A Scoring System with High-Resolution Computed Tomography to Predict Drug-Associated Acute Respiratory Distress Syndrome: Development and Internal Validation

Abstract

Drugs can cause acute respiratory distress syndrome (ARDS). However, there is no established clinical prediction rule for drug-associated ARDS (DARDS). We aimed to develop and validate a scoring system for DARDS prediction. We analysed data collected from a prospective, single-centre, cohort study that included ARDS patients. The ARDS diagnosis was based on the American-European Consensus Conference or Berlin definition. Drug-associated acute lung injury (DALI) was defined as previous exposure to drugs which cause ALI and presence of traditional risk factors for ALI. High-resolution computed tomography (HRCT; indicating extent of lung damage with fibroproliferation), Acute Physiology and Chronic Health Evaluation (APACHE) II, and disseminated intravascular coagulation (DIC; indicating multiorgan failure) scores and PaO2/FiO2 were evaluated for their ability to predict DARDS. Twenty-nine of 229 patients had DARDS. The HRCT, APACHE II, and DIC scores and PaO2/FiO2 were assessed. The model-based predicted probability of DARDS fitted well with the observed data, and discrimination ability, assessed through bootstrap with an area under the receiver-operating curve, improved from 0.816 to 0.875 by adding the HRCT score. A simple clinical scoring system consisting of the APACHE II score, PaO2/FiO2, and DIC and HRCT scores can predict DARDS. This model may facilitate more appropriate clinical decision-making.