We conducted a prospective longitudinal cohort study of 117 survivors of ARDS, recruited from four academic medical–surgical ICUs in Toronto, Canada between May 1998 and May 2001 (2). Patients were eligible for inclusion if they were at least 16 yr of age, had a Pa_O2:Fi_O2 ratio of 200 or less while receiving mechanical ventilation with a positive end-expiratory pressure of at least 5 cm of water, had evidence of airspace changes in all four quadrants on frontal chest radiograph, and an identifiable risk factor for ARDS. Patients were excluded if they were immobile before being admitted to the ICU, had a history of pulmonary resection, or had documented neurologic or psychiatric disease. Consent was obtained for 1-yr follow-up from patients' surrogate decision makers in the ICU and then directly from patients at ICU discharge. At the end of 1 yr, consent was obtained from patients for another 4 yr of follow-up. Two-year follow-up was completed in May 2003. Research ethics board approval was obtained at all participating hospitals.

We collected baseline demographic characteristics and measures of severity of illness in the ICU (2). These measures included the Acute Physiology, Age, and Chronic Health Evaluation (APACHE II; 0 to 71) (11), the Multiple Organ Dysfunction Score (MODS) (0 to 24) (12), and the modified Lung Injury Score (LIS) daily the first week and twice weekly thereafter (0 to 4) (13). Higher scores indicate more severe illness. The rate of change of LIS (LIS slope) is an indicator of the rate of resolution of lung injury; a greater LIS slope indicates slower recovery.

We also collected clinical and health-related quality of life data at scheduled study visits 3, 6, 12, 18, and 24 mo after discharge from the ICU. Each visit included an interview, physical examination, measures of pulmonary function, resting oximetry, the 6-min walk test (14–16) with continuous oximetry, and the SF-36 (17). The SF-36 includes eight domains of health-related quality of life: physical functioning, social functioning, physical role, emotional role, mental health, pain, vitality, and general health. Scores for each domain can range from 0 (worst) to 100 (best). We compared their scores with those of an age- and sex-matched control population.

We prospectively collected health care use data from administrative databases, hospital and outpatient clinical records, patients' monthly diaries, and self-reports at scheduled study visits. This approach provided patient-specific and activity-based resource use per day in the ICU, on the wards after discharge from ICU, and after hospital discharge. For health care costs, we took a third-party payer perspective (Ontario Ministry of Health) to estimate direct medical costs (including hospitalization costs, emergency room visit costs, professional fees, outpatient visit costs, drug costs, laboratory and radiologic procedural costs, outpatient and inpatient rehabilitation costs, home care costs, and chronic care hospitalization costs). We excluded indirect medical costs and direct nonmedical costs. Costs were adjusted for inflation rates, using the medical component of the Canadian Consumer Price Index (18), and were expressed in 2002 Canadian dollars. Our methodology for determining health care use and costs is detailed in appendix e1 (see the online supplement).

Details of the statistical analyses can be found in appendix e2. In brief, we summarized descriptive data on clinical outcomes, health care use, and costs. We used t tests to examine the changes in measures of pulmonary function and distance walked in 6 min, and the signed-rank test to assess within-subject changes in SF-36 scores between 1 and 2 yr after discharge from the index hospitalization (19). For missing SF-36 data, we used last observation carried forward for our analyses.

Using backward selection according to the Akaike's information criterion (AIC) (20), we performed multivariable regression analyses to examine how patients' baseline characteristics and time-dependent variables in the ICU influenced health care costs in the ICU, on the wards and from hospital discharge to 2 yr. To test the sensitivity of our model to whether patients were alive or dead, we repeated our analyses including only patients alive at 2 yr. We summarized overall model fits with R² and used Efron's enhanced bootstrap validation procedure to assess overfitting. For ease of interpretation, the results in Table 4 are the exponentiated parameters, which represent the adjusted ratio of mean costs associated with the corresponding predictor. A ratio greater than 1 indicates that costs associated with the variable are higher.

