Study reference

Study characteristics

Patient characteristics

Candidate predictors

Model development, performance and evaluation

Outcome measures and results

Comments

Interpretation of model

Asimakopoulou, 2014

Source of data¹ and date:

Cohort March 2009 – June 2011

Setting/ number of centres and country:

5 ICU’s in the greater Athens area

Funding and conflicts of interest:

Funding: not reported

Conficts of interest: none

Recruitment method²:

Consecutive patients who met the inclusion criteria:

“The ICU sample delivered from the list of all hospitalised

ICU patients during the described dates included 1,986 patients.

Out of this group, 1,002 patients died during ICU treatment

or after discharge. From the remaining sample of 984 patients,

834 were ineligible for the study for various reasons (cognitive

impairment; not traceable). The final sample consisted of 150

subjects, 48 of whom refused to participate (thus a 68% response

rate).”

Inclusion criteria:

• Adult patients aged between 18 and 70 years
• Admission to ICU for more than 24 hours
• Ability to communicate and contact
• Informed written consent

Exclusion criteria:

• Cognitive impairment
• Not traceable

Treatment received:
not described

Participants: N=102

Mean age ± SD:

44.95±15.17 years

Sex: % M / % F

65/102 (63.7%) M

37/102 (36,3%)

Other important characteristics:

ICU length of stay (days) 33.61 (range: 2-169)

Describe candidate predictors³ and method and timing of measurement:

“Demographic and clinical information was collected on all

patients, from the hospitals’ information systems.”

Predictor 1:

Gender (female/male)

Predictor 2:

Age (in years)

Predictor 3:

Organ loss (yes/no)

Predictor 4:
Disability (not described/patient selfcare)

Predictor 5:

Sedation (yes/no)

Predictor 6:

Length of stay (16-60 days/>61 days/1-15 days)

Number of participants with any missing value⁴?

Not reported

How were missing data handled⁵?

Not specified.

Development

Modelling method⁶:

“Independent variables that were found to bear a statistically significant association with the outcome in the context of univariate analyses

were then entered simultaneously into the logistic regression model”

Performance

Calibration measures⁷ and 95%CI:

Likely Hosmer-Lemeshow test: p>0.05

Discrimination measures⁸ and 95%CI:

Not reported

Classification measures⁹:

Not reported

Evaluation

Method for testing model performance¹⁰:

Not reported

Type of outcome:

Combination of major depression and PTSD after 3 months

Definition and method for measurement of outcome:

Mini International Neuropsychiatric Interview – MINI (version 5.0.0) (Sheehan et al., 1998)

Depression and PTSD
“Major depression and PTSD were recorded according to

MINI–DSM-IV clinical symptoms criteria. The phrasing of the

questions related to detection of symptoms – criteria for DSMIV diagnoses were modified in order to cover the time between the discharge from the hospital and the time of the interview.”

Endpoint or duration of follow-up:

3 months

Number of events/outcomes:

17.6% (n=18) comorbidity major depression with PTSD

RESULTS

Multivariable model¹¹:

Comorbidity major depression and PTSD

Alternative presentation of final model¹²:

Not reported

Interpretation:
The

strongest predictors of major depression–PTSD comorbidity in

the ICU group were the following: female gender, young age

and severity of illness

Davydow, 2013

Source of data¹ and date:

Cohort September 2010 – August 2011

Setting/ number of centres and country:

Harborview medical center, Seattle, WA

Funding and conflicts of interest:

Funding: non-commercial

Conficts of interest: Dr. Katon discloses that he has

received honoraria in the last 12 months for CME lectures funded indirectly by Lilly, Forest and Pfizer.

Recruitment method²:

Consecutive patients

Inclusion criteria:

• Admitted to any of the ICU’s of HMC

Exclusion criteria:

- Initial admission diagnosis of traumatic injury

- pre-existing cognitive impairment or dementia diagnosis noted in the medical record

- communication/language barrier

- ICU length of stay (LOS) ≤ 24 hours

- pre-existing medical illness with life expectancy of < 12 months

- admission for a suicide attempt.

Treatment received:
not described

Participants: N=150
120 (80%) completed a 12-month telephone

follow-up interview

Characteristics of patients who completed follow-up:

Mean age ± SD:

49.0±14.6 years

Sex: % M / % F

69/120 (57.5%) M

51/120 (42.5%) F

Other important characteristics:

ICU length of stay (days) 7.5 (8.2)

Describe candidate predictors³ and method and timing of measurement:

“Baseline patient characteristics and ICU

clinical factors were obtained through medical record review and in-person interview with

the patient.”

