Comparison of community-acquired, hospital-acquired, and intensive care unit-acquired acute respiratory distress syndrome: a prospective observational cohort study

Introduction Acute respiratory distress syndrome (ARDS) is a syndrome characterized by diffuse pulmonary edema and severe hypoxemia that usually occurs after an injury such as sepsis, aspiration and pneumonia. Little is known about the relation between the setting where the syndrome developed and outcomes in ARDS patients. Methods This is a 1-year prospective observational study conducted at a tertiary referred hospital. ARDS was defined by the Berlin criteria. Community-acquired ARDS, hospital-acquired ARDS and intensive care unit (ICU)-acquired ARDS were defined as ARDS occurring within 48 hours of hospital or ICU admission, more than 48 hours after hospital admission and ICU admission. The primary and secondary outcomes were short- and long- term mortality rates and ventilator-free and ICU-free days. Results Of the 3002 patients screened, 296 patients had a diagnosis of ARDS, including 70 (23.7 %) with community-acquired ARDS, 83 (28 %) with hospital-acquired ARDS, and 143 (48.3 %) with ICU-acquired ARDS. The overall ICU mortality rate was not significantly different in mild, moderate and severe ARDS (50 %, 50 % and 56 %, p = 0.25). The baseline characteristics were similar other than lower rate of liver disease and metastatic malignancy in community-acquired ARDS than in hospital-acquired and ICU-acquired ARDS. A multiple logistic regression analysis indicated that age, sequential organ function assessment score and community-acquired ARDS were independently associated with hospital mortality. For community-acquired, hospital-acquired and ICU-acquired ARDS, ICU mortality rates were 37 % 61 % and 52 %; hospital mortality rates were 49 %, 74 % and 68 %. The ICU and hospital mortality rates of community-acquired ARDS were significantly lower than hospital-acquired and ICU-acquired ARDS (p = 0.001 and p = 0.001). The number of ventilator-free days was significantly lower in ICU-acquired ARDS than in community-acquired and hospital-acquired ARDS (11 ± 9, 16 ± 9, and 14 ± 10 days, p = 0.001). The number of ICU-free days was significantly higher in community-acquired ARDS than in hospital-acquired and ICU-acquired ARDS (8 ± 10, 4 ± 8, and 3 ± 6 days, p = 0.001). Conclusions Community-acquired ARDS have lower short- and long-term mortality rates than hospital-acquired or ICU-acquired ARDS.


Introduction
Acute respiratory distress syndrome (ARDS) is a significantly heterogeneous syndrome that involves many different groups of patients that may influence outcomes [1][2][3]. The Berlin definition classifies ARDS into mild, moderate, and severe by hypoxemia severity, using the PaO 2 /FiO 2 ratio [4]. The predictive validity for mortality according to the Berlin definition has not been validated in recent studies [5][6][7]. The absolute predictive value is modest and suggests that some factors other than hypoxemia need to be investigated [8].
Differences in mortality rates have been demonstrated for patients with community-acquired pneumonia and hospital-acquired pneumonia [9]. For ARDS patients, a majority of patients developed acute lung injury (ALI)/ ARDS within the first five days, especially 48-72 h after admission [10]. A retrospective cohort study showed a trend of decreasing prevalence in hospital-acquired and intensive care unit (ICU)-acquired ARDS patients over an eight-year period [11]. A prospective, multi-center, observational study revealed that late-onset ALI/ARDS patients had longer ICU and hospital stays than earlyonset ALI/ARDS patients, but the mortality rate was not significantly different [12].
Little is known of the characteristics of patients in different ARDS subgroups according to the setting where the syndrome developed. The aim of this study is to investigate the outcomes of community-acquired, hospital-acquired and ICU-acquired ARDS patients. It has not been studied in the literature before and is a newly thought of patient population for ARDS. In addition to the severity of ARDS, the different categories may be an important factor for outcomes and, therefore, for clinical trials assessing the effects of potential interventions.

