An alternative method of acute lung injury classification for use in observational studies

Background: In observational studies using acute lung injury (ALI) as an outcome, a spectrum of lung injury and difficult-to-interpret chest radiographs (CXRs) may hamper efforts to uncover risk factor associations. We assessed the impact of excluding patients with difficult-to-classify or equivocal ALI diagnosis on clinical and genetic risk factor associations for ALI after trauma. Methods: This study was of a prospective cohort of 280 critically ill trauma patients. The primary outcome was the development of ALI. Patients were classified into one of three groups:(1) definite ALI (patients who fulfilled the American-European Consensus Conference [AECC] criteria for ALI), (2) equivocal ALI (patients who had difficult-to-interpret CXRs), and (3) definite non-ALI. We compared clinical and genetic ALI risk factor associations between two classification schemes: AECC classification (definite ALI vs rest) and alternative classification (definite ALI vs definite non-ALI, excluding equivocal ALI). Results: Ninety-three (35%) patients were classified as definite ALI, 67 (25%) as equivocal, and 104 (39%) as definite non-ALI. Estimates of clinical and genetic ALI risk factor associations were farther from the null using the alternative classification. In a multivariable risk factor model, the C statistic of the alternative classification was significantly higher than that derived from the AECC classification (0.82 vs 0.74; P < .01). Conclusions: The ability to detect ALI risk factors may be improved by excluding patients with equivocal or difficult-to-classify ALI. Such analyses may provide improved ability to detect clinical and genetic risk factor associations in future epidemiologic studies of ALI.