Kruskal-Wallis, Fisher exact, and chi-square tests, as appropriate.

For analyses using survey data, MoCA, HADS, EQ-5D-5L, and PTSS-10 scores were analyzed as dichotomous according to their respective cutoffs. Patients who did not complete a screening survey were excluded from that analysis.

Associations of benzodiazepines, steroids, paralytics, and delirium with post-ICU outcome measures were tested using Fisher exact and chi-square tests of independence, as appropriate. Associations between length of ICU stay and post-ICU outcome measures were assessed using Kruskal-Wallis tests. Analyses were conducted in R version 4.0.3 (47) and using the R packages gtsummary and ggplot2 (48, 49).

Results

Main Cohort

Between May 2020 and April 2021, 124 patients were seen by video visits or in person in the WCM post-ICU recovery clinic. Of those, 99 patients were admitted for COVID-19. Of the 99 patients with COVID-19, 94 were critically ill, and 87 were admitted contemporaneously with our comparison cohort and were included in the analysis (Figure 1A). Eighty patients (92%) required intubation and mechanical ventilation.

Figure 1.

(A) Flowchart of the patients included in the post–intensive care unit recovery clinic cohort. (B) Flowchart of patients included in the comparison cohort of coronavirus disease intubated survivors. COVID-19 = coronavirus disease; ICU = intensive care unit.

The patients in the post-ICU recovery clinic had a median age of 62 years, 74% were male, 34% were White, and 38% were Hispanic. Many patients had preexisting comorbidities, most commonly hypertension (46%), diabetes (36%), and cardiovascular disease (28%) (Table 1). No data were available regarding preexisting mental health diagnoses or preexisting cognitive dysfunction.

Table 1.

Baseline demographics and comorbidities and in-hospital treatments and outcomes

Characteristic	N = 87
Demographics and comorbidities
Age, yr	62 (50–70)
Sex
Female	23 (26)
Male	64 (74)
BMI on admission, kg/m2	30 (27–32)
BMI on discharge, kg/m2	25.1 (22.1–28.7)
Race
Asian	19 (22)
Black	5 (5.7)
White	30 (34)
Declined/other	33 (38)
Ethnicity
Hispanic or Latino or Spanish origin	33 (38)
Not Hispanic or Latino or Spanish origin	5 (5.7)
Unknown or not specified	49 (56)
Social Vulnerability Index	0.73 (0.36–0.83)
Hypertension	38 (44)
Cardiovascular excluding hypertension	22 (25)
Pulmonary	15 (17)
Renal	3 (3.4)
Diabetes	31 (36)
Psychiatric	3 (3.4)
Malignancy	8 (9.2)
In-hospital treatments and outcomes
Length of hospitalization, d	51 (38–80)
Length of intubation, d*	17 (11–24)
Length of ICU stay, d	22 (11–42)
Sequential Organ Failure Assessment score	12 (11–14)
Tracheostomy	38 (44)
Length of tracheostomy, d	59 (51–73)
Benzodiazepine exposure	57 (66)
Paralytics	58 (68)
Prone positioning	34 (40)
Chest tube	10 (12)
Renal replacement therapy	9 (10)
Vasopressors	69 (81)
Steroids	52 (61)
Therapeutic anticoagulation	50 (59)
Thrombotic event	27 (31)
Diabetic ketoacidosis/hyperglycemia	19 (22)
Delirium	61 (71)
Pharmacotherapy for delirium	49 (57)
Interleukin-6 inhibitor	27 (31)
Remdesivir	22 (26)

Definition of abbreviations: BMI = body mass index; ICU = intensive care unit.

Data are shown as median (interquartile range) or n (%).

^*Length of intubation was defined as the time from intubation to extubation or tracheostomy placement.