Study design and study population
A prospective observational study was conducted from September 2012 to August 2013 at Chang Gung Memorial Hospital, a tertiary care referral center with 3,700 ward beds and 278 adult ICU beds. These 278 adult ICU beds are distributed in 17 ICUs (nine medical ICUs, seven surgical ICUs, one burn ICU). All of the admitted patients with invasive mechanical ventilation were screened for eligibility by the Hospital Information System. Eligible patients were further evaluated when both chest-X-ray and PaO 2 /FiO 2 ratio criteria were present concurrently. Patients were included if they met the criteria of the Berlin definition of ARDS and those with ARDS were further reviewed by a second independent investigator (LC Chiu) blind to the previous screening results. Differences were resolved by discussion between the two senior intensive care physicians (KC Kao and HC Hu).
Patients were excluded if they were younger than 18 years old or had been admitted to another hospital and referred for admission. The Chang Gung Memorial Hospital's Institutional Review Board Ethics Committee approved the study protocol and informed consent was waived (CGMH IRB No.102-1729B).

Definitions
The Berlin definition of ARDS includes: (1) onset within one week of a known clinical insult or new or worsening respiratory symptoms; (2) bilateral opacities not fully explained by effusions, lobar/lung collapse, or nodules; (3) respiratory failure not fully explained by cardiac failure or fluid overload and needs objective assessment (e.g., echocardiography) to exclude hydrostatic edema if no risk factor is present; and (4) PaO 2 /FiO 2 ratio ≤300 mm Hg with positive-end expiratory pressure (PEEP) or continuous positive airway pressure (CPAP) ≥ 5 cm H 2 O [4].
According to the setting where the ARDS syndrome developed, we categorized ARDS patients into communityacquired, hospital-acquired and ICU-acquired ARDS. Community-acquired ARDS was defined as ARDS occurring upon admission or within 48 hours of hospital or ICU admission. Hospital-acquired ARDS was defined as ARDS occurring > 48 hours after hospital admission. ICUacquired ARDS was defined as ARDS occurring > 48 hours after ICU admission (11). All the ARDS patients had the known clinical insults or worsening respiratory symptoms within one week.

General management
It was recommended that patients be ventilated with protective ventilation using low tidal volume 4-8 mL/kg of predicted body weight plus moderate to high levels of PEEP for volume-controlled or pressure-controlled ventilation [13]. The predicted body weight of male patients was calculated as equal to 50 + 0.91 (height in centimeters -152.4) and of female patients as equal to 45.5 + 0.91 (height in centimeters -152.4). Ventilation was monitored by arterial blood gas measurements, with ventilator settings changed as needed. Pulse oximetry (SpO 2 ) was used to monitor oxygenation and the FiO 2 was adjusted to maintain SpO 2 > 90 % or PaO 2 > 60 mm Hg and to avoid raising the peak inspiratory pressure > 35 cm H 2 O.
The general medical management including fluid replacement, the use of antibiotics, corticosteroids in some selected patients and vasopressor agents, sedation with infusions of midazolam and paralysis with infusions of cisatracurium was directed by the intensivists-incharge. The patients had a peripheral arterial line and PiCCO plus monitoring (version 5.2.2; Pulsion Medical System AG, Munich, Germany) for hemodynamic monitoring if indicated.