Predictor 1:

Age in years

Predictor 2:

Sex

Predictor 3:

Race

Predictor 4:

Education

Predictor 5: marital/partnered state

Predictor 6:

lifetime history of major depression

Predictor 7:

number of prior traumatic event exposures

Predictor 8:

problem drinking at baseline

Predictor 9:

problem drug use at baseline

Predictor 10:

Charlson comorbidity score

Predictor 11:

SAPS II score

Predictor 12:

duration of mechanical ventilation

Predictor 13:

number of surgeries

Predictor 14:

number of blood product transfusions

Predictor 15:

days physically restrained

Predictor 16:

presence of probable delirium during the hospitalization

Predictor 17:

days of exposure to benzodiazepines

Predictor 18:

Opiods

Predictor 19:

Antipsychotics

Predictor 20:

Antidepressants

Predictor 21:

substantial acute stress syndrome.

Number of participants with any missing value⁴?

30 patients did not complete the 12 month interview.

How were missing data handled⁵?

Not specified.

Development

Modelling method⁶:

Mixed-model linear regression analysis with pre-specified confounding variables

Performance

Calibration measures⁷ and 95%CI:

Not reported

Discrimination measures⁸ and 95%CI:

Not reported

Classification measures⁹:

Not reported

Evaluation

Method for testing model performance¹⁰:

Not reported

Type of outcome:

1) symptoms of posttraumatic stress disorder

over the course of the year following medical-surgical intensive care unit admission

2) symptoms of depression following medical surgical

intensive care unit admission

Definition and method for measurement of outcome:

1) PTSD Checklist-civilian version. “Substantial PTSD symptoms can be ascertained with the PCL-C by following an algorithm that considers a score of 3 or more on

at least 1 intrusive symptom, 3 avoidant symptoms, and 2 arousal symptoms as consistent with DSM diagnostic criteria”

2) Patient Health Questionnaire-9 “We defined substantial depressive symptoms as a PHQ-9 score ≥ 10”

Endpoint or duration of follow-up:

12 months

Number of events/outcomes:

“At 3 months post-discharge from the ICU (mean: 88.8 days, SD: 28.0), 16% (95%Confidence Interval [95%CI]: 10%, 23%) of patients had substantial PTSD symptoms and 31% (95%CI: 23%, 39%) had substantial depressive symptoms.”

“One year after ICU admission (mean: 356.4 days, SD: 29.2), the prevalence of substantial depressive symptoms had decreased to 17% (95%CI: 10%, 23%) while the

prevalence of substantial PTSD symptoms was 15% (95%CI: 9%, 21%).”

RESULTS

Multivariable model¹¹:

PTSD

PTSD

Depression

Depression

Alternative presentation of final model¹²:

Not reported

Interpretation:
“In addition to lifetime history of major depression

and greater prior traumatic event exposure, we identified that a potentially modifiable factor,

in-hospital substantial acute stress symptoms, was independently associated with greater

severity of post-ICU PTSD and depressive symptoms.”

Hopkins, 2010

Source of data¹ and date:

Acute respiratory distress syndrome survivors from a

single-center mechanical ventilation randomized clinical

trial of higher vs. lower tidal volume target strategies

conducted from February 1994 to December 1999

Setting/ number of centres and country:

Single-center; USA

Funding and conflicts of interest:

Not reported

Recruitment method²:

First author invited ARDS-survivors from

single-center mech. ventilation RCT of higher vs. lower tidal volume target strategies conducted from February 1994 to December 1999 [20]

120 patients evaluated for this study, of which

• 42 patients had died
• 3 excluded due to cognitive disability/ Alzheimer’s disease
• 1 declined

Inclusion criteria:

• Tracheal intubation
• Ratio of arterial oxygen tension to inspired oxygen fraction (PaO2/FiO2) ≤150 mmHg,
• Pulmonary artery balloon occlusion pressure ≤18 mmHg (when available)
• no clinical evidence of left atria hypertension
• diffuse infiltrates in three out of four quadrants on chest radiographs
• age >16 years
• presence of an ARDS risk factor (e.g., aspiration, multiple trauma, pancreatitis, pneumonia, sepsis)