The median time to follow-up in the recovery clinic was 20 days. The median length of hospital stay was 51 days, the median length of ICU stay was 22 days, and the median length of intubation was 17 days (Table 1). The duration of intubation was defined as the time between endotracheal tube placement and either extubation or tracheostomy placement. Forty-four percent of patients received a tracheostomy. The most common complications during hospitalization were delirium, defined by the presence of a delirium diagnosis in the attending physician’s progress note (71%); thrombotic events (31%); and diabetic ketoacidosis or hyperglycemia (22%) (Table 1). Twelve percent of patients required a chest tube to treat pneumothorax, and 10% required initiation of renal replacement therapy. Sixty-six percent of patients were exposed to benzodiazepines. Sixty-eight percent of patients were treated with systemic paralytics, and 40% were treated with prone positioning. Eighty-one percent received vasopressors (Table 1). Regarding COVID-19-specific treatments, 61% of patients received steroids, 31% received an interleukin-6 inhibitor, and 26% received remdesivir (Table 1).

Scores on screening instruments for anxiety, depression, PTSD, cognitive function, and quality of life were available for 68 patients. Sixty-three patients completed the HADS and PTSS-10 or PCL-5. Of those, 29% screened positive for depression, 21% screened positive for anxiety, and 13% screened positive for post-traumatic stress symptoms. MoCA or MoCA-Blind scores were available for 59 patients, and 25% screened positive for cognitive impairment. Sixty-seven patients completed the EQ-5D-3L or EQ-5D-5L to assess quality of life. Results from the EQ-5D-5L and EQ-5D-3L revealed that 61% of patients reported difficulty with mobility, 57% reported difficulty with self-care, and 76% reported difficulty in performing usual activities (Table 2).

Table 2.

Intensive care unit recovery clinic surveys

Characteristic	N = 87
Post-ICU syndrome (any EQ-5D, HADS-T, MoCA, PTSS-10/PCL-5)	61/68 (90)
HADS-T screen positive	31/63 (49)
HADS-A screen positive	13/63 (21)
HADS-D screen positive	18/63 (29)
PTSS-10 or PCL screen positive	8/63 (13)
MoCA screen positive	15/59 (25)
EQ-5D mobility	41/67 (61)
EQ-5D self-care	38/67 (57)
EQ-5D usual activities	51/67 (76)
Any EQ-5D physical impairment*	54/67 (81)

Definition of abbreviations: EQ-5D = EuroQol-5D; HADS-A = Hospital Anxiety and Depression Scale–Anxiety; HADS-D = Hospital Anxiety and Depression Scale–Depression; HADS-T = Hospital Anxiety and Depression Scale–Total; ICU = intensive care unit; MoCA = Montreal Cognitive Assessment; PCL-5 = PTSD Checklist for the DSM-5; PTSD = post-traumatic stress disorder; PTSS-10 = Post-Traumatic Symptom Scale–10.

Data are shown as n/N (%).

^*Any response other than no on the EQ-5D mobility, self-care, or usual activities survey.

Among patients who attended the post-ICU clinic and completed screening surveys, the prevalence of PICS was 90%, as defined by impairment in at least one of the physical, psychological, or cognitive domains. Among post-ICU clinic attendees who completed at least one survey, 29 patients had a positive score in one domain (physical, psychological, or cognitive), 19 patients had positive scores in two domains, and 11 patients had positive scores in all three domains. The most common referrals made in the post-ICU clinic were to an interstitial lung disease clinic for post-COVID-19 pulmonary fibrosis (23%) or to a psychologist or psychiatrist (22%).

In patients who attended the recovery clinic and completed screening surveys, there were no associations between length of ICU stay and positive results for physical impairment, anxiety, depression, PTSD, or cognitive impairment (Figures 2A–2F). There were no associations between ICU treatments (use of benzodiazepines, steroids, or systemic paralytics) and positive results for anxiety, depression, PTSD, or cognitive impairment. Treatment with paralytics was associated with a positive result for pain on the EQ-5D. ICU delirium was not associated with any positive results (Tables 3–6).

Figure 2.

