Developing an Outcome of Adult Respiratory Distress Syndrome in a Trauma Population

• Activity Number: 90
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 9:00 AM to 10:50 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Christina A. Gaughan*+ and Jason D. Christie and Paul N. Lanken and Barbara B. Finkel and Barry Fuchs and Robert Gallop and J. Richard Landis
• Companies: University of Pennsylvania School of Medicine and University of Pennsylvania School of Medicine and University of Pennsylvania and University of Pennsylvania and University of Pennsylvania and West Chester University and University of Pennsylvania School of Medicine
• Keywords: biostatistics ; outcome classification ; sensitivity analysis

Abstract

An outcome is not always defined as straightforward categorical presence or absence of a disease or syndrome. This problem is particularly poignant when a syndrome has several defining criteria with variable precision in their definitions. Such is the case when identifying ARDS. By consensus, ARDS is defined as meeting three criteria within a 24-hour period: (1) bilateral pulmonary infiltrates on chest x-ray consistent with pulmonary edema; (2) absence of evidence of left atrial hypertension; and (3) poor systemic oxygenation (PaO2/FiO2 ratio <=200). Within a prospective cohort study of 273 trauma patients aimed at understanding the genetic and biological mechanisms of ARDS, we sought to assess the time of onset of ARDS as well as generate methods to minimize outcome misclassification. The following challenges were addressed: (1)agreement of chest radiograph interpretation; (2) interpretation of longitudinal changes in PaO2/FiO2 ratio; (3)classification of subjects meeting partial or nonsimultaneous criteria; (4) timing of classification relative to other time-varying covariates; (5) robustness of ARDS definition to variations in PaO2/FiO2 thresholds (sensitivity). NHLBI SCOR acknowledgment.