Exclusion criteria:

• disease states that deemed to be rapidly terminal (e.g., liver failure, malignancy, AIDS),
• traumatic brain injury
• prior neurologic disease (e.g., stroke, dementia, multiple sclerosis, etc.)
• enrollment in another ARDS study (e.g., [NIH/NHLBI] ARDS Network studies)

Treatment received:
mechanical ventilation before discharge, but treatment after discharge was not described

Participants: N=74

Mean age ± SD:

46±16 years

(range 16–81 years),

Sex: % M / % F

33/74 (45%) M /

41/74 (55%) F

Other important characteristics:

Hospital length of stay (days) 39±22

Number of ARDS risk factors, mean±SD: 3.1±1.3 0–1.8

Intubation duration (days), mean±SD:

28±19 3–94

ICU length of stay (days); mean±SD: 

34±20 5–98

Hospital length of stay (days); mean±SD: 39±22 7–122

Charlson comorbidity index [36]; mean±SD:  1±1.2 0–4

Study enrollment

  APACHE II score

  18.1±6.6 5–41

  MOF score

  7.2±3.5 0–18

  PaO2/FiO2

  107±32 54–151

  FiO2 (%)

  68±15 40–100

  PaO2 mmHg

  69±13 47–121

Describe candidate predictors³ and method and timing of measurement:

“Patient demographic and medical data were collected and

recorded prospectively using hospital-wide clinical electronic

medical records”

Predictor 1:

history of alcohol dependence;

defined using DSM IV criteria [23].

Predictor 2:

Gender

Predictor 3:

Age

Number of participants with any missing value⁴?

1 year FU:

N = 8/74 (11%);

(3x death, 5x decline 1y FU)

How were missing data handled⁵?

Not specified; patient excluded from analysis

Development

Modelling method⁶:

stepwise multiple regressions with bootstrap procedure

Performance

Calibration measures⁷ and 95%CI:

Not reported

Discrimination measures⁸ and 95%CI:

Not reported

Classification measures⁹:

R²=.177, R^2adj=.138

Evaluation

Method for testing model performance¹⁰:

internal validation; bootstrap procedure;

“A bootstrap procedure was used to confirm which independent variables significantly predicted these outcomes. The bootstrap method takes a random sample of

the original dataset with replacement. Sampling with

replacement means that every sample is returned to the data set after sampling. Thus, a particular data point from the original data set could appear multiple times in a given bootstrap sample. A regression analysis is then computed for this bootstrap sample. The regression coefficients (B and beta) are recorded and the procedure starting with a new sample is repeated (1000 times total in our analysis for each outcome). Confidence intervals (95%) are then computed for each regression coefficient. Any variable that is clearly related to the outcome will have confidence intervals for its coefficients that do not include zero. Thus, those variables which met the confidence interval criterion were selected and entered into a multiple regression analysis to determine the significance of the model.”

Type of outcome:

Anxiety (model does not fit inclusion criteria)

Depression at 1-year follow-up

Definition and method for measurement of outcome:

Depression:

Beck Depression Inventory (BDI); 0–9 indicate minimal, 10–16 mild, 17–29 moderate and 30–63 severe depression; cut-off: >16

Endpoint or duration of follow-up:

1 year

Number of events/outcomes:

16% (n=11) had scores N>16 on the Beck Depression Inventory at 1 year

Depression score, mean±SD were within normal range:

1 year

9.3±9.5 (range 0–41)

RESULTS

Multivariable model¹¹:

Depression – 1 year (n=64) - bootstrap

Model summary: R2=.177, R2adj=.138, F(3,63)=4.53, P=.006.

Alternative presentation of final model¹²:

Not reported

Interpretation: exploratory, i.e. more research needed.

Internal validation took place; not yet externally validated.

Age, gender and history of alcohol dependence are associated with a higher risk for depression at 1-year follow-up.

Myhren, 2010

Source of data¹ and date:

prospective cohort study, Feb 2005 - Dec 2006

Setting/ number of centres and country:

Single-center; Oslo University Hospital, Ullevaal,

academic, tertiary-care centre; Norway

Funding and conflicts of interest:

Funding not reported; authors declared not to have competing interests

Recruitment method²:

Not specified; individuals approached after stay in ICU

Inclusion criteria:

• age 18-75
• at least 24h in ICU (given the range in length of ICU stay, we decided to include this study)

Exclusion criteria:

• language difficulties,
• major psychiatric

illness (i.e. psychosis)

severe head injury or

cognitive failure

• other pre-existing mental illnesses

Treatment received: one physician and one team of nurses are assigned to the patient. Physical restraint is not used. During mechanical ventilation, the patients are treated with sedatives and analgesics.

