Despite numerous randomized controlled trials, a lung protective strategy during mechanical ventilation remains the only therapy shown to improve survival in patients with established ALI [³,⁴,^10-12]. According to the “two-hit” model of ALI development before ALI becomes clinically apparent, a pre-ALI state exists following the first injury to the lungs. The United States Critical Illness and Injury Trials Group-Lung Injury Prevention Study (USCIITG-LIPS 1) investigators focused on defining patient characteristics that would allow us to identify these patients before overt ALI develops. Interventions delivered in this early phase of care can offer potential prevention of ALI. As most clinical studies in ALI have primarily focused on mechanically ventilated patients, insight into a potentially preventable phase of ALI prior to its development is currently lacking [¹³,¹⁴]. Data suggest that ALI is rarely present at the time of initial Emergency Department (ED) evaluation; however, a search in top emergency medicine journals yielded many case reports of patients presenting with ALI [^15-18]. In reality, many ED patients may have unrecognized ALI and possess many predisposing conditions for ALI development. Recently, a National Heart Lung Blood Institute Workshop Report on future clinical research in ALI recommended to continue the development of strategies to perform ALI prevention trials and observational studies of patients without ALI undergoing prolonged mechanical ventilation [¹⁹]. Following the paradigm of trauma team care for major trauma, activation of the cardiac catheterization laboratory team for ST-elevation myocardial infarction and acute stroke teams for ischemic stroke, and early goal-directed therapy for severe sepsis, clinical benefit may be derived from early identification of and preventative interventions for patients at risk of developing ALI.

We evaluated the frequency of ALI development in at-risk hospitalized ED patients among the study sites, described the predisposing conditions and risk modifiers of ALI development, and determined the attribution of ALI to hospital mortality.

This is a subgroup analysis of data from a multicenter, observational cohort study, the United States Critical Injury and Illness Trials Group-Lung Injury Prevention Study 1 (USCIITG-LIPS 1) [²⁰,²¹]. All participating study sites received approval from their respective local institutional review board. The study flow diagram is illustrated in Additional file 1: Appendix 1.

From March through August 2009, 22 centers (20 American and 2 Turkish hospitals) enrolled patients with at least one ALI predisposition admitted from the ED. Patients were enrolled prospectively at 19 study sites and retrospectively at 3 sites.

Consecutive adult ED patients admitted to academic and community acute care hospitals were eligible for the study if they presented with one or more a priori defined conditions predisposing to ALI (shock, aspiration, sepsis, pancreatitis, pneumonia, high-risk trauma: traumatic brain injury, smoke inhalation, near drowning, lung contusion, multiple fractures; high-risk surgery: thoracic, spine, acute abdomen, cardiac, aortic vascular; and emergency surgery). Patients were excluded when ALI was present at initial assessment, if they were transferred from an in-patient setting, died in the ED, admitted for comfort or hospice care, or re-admitted during the study period. Hospital admission logs were reviewed to minimize the possibility that patients with predisposing conditions were missed. After identification of at-risk ED patients, they were followed through their hospitalization prospectively in 19 hospitals. In the three hospitals that enrolled retrospectively, investigators followed the same protocol and definitions, but data were collected after patient discharge.

Baseline characteristics, including demographics, comorbidities, and clinical variables, were collected during the first 6 h of initial ED evaluation. Predisposing conditions and ALI risk modifiers were identified and collected. Predisposing conditions were pre-defined, and ALI risk modifiers included: alcohol abuse, obesity chemotherapy, diabetes mellitus, smoking, tachypnea, hypoxemia, oxygen supplementation, hypoalbuminemia, and acidosis.

De-identified subject information was entered at each center into the secure, password-protected NIH-supported web form (REDCap http://www.project-redcap.org). Electronic range checks and validation rules were utilized to eliminate erroneous data entry and artifacts in numeric values. Prior to study initiation at each site, investigators and study coordinators reviewed the definitions of each risk factor (see Additional file 1: Appendix 2) and received mandatory structured online training for ALI assessment. Briefly, in determining if a chest radiograph is consistent with ALI, assessment begins with interpretability of the x-ray followed by evaluation for bilateral opacities generally described as infiltrates consistent with pulmonary edema. Only when the bilateral opacities are not fully explained by non-qualifying opacities (i.e., pulmonary fibrosis) and not limited to the lower lung zones with normal parenchyma above is the chest radiograph consistent with ALI. In addition, a formal training session was provided during the 2009 USCIITG meeting in Nashville, TN. The principal investigators from each site were responsible for data collection and entry, as well as quality control.