(A–F) Scatterplots illustrating the associations between length of intensive care unit (ICU) stay in days and the surveys completed in the post-ICU recovery clinic. EQ-5D = EuroQol-5D; HADS = Hospital Anxiety and Depression Scale–Total; MOCA = Montreal Cognitive Assessment; PCL = PTSD Checklist for the DSM-5; PTSD = post-traumatic stress disorder; PTSS = Post-Traumatic Symptom Scale–10.

Table 3.

Unadjusted associations with benzodiazepine exposure

Characteristic	Not Exposed (n = 27)	Exposed (n = 57)	P Value*
Post-ICU syndrome (any EQ-5D, HADS-T, MoCA, PTSS-10/PCL-5)	17/20 (85)	42/46 (91)	0.43
HADS-A screen positive	2/18 (11)	10/43 (23)	0.48
HADS-D screen positive	4/18 (22)	13/43 (30)	0.52
HADS-T screen positive	7/18 (39)	23/43 (53)	0.30
PTSS-10 or PCL-5 screen positive	3/19 (16)	5/42 (12)	0.70
MoCA screen positive	5/16 (31)	9/41 (22)	0.50
EQ-5D mobility	9/20 (45)	30/45 (67)	0.10
EQ-5D self-care	9/20 (45)	27/45 (60)	0.26
EQ-5D usual activities	16/20 (80)	33/45 (73)	0.56

Definition of abbreviations: EQ-5D = EuroQol-5D; HADS-A = Hospital Anxiety and Depression Scale–Anxiety; HADS-D = Hospital Anxiety and Depression Scale–Depression; HADS-T = Hospital Anxiety and Depression Scale–Total; ICU = intensive care unit; MoCA = Montreal Cognitive Assessment; PCL-5 = PTSD Checklist for the DSM-5; PTSD = post-traumatic stress disorder; PTSS-10 = Post-Traumatic Symptom Scale–10.

Data are shown as n/N (%), where N is the total with the post-ICU survey measured. Of our cohort, 68 completed at least one survey, 63 completed the HADS, 63 completed the PTSS-10 or PCL-5, 59 completed the MoCA, and 67 completed the EQ-5D.

^*Statistical tests performed: Fisher exact test and chi-square test of independence.

Table 4.

Unadjusted associations with steroids

Characteristic	No Steroids (n = 33)	Received Steroids (n = 52)	P Value*
Post-ICU syndrome (any EQ-5D, HADS-T, MoCA, PTSS-10/PCL-5)	23/26 (88)	37/41 (90)	>0.99
HADS-A screen positive	5/23 (22)	7/39 (18)	0.75
HADS-D screen positive	9/23 (39)	8/39 (21)	0.11
HADS-T screen positive	11/23 (48)	19/39 (49)	0.95
PTSS-10 or PCL-5 screen positive	3/23 (13)	5/39 (13)	>0.99
MoCA screen positive	7/22 (32)	7/36 (19)	0.29
EQ-5D mobility	15/26 (58)	25/40 (62)	0.70
EQ-5D self-care	15/26 (58)	22/40 (55)	0.83
EQ-5D usual activities	19/26 (73)	31/40 (78)	0.68
EQ-5D pain/discomfort	16/26 (62)	22/40 (55)	0.60
EQ-5D anxiety/depression	15/26 (58)	16/40 (40)	0.16

Definition of abbreviations: EQ-5D = EuroQol-5D; HADS-A = Hospital Anxiety and Depression Scale–Anxiety; HADS-D = Hospital Anxiety and Depression Scale–Depression; HADS-T = Hospital Anxiety and Depression Scale–Total; ICU = intensive care unit; MoCA = Montreal Cognitive Assessment; PCL-5 = PTSD Checklist for the DSM-5; PTSD = post-traumatic stress disorder; PTSS-10 = Post-Traumatic Symptom Scale–10.

Data are shown as n/N (%), where N is the total with the post-ICU survey measured. Of our cohort, 68 completed at least one survey, 63 completed the HADS, 63 completed the PTSS-10 or PCL-5, 59 completed the MoCA, and 67 completed the EQ-5D.

^*Statistical tests performed: Fisher exact test and chi-square test of independence.