Participants: N=255 (but, 194 at 1-year follow-up)

Mean age(SD):

47.9y (15.7)

Sex: % M

160/255 (62.7%) male

Other characteristics:

Length of ICU stay, days: 12.0 (10.3 to 13.8)

SAPS score, mean (CI):

  37.0 (35.3 to 38.7)

NEMS (Nine equivalents of nursing manpower use), mean/day (CI)

29.6 (28.8 to 30.5)

Length ICU stay, mean days (CI):

12.0 (10.3 to 13.8)

Mech. Ventilation, n (%) 216 (84.7)

Duration of mech. Vent., mean days (CI)

11.0 (9.3 to 12.7)

Disease category, n (%)

  Medical 107 (42.0)

  Surgical without trauma

  62 (24.3)

  Trauma 86 (33.7)

Mild/moderate head injury/disease, n (%) 72 (28.2)

Transferred to local hospitals ICU, n (%) 132 (51.8)

Transferred while still on MV, n (%) 66 (25.9)

Life Orientation Test, mean (CI)

15.7 (15.1 to 16.3)

Describe candidate predictors³ and method and timing of measurement:

Predictor 1:

Age; pretrauma variable

Predictor 2:

Gender; pretrauma variable

Predictor 3:

Educational status; pretrauma variable

Predictor 4:

Employment status; pretrauma variable

Predictor 5:

Personality trait; administered after ICU admission

Predictor 6:

Memory or pain; administered after ICU admission

Predictor 7:

Factual recall; administered after ICU admission

Predictor 8:

Delusional memories; administered after ICU admission

Number of participants with any missing value⁴?

35/194 (18%) not included in multivariate analysis due to any missing value

How were missing data handled⁵?

• One missing item was accepted in each subscale of IES and HADS.
• Missing item was replaced with mean of other items for that patient. Although one missing item was accepted some patients did not get a sum score on these scales resulting in 180 patients with IES-total score at 12 months and 192 patients with HADS score.
• We refer to the highest cut-off score (IES ≥ 35, HADS ≥ 11) concerning symptom levels that probably needs treatment (case level).

Development

Modelling method⁶:

Multivariable logistic regression analysis, using forward Wald variable selection and adjusting for age and gender. Variables were only included if the dependent variable in the univariate analyais (p>0.2)

Performance

Calibration measures⁷ and 95%CI:

See results (only reported for PTSD)

Discrimination measures⁸ and 95%CI:

Not reported

Classification measures⁹:

See results

Evaluation

Method for testing model performance¹⁰:

No validation reported

Type of outcome:

• PTSD; dichotomous
• Anxiety; dichotomous
• Depression; dichotomous

Definition and method for measurement of outcome:

PTSD:

• Impact of Event Scale (IES); two subscales
• (7 items intrusion, 8 items avoidance), with score 0-5 per item and total range from 0 to 75
• Higher scores indicate more severe PTSD-related symptoms.
• Score >20 indicates reactions of clinical importance; score >35 indicates severe symptoms with high probability of PTSD diagnosis

Anxiety / Depression

• Hospital Anxiety and Depression scale (HADS)
• 14 items (7 anxiety, 7 depression); score 0-3 per item, total score is 21 on subscales.
• Allocated to category: score 0 to 7 = non-case; 8 to 10 = borderline case; and 11 or more = definite case.

Endpoint or duration of follow-up:

12 months

Number of events/outcomes:

• PTSD (score ≥ 35): 27% (48/180)
• Anxiety (score ≥ 11) 18% (35/192)
• Depression (score ≥11) 12% (22/192)

RESULTS

Multivariable model¹¹:

Posttraumatic stress symptoms – 1 year

Multivariate model showed a good fit to the data: Hosmer-lemeshow statistic of 4.93 of 8 degrees of freedom (P=0.77).

• Adjusted for age and sex
• Educational status: low = 0, high = 1.
• Personality trait: pessimism = low score; optimism = high score.