One hundred and seventeen eligible patients consented to participate in the ICU; of those, 109 consented for 1-yr follow-up at discharge from ICU (Figure 1). The eight survivors of ARDS who did not consent to follow-up after ICU discharge did consent to evaluation of their ICU records and cost data. Our 2-yr follow-up captured 85% of eligible patients for clinical data and 96% for economic data. Among patients for whom follow-up data were available, 86 of 98 (88%) survived to 1 yr and 78 of 92 (85%) survived to 2 yr. Because ICU mortality for our cohort was 40% (2), overall 2-yr mortality from diagnosis to 2 yr after ICU discharge was 49%. Most deaths occurred during the first 6 mo after discharge from the ICU and were related to preexisting medical problems. The median age of patients with ARDS who survived to be discharged from the ICU was 45 yr, and 56% were male (Table 1). They were severely ill (median APACHE II score of 23, median maximal LIS of 3.7, and median MODS of 9 on day of eligibility). Twelve percent of patients (14 of 117) required renal replacement therapy in the ICU; only one continued to undergo dialysis after discharge from the ICU because of preexisting end-stage renal disease. Our cohort spent a median of 25 d in the ICU and 48 d in the acute care hospital.

Figure 1. Flow chart of cohort of survivors of acute respiratory distress syndrome. *The number in the cohort includes those who died, but excludes those who are lost to follow-up and those who declined follow-up.

Download Figure | Download Powerpoint

[More] [Minimize]

Exercise capacity of survivors of ARDS, measured by the 6-min walk test, did not significantly improve from 1 to 2 yr after ICU discharge and continued to be lower than normal, based on an age- and sex-matched control population (Table 2). Patients attributed exercise limitation to muscle weakness, although self-reported muscle weakness was less pronounced at 2 yr compared with at 1 yr. Apart from emotional role and mental health domains, all other domains of health-related quality of life as measured by SF-36 remained below those of the normal population. Although the absolute median scores for physical functioning, general health, and social functioning increased from 1 to 2 yr, none reached statistical significance. There was a trend toward better physical role at 2 yr compared with 1 yr (p = 0.0586), but this was not reflected in the distance walked in 6 min, which did not improve over time. There were no significant changes in measures of pulmonary function. Median lung volumes and spirometry remained within the predicted normal range; the median carbon monoxide diffusion capacity in the cohort rose slightly to within the normal range but these changes did not reach statistical significance. Two additional patients had arterial oxygen desaturation below 88% during the 6-min walk test at 2 yr (7 of 62 at 2 yr vs. 5 of 78 at 1 yr), but neither required supplemental oxygen at rest or on exertion. The proportion of patients who returned to their baseline body weight was stable from 1 yr (71%) to 2 yr (70%).

Eleven more survivors of ARDS returned to work at 2 yr compared with at 1 yr (65 vs. 49%). Most patients returned to their original job. Those who had not returned to work cited the following reasons: depression, post-traumatic stress disorder, muscle weakness and fatigue, short-term memory loss from closed head injury, and disability secondary to orthopedic injuries. All patients, except one, were ambulatory and lived independently in the community; one patient developed progressive neurologic dysfunction after ICU discharge and was cared for in a retirement home.

Thirty-one of the 109 survivors of ARDS (28%) were transferred to community hospitals: 10 (9%) to another ICU and 21 (19%) to another hospital ward. Once discharged from the acute care hospital, 36 (33%) patients received inpatient rehabilitation, with a median length of stay of 36 d (mean, 52 d). Of those discharged home, 52 (51%) received formal home care support services. During the 2 yr of follow-up, the median number of home care visits for our cohort was 28 (mean, 74 visits), of which 27 (mean, 55 visits) were in the first year; most were for nursing care. Over the 2-yr period, our cohort had a median of 7 (IQR, 2 to 15) visits to their primary care physicians and a median of 7 (IQR, 1 to 15) visits to specialists, with more visits occurring in the first year than in the second year.