Data collection
Demographics and baseline clinical information were collected on enrollment. Data on patient outcome 90 days after inclusion were tracked. The following data were recorded upon ICU admission: date of hospital and ICU admission, age, gender, body weight and height, underlying disease, and risk factors of ARDS. The mechanical ventilator settings (i.e., artery blood gas, tidal volume, lowest PaO 2 /FiO 2 ratio with the highest PEEP, and peak airway pressure) were recorded during mechanical ventilation at the time of ARDS diagnosis. The Charlson comorbidity index (CCI) [14], Acute Physiology and Chronic Health Evaluation (APACHE) II score [15], Sequential Organ Failure Assessment (SOFA) score [16], Multiple Organ Dysfunction (MOD) score [17], and Lung Injury score (LIS) [18] were recorded on the day of inclusion and on days 3, 7, and 14 after inclusion. Regarding the definitions of liver disease in CCI, mild liver disease was chronic hepatitis or cirrhosis without portal hypertension. Moderate and severe liver diseases were cirrhosis combined with portal hypertension without and with history of varices bleeding [14].
The primary outcome was mortality (in ICU, at 28 days, at 60 days, and at 90 days, in hospital) and the secondary outcomes were ICU-free days and ventilator-free days. The number of ventilator-free days or ICU-free days was the mean number of days from day 1 to day 28 on which the patient had been breathing without assistance for at least 48 consecutive hours or which the patient had been transferred to ward from ICU. Patients who did not survive to 28 days were assigned zero ventilator-free days and zero ICU-free days.

Statistical analysis
Descriptive statistics were expressed as mean ± SD (standard deviation). All variables were tested for normal distributions using the Kolmogorov-Smirnov test. Student t test was used to compare the means of continuous variables with normal distribution and the Mann-Whitney U test for the rest. Categorical data were compared using the chi square test. Risk factors for hospital mortality were analyzed using univariate analysis, and the variables statistically significant (p < 0.05) in the univariate analysis were included in the multivariate analysis by applying a multiple logistic regression based on backward elimination of data. The Hosmer-Lemeshow goodness-of-fit test was used for calibration when evaluating the number of observed and predicted deaths in risk groups for the death probabilities. Cumulative survival curves as a function of time were generated using the Kaplan-Meier approach and compared using the log-rank test. All statistical tests were two-tailed and p < 0.05 was considered statistically significant. All statistical analyses were performed using the SPSS (SPSS for Windows, SPSS Inc., Chicago, IL, USA) statistical package.

Results
During the study period, 3,002 admitted patients with invasive mechanical ventilation were screened ( Fig. 1). After excluding 2,664 patients who did not meet the ARDS criteria and 42 ARDS patients referred from other hospitals, 296 ARDS patients were included for analysis. Of these 296 ARDS patients, 24 patients underwent adjudication to have a definitive diagnosis of ARDS. The most common was ICU-acquired ARDS at 48.3 %, followed by hospital-acquired ARDS at 28 %, and community-acquired ARDS at23.7 %.
The demographic and clinical characteristics of different types of ARDS patients are shown in Table 1. Comparison among the three diagnosis groups revealed no statistically significant differences regarding age, gender, body weight, severity, causes, ventilation, oxygenation, and initial mechanical ventilation setting. Of the community-acquired ARDS patients, nine (12.9 %) were classified as mild ARDS, 24 (34.3 %) as moderate ARDS, and 37 (52.9 %) as severe ARDS. Of the hospital-acquired ARDS patients, 14 (16.9 %) were classified as mild, 31 (37.3 %) as moderate, and 38 (45.8 %) as severe ARDS. Of the ICU-acquired ARDS patients, 28 (19.6 %) were classified as mild, 54 (37.8 %) as moderate, and 61 (42.7 %) as severe ARDS. The distribution of severity in these three groups was not statistically significant (p = 0.65).
Prone position was applied to three patients with community-acquired ARDS, five with hospital-acquired ARDS, and three with ICU-acquired ARDS. Venovenous extra-corporeal membrane oxygenation was used in three patients with community-acquired ARDS, four with hospital-acquired ARDS, and three with ICUacquired ARDS.