Table 5.

Unadjusted associations with paralytics

Characteristic	No Paralytics (n = 27)	Received Paralytics (n = 58)	P Value*
Post-ICU syndrome (any EQ-5D, HADS-T, MoCA, PTSS-10/PCL-5)	19/21 (90)	41/46 (89)	>0.99
HADS-A screen positive	4/18 (22)	8/44 (18)	0.73
HADS-D screen positive	5/18 (28)	12/44 (27)	0.97
HADS-T screen positive	8/18 (44)	22/44 (50)	0.69
PTSS-10 or PCL-5 screen positive	3/19 (16)	5/43 (12)	0.69
MoCA screen positive	4/16 (25)	10/42 (24)	>0.99
EQ-5D mobility	13/21 (62)	27/45 (60)	0.88
EQ-5D self-care	11/21 (52)	26/45 (58)	0.68
EQ-5D usual activities	15/21 (71)	35/45 (78)	0.58
EQ-5D pain/discomfort	8/21 (38)	30/45 (67)	0.029
EQ-5D anxiety/depression	9/21 (43)	22/45 (49)	0.65

Definition of abbreviations: EQ-5D = EuroQol-5D; HADS-A = Hospital Anxiety and Depression Scale–Anxiety; HADS-D = Hospital Anxiety and Depression Scale–Depression; HADS-T = Hospital Anxiety and Depression Scale–Total; ICU = intensive care unit; MoCA = Montreal Cognitive Assessment; PCL-5 = PTSD Checklist for the DSM-5; PTSD = post-traumatic stress disorder; PTSS-10 = Post-Traumatic Symptom Scale–10.

Data are shown as n/N (%), where N is the total with the post-ICU survey measured. Of our cohort, 68 completed at least one survey, 63 completed the HADS, 63 completed the PTSS-10 or PCL-5, 59 completed the MoCA, and 67 completed the EQ-5D.

^*Statistical tests performed: Fisher exact test and chi-square test of independence.

Table 6.

Unadjusted associations with delirium

Characteristic	No Delirium (n = 25)	Delirium Present (n = 61)	P Value*
Post-ICU syndrome (any EQ-5D, HADS-T, MoCA, PTSS-10/PCL-5)	16/21 (76)	45/47 (96)	0.025
HADS-A screen positive	4/17 (24)	9/46 (20)	0.74
HADS-D screen positive	6/17 (35)	12/46 (26)	0.47
HADS-T screen positive	8/17 (47)	23/46 (50)	0.84
PTSS-10 or PCL-5 screen positive	3/18 (17)	5/45 (11)	0.68
MoCA screen positive	4/17 (24)	11/42 (26)	>0.99
EQ-5D mobility	11/20 (55)	30/47 (64)	0.50
EQ-5D self-care	9/20 (45)	29/47 (62)	0.21
EQ-5D usual activities	13/20 (65)	38/47 (81)	0.21
EQ-5D pain/discomfort	8/20 (40)	31/47 (66)	0.049
EQ-5D anxiety/depression	9/20 (45)	23/47 (49)	0.77

Definition of abbreviations: EQ-5D = EuroQol-5D; HADS-A = Hospital Anxiety and Depression Scale–Anxiety; HADS-D = Hospital Anxiety and Depression Scale–Depression; HADS-T = Hospital Anxiety and Depression Scale–Total; ICU = intensive care unit; MoCA = Montreal Cognitive Assessment; PCL-5 = PTSD Checklist for the DSM-5; PTSD = post-traumatic stress disorder; PTSS-10 = Post-Traumatic Symptom Scale–10.

Data are shown as n/N (%), where N is the total with the post-ICU survey measured. Of our cohort, 68 completed at least one survey, 63 completed the HADS, 63 completed the PTSS-10 or PCL-5, 59 completed the MoCA, and 67 completed the EQ-5D.

^*Statistical tests performed: Fisher exact test and chi-square test of independence.

