Objective: To determine and compare the diagnostic accuracy of three clinical definitions of acute respiratory distress syndrome (ARDS): (1) the American-European consensus conference definition; (2) the lung injury score; and (3) a recently developed Delphi definition. A second objective was to determine the accuracy of clinical diagnoses of ARDS made in daily practice.
Design: Independent comparison of autopsy findings with the daily status of clinical definitions, constructed with data abstracted retrospectively from medical records.
Setting: Tertiary intensive care unit.
Patients: One hundred thirty-eight patients from the period 1995 through 2001 who were autopsied after being mechanically ventilated.
Interventions: Clinical ARDS diagnoses were determined daily without knowledge of autopsy results. Charts were reviewed for any mention of ARDS in the clinical notes. Autopsies were reviewed independently by two pathologists for the presence of diffuse alveolar damage. The sensitivity and specificity of the definitions were determined with use of diffuse alveolar damage at autopsy as the reference standard.
Measurements and main results: Diffuse alveolar damage at autopsy was documented in 42 of 138 cases (30.4%). Only 20 of these 42 patients (47.6%) had any mention of ARDS in their chart. Sensitivities and specificities (95% confidence intervals) were as follows: American-European definition, 0.83 (0.72-0.95), 0.51 (0.41-0.61); lung injury score, 0.74 (0.61-0.87), 0.77 (0.69-0.86); and Delphi definition, 0.69 (0.55-0.83), 0.82 (0.75-0.90). Specificity was significantly higher for both the lung injury score and Delphi definition than for the American-European definition (p < .001 for both), whereas comparisons of sensitivity, which was higher for the American-European definition, were not significantly different (p = .34 and p = .07, respectively).
Conclusions: Acute respiratory distress syndrome appears underrecognized by clinicians in patients who die with this syndrome. In this population, the specificities of existing clinical definitions vary considerably, which may be problematic for clinical trials.