Discussion
This one-year prospective observational study found that community-acquired ARDS patients had better shortand long-term mortality rates than hospital-acquired and ICU-acquired ARDS patients. Community-acquired ARDS patients had less underlying moderate to severe liver disease than hospital-acquired and ICU-acquired ARDS patients. The community-acquired ARDS patients had more ICU-free days than hospital-acquired or ICUacquired ARDS patients and the ICU-acquired ARDS patients had fewer ventilator-free days than communityacquired or hospital-acquired ARDS patients.
The incidence of ARDS varies widely. Differences in demographics, healthcare systems, and definitions may account for different incidences of ARDS in different areas or countries. In adult population-based studies, the incidence of ARDS by the American-European consensus (AECC) definition ranged from 5-7.2 cases/100,000/ year in Europe to 33.8/100,000/year in USA [11,[19][20][21]. For ICU patients, the reported incidence of ALI/ARDS For critically-ill patients, the lead time of events is an important prognostic factor, but the correlation between the time of disease onset and outcomes is controversial. Late-onset septic shock (>24 h after ICU admission) was associated with a higher mortality rate but not significant in patients with early-onset septic shock (<24 h after ICU admission) (88 % vs. 63 %, p = 0.071) [25]. There was no difference in the mortality rates of patients with sepsis on ICU admission, those who developed sepsis within 48 h after ICU admission, and those who developed sepsis > 48 h after ICU admission (27 % vs. 20 % vs. 28 %, p = 0.526) in the Sepsis Occurrence in Acutely Ill Patients (SOAP) study [26]. However, in a multivariate analysis of patients with shock from the SOAP database, late-onset (>48 h after ICU admission) shock was an independent predictor of higher ICU mortality (odds ratio, 2.6; 95 % CI, 1.6-4.3; p < 0.001) [27]. For renal failure, later onset was associated with worse prognosis than early onset in ICU patients with mechanical ventilation [28][29][30]. For post-traumatic ARDS patients, there was no difference in mortality rates between the early (within 48 h of hospital admission) and late (>48 h of hospital admission) groups (27 % vs. 21 %) [31]. Sub-analysis of the SOAP database demonstrated no significant difference in ICU mortality rates between early-and late-onset ARDS (45.7 % vs. 35.5 %) [12].
The overall ICU mortality rate was not significantly different in mild, moderate, and severe ARDS (50 %, 50 %, and 56 %, p = 0.25) in this study. The previous study in ARDS patients found that the presence of serious comorbidities, such as acquired immune deficiency syndrome, metastatic cancer, immunocompromised, liver cirrhosis, and hepatic failure, almost tripled the mortality rate after hospital discharge compared to   [32]. The present study revealed that community-acquired ARDS patients had a significantly lower ICU mortality rate than hospital-acquired and ICU-acquired ARDS patients. Further analysis of the comorbidities showed that moderate to severe liver disease is less prevalent in the community-acquired ARDS patients than hospital-acquired and ICU-acquired ARDS patients (0 %, 19.3 %, and 21.7 %, p = 0.000). More patients with severe liver dysfunction may contribute to the high mortality rates in hospital-acquired and ICU-acquired ARDS patients than community-acquired ARDS patients. The patients with underlying severe chronic liver disease were always excluded in the most randomized controlled trials of ARDSnet [13,[33][34][35][36][37]. It is difficult to determine whether comorbidities with advanced liver disease or the severity of ARDS most influence the survival outcomes. The comorbidities could affect survival in ARDS patients and this needs further investigation. For ARDS patients, there are many conditions associated with mortality. ARDS is a syndrome that is not, by itself, a cause of mortality but rather accompanies other disease processes. The cause of ARDS such as pneumonia, aspiration or sepsis, severity of ARDS, underlying comorbidities, and cause of death may relate to the mortality. In this study, the cause and severity of ARDS were not significantly different between the three categories. In addition to moderate to severe liver disease, more patients had underlying metastatic solid tumors in hospital-acquired and ICU-acquired ARDS than community-acquired ARDS. Given the poor prognosis of cancer patients with ARDS, it may explain, at least in part, the poor outcome in patients with hospitalacquired and ICU-acquired ARDS compared to community-acquired ARDS. Furthermore, it is possible that hospital-acquired and ICU-acquired ARDS may have worse prognosis due to the presence of hospital exposures that are known to increase the risk for  hospital-acquired ARDS, such as aspiration, nonprotective ventilation, inadequate antibiotics and source control, surgical or medical errors, and too much fluid and transfusion [38].
In this study, ICU-acquired ARDS patients had fewer ventilator-free days than the other two types of ARDS patients and had fewer ICU-free days than communityacquired patients ( Table 2). These patients had already been treated in the ICU for more than two days due to previous insults or poor conditions before ARDS occurred. After suffering from ARDS, the longer ventilator use and ICU stay might be due to the fragile characteristics of these ICU patients. The medical cost in the ICU would be effectively reduced if we could recognize ARDS early and prevent its occurrence in these ICU patients.
Although some randomized controlled trials have reported improvements in mortality, the current overall hospital mortality is about 40-50 % in most series of ARDS patients [20,39]. The different reported mortality rates in ARDS patients may be due to differences in patient selection, associated underlying diseases, predisposing or risk factors for ARDS, severity of hypoxemia, and setting of mechanical ventilation. For example, trauma-induced ARDS had favorable prognosis, with approximately 10 % 60-day mortality rate [40,41]. However, the hospital mortality rate was higher at 68.8 % in cancer patients with ARDS [42]. In the present observational cohort study, the ICU mortality varies from 37 % to 61 % in three different types of ARDS. Other than the presence of malignancies and a high prevalence of liver failure, one of the reasons for the high mortality rate could be the use of high tidal volume (>9 ml/kg PBW) in a substantial proportion of this cohort ( Table 1). As a consequence of high mortality and considerable variability in outcomes in ARDS patients, classification of phenotype for mortality is essential for predicting prognosis, guiding clinical decision-making, and designing prospective randomized controlled trials.
There are some limitations in this study. First, misdiagnosis of ARDS is a potential limitation derived from reliance on available diagnostic criteria. To diminish this possible bias, the diagnosis of ARDS was based on established criteria and the accuracy of diagnosis was verified through a separate case review by three independent intensive care physicians. Second, the study design limited patients to those who required invasive mechanical ventilation to identify those most at risk of subsequent mortality, thereby losing patients who met the criteria for ARDS but only received non-invasive ventilation. Third, a low tidal volume strategy was not fully applied in all of the ARDS patients and ventilator-induced lung injury might have contributed to mortality. However, this would have equally exerted an impact on all three types of ARDS patients and should not influence the results. Fourth, different management strategies may alter the ARDS-related outcomes, but this study does not identify the correlation between management strategies and mortality. Lastly, perhaps the main limitation is the single-center nature of the study population. Although strengthened by the prospective, cohort, and observational design, this may limit the generalization of the study results and warrant external validation.