Forty-three patients (39%) were readmitted to hospital in the first 2 yr after ICU discharge. Twenty-two patients (20%) were admitted two or more times for recurring medical problems related to their risk factor for ARDS or complications of their prolonged, complex critical illness. The reasons for readmission were as follows: (1) colostomy reversal, incisional hernia repair, bowel obstruction or anastomotic leak after pancreatitis or complex intraabdominal sepsis (nine patients); (2) treatment for infection, line sepsis, or rejection in oncology or transplant patients (three patients); (3) orthopedic procedures or skin grafting in trauma and/or burn patients (four patients); and (4) inpatient psychiatry for depression and posttraumatic stress disorder (two patients). One patient was admitted for complications directly related to her necrotizing pneumococcal pneumonia, ARDS, and severe organ dysfunction. She was admitted eight times for recurrent lung infections and hemoptysis secondary to bronchiectatic disease resulting from her severe pneumonia and for amputation of bilateral gangrenous toes. One patient had admissions for debridement of heterotopic ossification. One patient, who was initially admitted for mitral valve replacement after endocarditis, sepsis, and ARDS, had a second hospital admission to deliver a term baby. The final patient had admissions for extraction of all his teeth secondary to severe dental caries. The remaining 21 patients (19%) had a single, isolated admission. Five of these patients had admissions similar to those listed above: hernia repair (two patients), reversal of colostomy (two patients), and Hickman line sepsis (one patient); the rest of the patients had admissions unrelated to the episode of ARDS or critical illness. There were a total of 62 readmissions in the first year and 39 in the second year. The mean number of days of subsequent hospitalization was 4.2 ± 15.9 d in the first year and 2.4 ± 8.3 d in the second year.

The mean health care costs for a survivor of ARDS from ICU admission to 2 yr post-ICU discharge are shown in Table 3 and Figure 2. The largest portion of this total cost was the initial hospital stay (mean, Can $128,860; 95% confidence interval, Can $111,970–$151,190), with ICU costs accounting for 76% of these hospital costs. Nursing costs dominated total initial hospital costs, accounting for 75% of total ICU costs and 66% of total ward costs (Figure 2). Other significant cost items in the ICU include physician costs, costs of laboratory examinations and blood processing, imaging costs, and pharmacy costs. On the wards, pharmacy costs were the second most important cost item after nursing costs. The eight survivors of ARDS who did not consent to follow-up after ICU discharge had higher ICU costs than those who did consent (data not shown).

Figure 2. Distribution of health care costs from intensive care unit (ICU) admission to 2 yr after ICU discharge for survivors of acute respiratory distress syndrome.

Download Figure | Download Powerpoint

[More] [Minimize]

The mean costs after discharge from the index hospitalization up to 2 yr were Can $28,350 (95% confidence interval, Can $20,580–$38,350), with subsequent hospitalization costs and inpatient rehabilitation costs predominating. Home care costs, outpatient pharmacy costs, and physician costs were the other significant costs. Nursing was the major component of home care costs. Mean total cost after index hospitalization was Can $16,200 in the first year and Can $12,100 in the second year.

In the ICU, patients with more organ dysfunction acquired during their ICU stay, slower resolution of lung injury, burns, and informal caregivers had higher costs (unadjusted model R² = 0.50, bootstrap-corrected R² = 0.44; Table 4). On the wards, those who had more premorbid organ dysfunction, slower resolution of lung injury in the ICU, and who were obese (body mass index ≥ 30) had higher costs (unadjusted model R² = 0.26, bootstrap-corrected R² = 0.25). From hospital discharge to 2 yr, patients who were older and had acquired more organ dysfunction during their ICU stay had higher costs (unadjusted model R² = 0.13, bootstrap-corrected R² = 0.06) (determinants of health care costs from hospital discharge to 1 yr were the same). When we reanalyzed our data including only those who were alive at 2 yr, we found that the determinants of health care costs were similar (see Table E2 in the online supplement).