Comparison Cohort

We compared baseline characteristics and features of ICU admission between a cohort of 87 COVID-19 ICU survivors who attended the post-ICU recovery clinic and a cohort of 193 COVID-19 ICU survivors who did not attend the recovery clinic. Age, sex, and race were similar between the two groups, but more Hispanic patients attended the post-ICU clinic than did not. The Social Vulnerability Index was similar between the two groups. The highest SOFA score on admission was similar between the two groups (Table 7). The use of steroids, prone positioning, and tracheostomy was similar between the two groups. Both cohorts had long hospitalizations (median length > 40 d) and long lengths of intubation (median length > 17 d). Patients who attended the post-ICU clinic were more likely to have received remdesivir and were more likely to have preexisting cardiovascular disease. Patients who did not attend the post-ICU clinic were more likely to have preexisting renal disease and were more likely to have been discharged to an acute, a subacute, or a skilled nursing facility, as opposed to discharged home (Table 7).

Table 7.

Comparison with entire intubated cohort

Characteristic	Intubated and Survived (N = 193)	Post-ICU Clinic Cohort (N = 87)	P Value*
Baseline demographics and comorbidities
Age, yr	62 (51–71)	62 (50–70)	0.80
Sex			0.23
Female	65 (34)	23 (26)
Male	128 (66)	64 (74)
BMI, kg/m2	28.3 (25.3–33.2)	30.0 (26.6–31.9)	0.46
Race			0.60
Asian	36 (19)	19 (22)
Black	20 (10)	5 (5.7)
White	62 (32)	30 (34)
Declined/other	75 (39)	33 (38)
Ethnicity			<0.001
Hispanic or Latino or Spanish origin	51 (26)	33 (38)
Not Hispanic or Latino or Spanish origin	83 (43)	5 (5.7)
Unknown or not specified	59 (31)	49 (56)
Social Vulnerability Index	0.68 (0.41–0.84)	0.73 (0.36–0.83)	0.75
Diabetes	66 (34)	31 (36)	0.82
Renal	20 (10)	3 (3.4)	0.05
Pulmonary	38 (20)	15 (17)	0.63
Malignancy	8 (4.1)	8 (9.2)	0.10
Cardiovascular excluding hypertension	32 (17)	22 (25)	0.09
Hypertension	106 (55)	38 (44)	0.08
In-hospital outcomes and treatments
Length of hospitalization, d	42 (26–57)	51 (38–80)	<0.001
Length of intubation, d	19 (11–36)	17 (11–24)	0.028
Worst SOFA score	12 (11–14)	12 (11–14)	0.08
Steroids	130 (67)	52 (61)	0.34
Prone positioning	88 (46)	34 (40)	0.43
Renal replacement therapy	35 (18)	9 (10)	0.11
Tracheostomy	83 (45)	38 (44)	0.90
Remdesivir	24 (12)	22 (26)	0.009
Discharge location
Acute rehabilitation	55 (28)	10 (11)	<0.001
Home	54 (28)	59 (68)
Hospice	3 (1.6)	0 (0)
Other or unknown	17 (8.8)	7 (8.0)
Skilled nursing facility	19 (9.8)	2 (2.3)
Subacute rehabilitation	45 (23)	9 (10)

Definition of abbreviations: BMI = body mass index; ICU = intensive care unit; SOFA = Sequential Organ Failure Assessment.

Data are shown as median (interquartile range) or n (%). The comparison cohort contains all patients who arrived at Weill Cornell Medicine or were transferred to Weill Cornell Medicine and were intubated between March 3, 2020, and May 19, 2020. Patients who came to our ICU clinic and were admitted during this time are now removed from the comparison cohort (intubated survivors). Note that variables may not be defined identically. BMI was measured on discharge for the post-ICU cohort and at any point during hospitalization for the comparison cohort. Race and ethnicity were condensed for the post-ICU cohort to match the comparison cohort. “Renal” refers to chronic kidney disease or end-stage renal disease. “Pulmonary” refers to asthma, chronic obstructive pulmonary disease, interstitial lung disease, or obstructive sleep apnea for both cohorts. “Cardiovascular without hypertension” refers to coronary artery disease or heart failure.