Conclusions
In this study, the overall ICU mortality rate was not significantly different in mild, moderate, and severe ARDS. According to the timing of onset and admission source, three types of patients with ARDS are classified as community-acquired, hospital-acquired, and ICUacquired ARDS. The most common is ICU-acquired ARDS with 48.3 %, followed by hospital-acquired ARDS with 28 %, and community-acquired ARDS with 23.7 %. Community-acquired ARDS patients have better shortand long-term mortality rates than hospital-acquired or ICU-acquired ARDS patients. Underlying advanced liver disease may contribute to the different outcomes between these three groups. Patients with ICU-acquired ARDS have lower numbers of ventilator-free days and ICU-free days than those with community-acquired or hospital-acquired ARDS. These data provide relevant information on ARDS patients for evaluating individual outcomes and designing clinical trials.

Key messages
In this prospective study, according to the timing of onset and admission source, the most commonly occurring ARDS is ICU-acquired ARDS with 48.3 %, followed by hospital-acquired ARDS with 28 %, and community-acquired ARDS with 23.7 %. Community-acquired ARDS patients have better short-and long-term mortality rates than hospitalacquired or ICU-acquired ARDS patients. ICU-acquired ARDS patients have lower numbers of ventilator-free days and ICU-free days than those with community-acquired or hospital-acquired ARDS. In terms of different outcomes, classification according to the setting where the ARDS developed may be considered in designing clinical therapeutic trials in the future.