Anxiety – 1 year

Retrieved from text; only statistically significant associations were reported

Depression – 1 year

Retrieved from text; only statistically significant associations were reported

Alternative presentation of final model¹²:

Not reported

Interpretation: Exploratory, i.e. more research needed.

• Adjusted for age and gender, low educational level, personality trait (pessimism), memory

of pain and factual recall were independent predictors of posttraumatic stress symptoms.

• Adjusted for age and gender, employment status (unemployment higher odds), personality trait (optimism higher odds) were independent predictors of anxiety symptoms.
• Adjusted for age and gender, personality trait (optimism higher odds) and disease category (surgery higher odds) were independent predictors of depressive symptoms.

Paparrigopoulos, 2014

Source of data¹ and date:

Consecutive adult patients admitted for at least 24 h to the ICU of

a general hospital during a two-year period

Setting/ number of centres and country:

Not specified; Greece

Funding and conflicts of interest:

Funding not reported; no conflicted to declare

Recruitment method²:

308 patients were contacted by telephone after 18-24 months after ICU discharge;

• 127 (41%) deceased
• 127 (41%) could not be reached
• 6 (2%) excluded (reasons not described)
• 48 (16%) included

Incl/excl. criteria:

“unselected sample of adult patients without using any strict exclusion criteria”

Treatment received?
not reported

Participants: N=48

Mean age ± SD:

52.7±2.8

Sex: % M / % F

69% male, 31% female

Other important characteristics:

Duration of ICU stay, mean±SD, days: 13±3

Marital status:

   Single 33%

   Married 50%

   Divorced/widowed 17%

SES:

   Above average 13%

   Average 66%

   Below average 21%

Education:

   Illiterate/basic 35%

   Intermediate 38%

   Higher/Highest 27%

Describe candidate predictors³ and method and timing of measurement:

Demographic information was obtained from medical record, other information was “self-reported during a semi-structured interview”

Predictor 1:

Sex

Predictor 2:

Age

Predictor 3:

Adulthood traumatic events (as determined during the interview)

Predictor 4:

Childhood traumatic events (as determined during the interview)

Predictor 5:

Stressful Life Events (as determined during the interview)

Predictor 6:

Lifetime Psychiatric History

Predictor 7:

Social Issues

Number of participants with any missing value⁴?

0; patients included that answered phone and were eligible to participate (no reasons for exclusion described)

How were missing data handled⁵?

Not reported; not applicable as only patients included that could be reached by phone.

Development

Modelling method⁶:

Multivariate linear and  logistic regression analysis with forward conditional variable selection was implemented to confirm the observed associations between risk factors and positively screened depression and PTSD cases.

Performance

Calibration measures⁷ and 95%CI:

Not reported

Discrimination measures⁸ and 95%CI:

Not reported

Classification measures⁹:

Depression was explained by 34% of the variance in the linear analysis.

Evaluation

Method for testing model performance¹⁰:

No validation reported

Type of outcome:

Depression: dichotomous

PTSD: dichotomous

Definition and method for measurement of outcome:

Depression: Center for Epidemiologic Studies for Depression (CES-D); cut-off at 16

PTSD: 17-item Davidson Trauma Scale (DTS) was used for the assessment of PTSD Symptoms; cut-off at 40

Endpoint or duration of follow-up:

Phone interview (18-24 months after discharge)

Number of events/outcomes:

Depression: n=15, 31%

PTSD: n=12, 25%

RESULTS

Multivariable model¹¹:

logistic regression

Depression

PTSD

Linear regression

Depression

PTSD

Alternative presentation of final model¹²:

Not reported

Interpretation: exploratory, i.e. more research needed;

• Age and lifetime were predictors of depression
• Lifetime Psych. History was a predictor for PTSD

Pfoh, 2016

Source of data¹ and date:

Medical record and new assessment; enrolment 2004-2007 in primary ICAP study. ARDS patients followed until late 2012.

Setting/ number of centres and country:

13 ICU’s from 4 hospitals, Baltimore, USA

Funding and conflicts of interest:

This research was supported by the National Institutes of Health

(P050HL73994, R01HL088045, and K24HL088551) along with the Johns Hopkins Institute for Clinical and Translational Research (ICTR) (UL1 TR 000424-06). Dr. Pfoh’s time was supported by an Institutional National Research Service Award (T32HP10025B0) to Johns Hopkins School of Medicine; all authors declare that they have no conflict of interest to disclosure.