^*Statistical tests performed: chi-square test of independence and Wilcoxon rank sum test. P values in boldface type denote statistical significance.

Discussion

In this study, we describe a population of 87 COVID-19 survivors who presented to a well-established post-ICU recovery clinic. To the best of our knowledge, this is the largest study to examine PICS in COVID-19 ICU survivors seen in a post-ICU recovery clinic. It is also the only study to compare ICU survivors who attended a post-ICU recovery clinic with ICU survivors who did not.

The prevalence of PICS in our cohort, defined as impairment in at least one of the physical, psychological, or cognitive domains, was 90%. This is higher than the previously reported prevalence of PICS in ARDS survivors (4) (up to 60%) but similar to the only other published prevalence of PICS in COVID-19 ICU survivors (23). Positive screening results in more than one domain were seen in 44.8% of patients. This overlap is higher than reported in the existing ARDS literature, in which only 25% of patients had impairments in cooccurring domains at 3-month follow-up (50). This may be partially explained by the large percentage of our patients (81%) with physical impairments and the severity of illness in our cohort, as well as the short median time to follow-up in the post-ICU clinic (20 d).

We found no associations of length of ICU stay, use of steroids, systemic paralytics, or benzodiazepines, and ICU delirium with positive screening results for anxiety, depression, PTSD, or cognitive impairment. Treatment with paralytics was associated with a positive result for pain in our quality-of-life survey. These results suggest that severe COVID-19 is a risk factor for post-ICU syndrome irrespective of ICU factors. However, our relatively small sample size and analytic approach limit the conclusions we can make. There were no clinically significant differences between the baseline characteristics in the cohort of COVID-19 ICU survivors who attended the post-ICU clinic compared with a cohort of COVID-19 ICU survivors who did not attend the clinic. However, those who attended the post-ICU clinic less commonly had underlying kidney disease and more often had cardiovascular disease. These findings seem unlikely to affect the prevalence of PICS in ICU survivors. The patients who attended the clinic were more likely to be discharged home rather than to a rehabilitation or skilled nursing facility. These data support the notion that admission to inpatient rehabilitation remains a barrier to post-ICU follow-up care.

The severity of illness in our cohort warrants emphasis. Our patients were older than those in previously published cohorts and had a longer median duration of mechanical ventilation and a longer median length of hospital stay compared with other published studies. For example, a recently published cohort of 45 patients in another New York City post-ICU clinic had a mean age of 53 years, a median length of hospitalization of 18 days, and a median length of intubation of 8 days (23). Furthermore, fewer patients in that cohort required paralysis, prone positioning, and tracheostomy (23).

It is worthwhile to break down the specific components of PICS within our cohort and to compare them with existing prepandemic data.

Before the COVID-19 pandemic, rates of cognitive impairment among ICU survivors ranged from 24% to 61% (15, 51–53). Our cohort possessed many high-risk characteristics: ARDS, a long duration of intubation, a high rate of delirium, and benzodiazepine exposure (53). Furthermore, reports of “brain fog” following COVID-19 infection, even when mild, have been widespread (54). However, only 25% of our cohort screened positive for cognitive impairment. We used the MoCA to screen for cognitive impairment, while some studies of ARDS survivors have shown a much higher incidence (50%–73%) of neurocognitive dysfunction (14, 15), perhaps because of more rigorous testing. Finally, there is a large rate of attrition from those who survive the ICU to those who make it to the post-ICU clinic, which may select for patients with less cognitive impairment.

The rates of depression, anxiety, and PTSD in our COVID-19 post-ICU cohort (29%, 21%, and 13%, respectively) were lower than reported in the existing literature (55–57). One study reported associations between post-ICU depression, younger age, and female gender (58). Our population was older and mostly male, which could contribute to the lower prevalence described (58). In addition, during the pandemic, WCM opened an inpatient rehabilitation unit that provided patients with access to neuropsychologists and group therapy with other COVID-19 survivors (59). In this unique setting, many patients already noted improvement in their psychological symptoms before discharge home.