The primary outcome was the development of AECC-defined ALI during the hospital admission of at-risk ED patients. Secondary outcomes included time to ALI development; proportion and duration of invasive and non-invasive mechanical ventilation; vasopressor requirement; acute renal failure requiring hemodialysis; and ICU and hospital length of stay and mortality.

The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were followed in the design and reporting of this observational study [²²]. PCH and AM analyzed the data, which were summarized as number (in percentage) and median (with inter-quartile range). Missing data were coded explicitly as described and handled by using logic expression. Continuous variables were dichotomized at the median. The odds ratios and 95 % confidence intervals were computed from performing logistic regression. Statistical significance was set at 0.05 for regression analysis. Frequency of ALI was calculated per number of ED patients presenting with predisposing condition at the time of hospital admission.

In order to evaluate our secondary outcome measures, we compared hospital and ICU mortality and length of stay between ED patients at risk who developed ALI and those who did not. To determine the mortality burden attributed to the development of ALI, we performed a logistic regression analysis adjusted for the baseline Acute Physiology and Chronic Health Evaluation (APACHE II) score [²³]. In addition, we described the utilization and duration of invasive and non-invasive mechanical ventilation, and performed an exploratory analysis comparing patients who did and did not develop ALI with respect to their initial vent settings, specifically tidal volume per predicted body weight, plateau pressure, and positive end expiratory pressure (PEEP). Lastly, we illustrated the frequency of ALI development for each predisposing condition, at hospital day onset, and at each hospital setting. All statistical analysis was performed in SAS 9.2 (SAS Institute, Cary, NC).

Sites with retrospective data and extreme frequencies of ALI development were excluded, and a sensitivity analysis of the prospective cohort from the entire ED cohort was performed.

ALI developed in 303 (7.0 %) admitted ED patients with a median of 2 days, inter-quartile range (IQR) 2–5 days. Among patients who developed ALI, a subset of 198 (65.3 %) met the ARDS criteria. The follow-up to hospital discharge was complete in all patients. Baseline characteristics, severity of illness, and predisposing conditions and ALI risk modifiers differed between patients who did and those who did not develop ALI (Table 1). Patients who developed ALI were more likely male and heavier, and had a higher APACHE II score. The majority of patients had all measurements available at the time of hospital admission except for serum albumin (n = 2,423) and arterial pH (n = 1,499). As these tests are usually ordered based on clinical suspicion, missing data were considered normal (i.e., if serum albumin or arterial pH was not measured, hypoalbuminemia and acidosis were coded as absent), similarly to APACHE score calculation. The odds ratios for ALI development were highest in risk modifiers of tachypnea and a fraction of inspired oxygen 35 % or greater; and in predisposing conditions of aortic vascular, cardiac, and spinal surgeries, near-drowning, and smoke inhalation. However, the numbers of subjects in some of these categories were small compared to others. The frequency of ALI varied according to predisposing condition with the highest rate occurring after emergent aortic surgery (42.9 %) and the lowest rate occurring in pancreatitis (2.8 %) (Figure 3).

Outcome data for the study cohort are shown in Table 2. More than half of the entire cohort and 91 % of patients who developed ALI were treated in the ICU; and 31 % of the entire cohort and 95 % of patients with ALI were either invasively or non-invasively mechanically ventilated. Compared to at-risk patients who did not develop ALI, those who developed ALI were more likely to be ventilated invasively (88 % vs 19 %) and non-invasively (30 % vs 10 %), respectively. Similarly, comparing to patients who did not develop ALI, those who did had increased resource utilization as reflected in more vasopressor usage (38 % vs 8 %), a higher percentage of acute hemodialysis requirement (11 % vs 3 %), and longer ICU (9 vs 2 days) and hospital (16 vs 5 days) lengths of stays. More importantly, patients who developed ALI had increased ICU (24 % vs 3 %) and hospital (28 % vs 5 %) mortality. When adjusted for severity of illness using the APACHE II score, the development of ALI markedly increased the risk of in-hospital death by more than four-fold [OR 4.45, 95%CI (3.23, 6.14)].