Recruitment method²:

5-year longitudinal analysis is part of Improving Care for ALI Patients (ICAP) study which recruited

mechanically ventilated patients with acute lung injury

(ARDS).

Inclusion criteria:

• ICAP participants with completed ≥1 physical assessments at 3 or 6 months of follow-up

Exclusion criteria:

• Not specified; referred to electronic supplement, not described

Treatment received?
mechanical ventilation (before discharge)

Participants: N = 193

Age, median (IQR):

49 (IQR 41–58) years

Sex: n (%) male

107 (55%) male

Other important characteristics:

APACHE II score at ICU admission:

 23 [19– 28]

Maximum daily SOFA organ failure score in ICU:

median 9 [IQR 7–11]

Dialysis in ICU:

44 (23%)

Severity of ARDS (respiratory SOFA score)

median 4 (IQR 3, 4)

Physical therapy in the ICU

100 (52%)

Mean length of stay at ICU:

Median 15 (IQR 10-23)

Describe candidate predictors³ and method and timing of measurement:

Patient baseline (pre-ARDS):

Predictor 1:

• Age

Predictor 2:

• Sex

Predictor 3:

• Charlson Comorbidity Index (CCI)

Predictor 4:

• Functional Comorbidity Index

These predictors were analysed in a model separate from the ‘other predictors’ that are described below.

Other predictors:

ICU severity of illness measures:

• Acute Physiology and Chronic Health Evaluation II score at ICU admission
• organ failure status [maximum daily Sequential Organ Failure Assessment(SOFA) score in ICU]
• acute renal failure requiring dialysis (ever vs. never)

ICU variables:

• Mean daily blood glucose level (modeled as >150 vs. ≤150 mg/dl based on prior research, along with a separate indicator for pre-existing diabetes);
• mean daily doses of

benzodiazepines (in midazolam-equivalents)

• opioids(in intravenous morphine-equivalents)
• systemic corticosteroids (prednisone equivalents + modeled as ever /never)
• coma (proportion of

ICU days with Richmond Agitation Sedation Scale score −4 or −5)

• delirium (proportion of non-comatose ICU days with a positive Confusion Assessment Method score for ICU assessment)
• durations of mechanical
• ventilation
• bed rest (see ESM)
• ICU stay

Number of participants with any missing value⁴?

six patients (3 %) were missing one or more of the physical status outcomes

How were missing data handled⁵?

• Prior to analysis, patients with a missing outcome had their data reviewed to determine if the data were missing due to due to a known decline in physical function (e.g., testing not done due to the patient being bed bound). In this case, the outcome for that time point was imputed as “declined” with 11 such imputations for strength (2% of observations), 17 (3%) for exercise capacity, and three (<1%) for physical function.
• For the subsequent assessment, imputed values were counted as missing since there was no comparison value in the previous year.
• Missing data that could not be imputed remained missing.

Development

Modelling method⁶:

• generalized linear mixed

model, with a random intercept for each patient and main effects for each follow-up time

Performance

Calibration measures⁷ and 95%CI:

Not reported

Discrimination measures⁸ and 95%CI:

Not reported

Classification measures⁹:

Not reported

Evaluation

Method for testing model performance¹⁰:

No validation reported

Type of outcome:

• Exercise capacity [evaluated using the 6-min walk test (6MWT)],  measured “activity limitation”.
• The 6MWT was performed based on American Thoracic Society guidelines [16], using a single test and the longest walking distance available, and the results are reported as percent predicted value based on established norms.
• Continuous variable (percentage)

Definition and method for measurement of outcome:

Exercise capacity is reported as the percentage of predicted value for the 6-min walk test

each patient was defined as “declined” when a comparison of his/her current and prior score revealed a decrease

that was greater than the Reliable Change Index (RCI )at each follow-up

Endpoint or duration of follow-up:

5-year follow-up period

Number of events/outcomes:

103 (53 %) had ≥1 decline(s) in exercise capacity (6MWT),

RESULTS

Multivariable model¹¹:

Exercise capacity – 5-year follow-up (n=193)

Alternative presentation of final model¹²:

Not reported

Interpretation:

• exploratory, i.e. more research needed
• Age and pre-ARDS comorbidity (Charlson index) were independently associated with decline in exercise capacity defined by percentage of expected results of the 6MWT.

Study reference

(first author, year of publication)

Study participation

Study sample represents the population of interest on key characteristics?