Many ICU survivors have persistent physical impairment, including weakness, critical illness neuropathy and/or myopathy, and muscle atrophy (60, 61). Known risk factors for critical illness–related neuromuscular abnormalities include prolonged length of ICU stay, sepsis, multiorgan dysfunction, renal replacement therapy, and administration of vasopressors (10, 61). The prevalence of physical impairment in our COVID-19 cohort was 81%, which is consistent with previous reports (60, 61).

We hypothesized that benzodiazepine exposure, treatment with corticosteroids, and use of systemic paralytics were likely to be associated with positive screening results for anxiety, depression, PTSD, cognitive impairment, or decreased quality of life. However, no significant associations were found. Of note, prior studies have assessed duration of delirium and found statistically significant associations with PICS (29), but we did not have access to reliable data on the duration of delirium in our cohort. The lack of associations in our study may suggest that other factors contribute to PICS or perhaps that this population of COVID-19 ICU survivors is so uniformly high risk for PICS that individual risk factors cannot be isolated. Furthermore, the study may have been underpowered to detect a statistically significant association.

There were few significant differences between the baseline characteristics or ICU treatment in the cohort of COVID-19 ICU survivors who attended the post-ICU clinic compared with a cohort of COVID-19 ICU survivors who did not attend the clinic. This is important because the ability to attend a post-ICU clinic may select for healthier survivors with fewer impairments, contributing to selection bias (28). In this case, length of hospital stay, duration of mechanical ventilation, SOFA score, and the frequency with which patients were treated with systemic paralysis, prone positioning, and tracheostomy were similar between the two cohorts, reducing the likelihood of selection bias and strengthening the generalizability of these results to all ICU survivors. Furthermore, the Social Vulnerability Index, a marker of the negative health effects from external stressors, was similar in both groups, indicating neither cohort reflected a more vulnerable population.

The strengths of this study include the relatively large cohort size of 87 patients, 80 of whom were intubated, and the severity of their illness. Multidimensional data on physical, psychological, and cognitive domains were available. Furthermore, there were few differences in demographics or severity of illness between survivors who attended the post-ICU clinic and those who did not, suggesting that these data may be generalizable to COVID-19 ICU survivors who do not attend an ICU recovery clinic. Finally, with relatively robust numbers for a post-ICU clinic, we not only described this population but also examined associations between PICS and modifiable clinical interventions (62).

Our study is not without limitations. Some patients (n = 19) did not complete the screening surveys during their clinic visits, mostly citing time constraints, which could have influenced the measured prevalence of PICS in our cohort. Duration of mechanical ventilation could not be measured, because of charting restraints during the height of the pandemic, so length of ICU stay and duration of endotracheal intubation were used as a surrogates. However, it is understood that neither is a perfect approximation of duration of mechanical ventilation. The surveys were available only in English, and if patients did not speak English, we allowed family members to assist with survey completion, introducing a possible bias. Despite the representative sample, we cannot exclude selection bias as a factor in the relatively low rates of anxiety, depression, PTSD, and cognitive impairment or in the lack of evidence for associations with in-hospital treatments. Our association measures between hospital treatment and PICS may be confounded by the overall illness severity in our population, with a median SOFA of 12.

Conclusions

We describe a population of COVID-19 ICU survivors and their impairments. PICS was prevalent but was not associated with length of ICU stay or use of benzodiazepines, steroids, or paralytics in this single-center cohort. As COVID-19 vaccination efforts accelerate and intensivists see a decline in acute COVID-19 infections, we must not forget the suffering and debility that ICU survivors and their families experience. The existence of post-ICU recovery clinics is vital to providing multidisciplinary care for this vulnerable population. It is our responsibility as intensivists, and as a healthcare system, to care for our patients not only in the ICU but also as they face the aftermath of critical illness. To succeed in this important goal, more research is needed to confirm modifiable risk factors for PICS in COVID-19 survivors (62).