The variation in the frequency of ALI development among the 22 study sites is illustrated in Figure 4. Even after adjustment for each site’s APACHE II score, a significant variation in the frequency of ALI development remained. By excluding the one outlier (44 %), the frequency of ALI development varied from 0.7 to 12.8 %.

Since the data set is fairly large, the utilization of invasive mechanical ventilation and onset of ALI development were explored. Figure 5 provides a description of ALI development and initiation of invasive mechanical ventilation by hospital day. The majority of patients who developed ALI and who received invasive mechanical ventilation did so within the first 2 days after hospital admission. As illustrated in Figure 6, the majority of patients who developed ALI had initiation of invasive mechanical ventilation on the day of ALI onset. Unfortunately, no data were recorded regarding the exact timing (in hours and minutes) and the reason for initiation of invasive mechanical ventilation with respect to the onset of ALI development. Although relatively few patients developed ALI prior to initiation of invasive mechanical ventilation, many developed ALI a day or more after initiation of invasive mechanical ventilation.

When data from the two non-US study sites were removed (n = 4,233) from the analysis (see Additional file 1: Appendix 3), ALI developed in 266 (6.3 %) patients with a median of 2 days (IQR 2–5 days) after admission to the hospital. The frequency of ALI varied according to predisposing condition with the highest rate of ALI occurring after aortic surgery (36.4 %) and the lowest rate occurring in pancreatitis (2.9 %). Baseline characteristics, severity of illness, predisposing conditions and ALI risk modifiers, and secondary outcomes remained different between patients who did and those who did not develop ALI. However, there were no significant differences in the group characteristics from the US study sites when comparing to the entire cohort.

In addition, when data from the three retrospective US study sites were removed (n = 3,981) from our descriptive analysis (see Additional file 1: Appendix 4), ALI developed in 241 (6.1 %) patients with a median of 2 days (IQR 2–4 days) after admission to the hospital. The frequency of ALI varied according to predisposing condition with the highest rate of ALI occurring after aortic surgery (40 %) and the lowest rate occurring in pancreatitis (2.4 %). Baseline characteristics, severity of illness, predisposing conditions and ALI risk modifiers, and secondary outcomes maintained their differences between patients who did and did not develop ALI. However, there were no significant differences in the group characteristics when comparing the prospective US sites and the entire US cohort.

Since there is growing evidence to suggest that a lung protective strategy with low tidal volume ventilation may be protective in at-risk patients, we explored the effect of the initial vent setting on ALI development [²⁴,²⁵]. We evaluated the initial tidal volume divided by predicted body weight (TV/PBW in cc/kg) in 768 patients who underwent invasive mechanical ventilation (Table 2). Comparing those who did not, patients who developed ALI received a significant lower initial TV/PBW [8.1 vs 8.4, OR 0.70 (95 % CI 0.51-0.96), p = 0.029] but had a significantly higher plateau pressure [21 vs 19, OR 1.77 (95 % CI 1.17-2.69), p = 0.007] and peak end-expiratory pressure (PEEP) [OR 2.30 (95 % CI 1.32-4.01), p = 0.003]. Regarding the mode of mechanical ventilation, a significantly higher proportion of patients who developed ALI was initiated on pressure control ventilation [OR 2.13 (95 % CI 1.41-3.2), p = 0.01] and non-invasive ventilation [OR 3.87 (95 % CI 2.93-5.11), p < 0.001] compared to those who did not develop ALI; in contrast, no significant difference was found in those who underwent initial volume control ventilation.

Our study carries the limitations of observational cohort studies and those inherent to clinical research with ALI: inter-observer reliability of portable chest X-ray interpretation, ALI imitators, and consistency in determining exclusion of left atrial hypertension as the principal cause of pulmonary edema [^26-28]. We instituted a mandatory structured training in ALI assessment and held the site-principal investigators responsible for quality control. These measures were intended to mitigate these limitations.

Regarding the variation in the frequency in ALI development among the study sites, two non-US sites had the highest rate and could be explained by differences in their health-care delivery system, specifically emergency medical care and critical care services, and possibly population and environmental factors. When the data from these two non-US centers were excluded, no significant difference in the frequency of ALI development (p = 0.216) was found between the US sites. In addition, the vast majority of patients were enrolled prospectively ensuring close follow-up and reducing the risk of misclassification from medical record review. When data from those centers enrolling retrospectively were excluded, no significant difference in the frequency of ALI development (p = 0.665) was found comparing the prospective US sites to the entire US cohort. Hence, although a significant variation in the frequency of ALI development across institutions was found, the US sites and the prospectively enrolling US sites were not significantly different when each subgroup was compared to the entire cohort.