(high/moderate/low risk of selection bias)

Study Attrition

Loss to follow-up not associated with key characteristics (i.e., the study data adequately represent the sample)?

(high/moderate/low risk of attrition bias)

Prognostic factor measurement

Was the PF of interest defined and adequately measured?

(high/moderate/low risk of measurement bias related to PF)

Outcome measurement

Was the outcome of interest defined and adequately measured?

(high/moderate/low risk of measurement bias related to outcome)

Study confounding

Important potential confounders are appropriately accounted for?

(high/moderate/low risk of bias due to confounding)

Statistical Analysis and Reporting

Statistical analysis appropriate for the design of the study?

(high/moderate/low risk of bias due to statistical analysis)

Asimakopoulou, 2014

High risk

Sample is inadequate representation of whole group of patients; only 5,1% of approached patients included in study

Low risk

Missing data is not reported.

High risk

Predictor disability was not clearly described. Method and timing is unclear.

Low risk

Outcome defined (depression and PTSD) and adequately measured

Low risk

Relevant analysis age and gender adjusted

Low risk

Appropriate analysis; logistic regressions

Davydow, 2013

High risk

Only 150 of eligible patients (1183) participated.

Moderate risk

“Patients who did not complete follow-up

were more likely to have substantial acute stress symptoms during their hospitalization”

Low risk

Low risk

Low risk

Low risk

Hopkins, 2010

Moderate risk

All patients participate an RCT

Low risk

“Of the 74 ARDS survivors enrolled in the study, three died in the first year following hospital discharge from: pulmonary fibrosis/cor pulmonale, liver failure or diabetic complications. Five survivors declined to return for 1-year follow up (e.g., busy schedules or not interested), which resulted

in 66 survivors who completed the 1-year evaluation. Two ARDS survivors died in the second year (bowel obstruction

or cardiac failure), and two declined to return for the

2-year follow-up, resulting in 62 survivors who completed

the 2-year evaluation.”

Low risk

Defined (“history of alcohol dependence, history

of smoking, gender, ratio of PaO2 to FiO2, hours of

mechanical ventilation, APACHE II score and presence of cognitive sequelae at discharge and 1 year after discharge”)

and adequately measured.

Low risk

Outcome defined (depression and anxiety) and adequately measured

Low risk

Relevant analysis age and gender adjusted

Low risk

Appropriate analysis, stepwise multiple regressions; and bootstrap to internally validate model

Myhren, 2010

Moderate risk

Source and dates are not adequately described

Moderate risk

Lost- to follow-up group differs in variable of interest at baseline.

“Patients that were lost to follow up (n = 61) scored

significantly higher on HADS-Anxiety at baseline compared

with those who completed follow up (6.6 vs. 5.3,

P = 0.041), but not significantly different on HADS Depression

(5.5 vs. 4.5, P = 0.116) or IES-total (25.0 vs. 21.8, P = 0.207).”

Low risk

PFs defined and adequately described.

Low risk

Outcome defined and adequately described.

Low risk

All analyses age and gender adjusted

Low risk

Appropriate analysis; logistic regressions, age/gender adjusted

Paparrigopoulos, 2014

High risk

Sample is inadequate representation of whole group of patients; only 16% of approached patients included in study (41% had died, 41% could not be reached and 2% was excluded).

Low risk

Not applicable as patients were approached at moment of follow-up measurement, 18-24 months after ICU discharge.

Low risk

Defined (sex, age, adulthood/childhood traumatic events, stressful life events, lifetime psychiatric history, social issues) and adequately measured.

Low risk

Outcome defined and adequately described.

Low risk

Relevant analysis age and gender adjusted

Low risk

Appropriate analysis; logistic regressions

Pfoh, 2016

Moderate risk

Patients were included from another trial and were only included if completed 1 or more physical assessment at 3 or 6 months of follow-up. The included group was younger and healthier (lower Charlson comorbidity score) compared to patients that were not included (died or ineligible for evaluation).

Low risk

Missing data imputed as ‘declined’ if retrieved why data was not present; otherwise remained missing. Per outcome 1-3% of observations remained missing.

Low risk

Defined (age per decade, male, functional comorbidity

Index, Charlson comorbidity

index) and adequately measured.

Low risk

Outcome defined and adequately described.

Low risk

Relevant analysis age and gender adjusted

Low risk

Appropriate analysis; generalized linear mixed model