Even though ALI is classified as a rare disease [²⁹], it is a major public health concern. It is, however, unclear how much if variations in initial management of ALI-prone patients by emergency physicians contribute to its development. To our knowledge, this is the largest detailed study of a cohort of hospitalized adult ED patients at risk for ALI development. The strengths of this study include the large sample size from a geographically diverse population of patients at both academic and community hospitals. Using routinely available clinical data, we identified ED patients at risk for ALI development early in the course of their illness. The early identification of predisposing conditions and risk modifiers as well as subsequent interventions in the ED and ICU may potentially prevent disease development by minimizing or avoiding secondary insults.

Interestingly, in our study we found that ALI developed with lower frequencies than previously reported following conditions recognized to predispose patients at risk for ALI such as: aspiration, pneumonia, sepsis, and trauma [³⁰,³¹]. This may be explained by the fact that our patients were enrolled early (in the ED) without any signs of ALI (with at least one predisposing factor present) and we excluded those who developed ALI within 6 h of ED presentation, while those other studies enrolled patients upon ICU admission. Consistent with our findings, a recent study by Ferguson et al. showed that 7 % of patients with sepsis, 2 % of patients with pancreatitis, 10 % of patients with pneumonia, and 15 % of patients with witnessed aspiration developed ALI [³²]. Similarly, the majority of patients with predisposing conditions never developed ALI, and many were not admitted to the ICU.

Our inclusion criteria required the presence of at least one ALI risk factor at the time of hospital admission, potentially missing the patients who acquire a predisposing condition and received a secondary injury later in the hospital stay. We cannot rule out that a minority of our patients identified as high risk were already progressing to develop full-blown ALI at the time of enrollment, although the exclusion of those who developed ALI within 6 h of ED presentation intended to minimize this possibility. However, we do believe that earlier identification of such patients would also be of benefit, and could help limit the progression of ALI development and improve patient outcomes by alerting providers to make efforts to limit second-hit exposures. To support this notion, a population-based study in Olmsted County, Minnesota, has recently shown a steady decline in ARDS incidence. This was attributable entirely to a reduced incidence of hospital-acquired ARDS and suggests that recent improvements in prevention, early recognition, and critical care delivery may in part be responsible for this [³³].

From Olmsted County’s experience, major system changes within the hospital throughout the years were progressively made. These include electronic medical records with computerized order entry to monitor data and institute decision support, restrictive transfusion protocol with leukoreduction and male donor predominant plasma transfusion, respiratory therapy protocol on limiting initial volume according to predicted body weight on all patients, increased staffing of intensivists with 24-h on site, sepsis resuscitation protocols and teams, rapid response teams, standardization of inpatient pneumonia care, and staff education and training. Regarding potential interventions that may affect the development of ALI, a checklist for lung injury prevention (CLIP) has been proposed and developed by experts in the field [³⁴]. The CLIP domains (and elements) consist of routine ICU practices of morbidity prevention and include: respiratory support (lung protective strategies, minimizing oxygen toxicity) [¹²,²⁴,²⁵]; aspiration precautions (rapid sequence intubation, head-of-bed elevation, oral care with chlorhexidine) [^35-37]; infection control (early and appropriate antibiotic therapy, source control, prevention of nosocomial infection transmission) [³⁸]; fluid management (early fluid resuscitation in severe sepsis and septic shock, fluid restriction after shock resolution) [³⁹,⁴⁰]; transfusion management (restrictive red blood cell transfusion threshold, transfusion guidelines for blood products) [⁴⁰,⁴¹]; and communication (validated structured handoffs such as SBAR: situation, background, assessment, and recommendation) [⁴²]. As a continuum in the care of the critically ill starting in the ED and transitioning to the ICU, all of these domains and many of these proposed elements are being followed to a certain extent, but inconsistently as reflected in our own experiential observations and practice variations. However, such a proposed checklist will require validation for its utility in adherence to best practices and ALI prevention.

PRW: Predicted body weight; RR: respiratory rate; tachypnea = RR > 30; SpO₂: oxygen saturation; FiO₂: fraction of inspired oxygen.

ICU: Intensive care unit; LOS: length of stay; TV: tidal volume; PBW: predicted body weight; PEEP: peak end-expiratory pressure.