查看文章

Introduction Acute respiratory distress syndrome (ARDS) is a heterogenous entity making definitive diagnosis challenging. Misdiagnosis of ARDS occurs commonly leading to low adherence to evidence-based therapies and research limitations. Automated algorithms utilizing electronic health record (EHR) data may allow for large-scale accurate ARDS identification. Methods Machine-learning models were trained to detect ARDS using readily available EHR data in acute hypoxemic respiratory failure (AHRF) patients at a single center. The incremental value of including different data types and differing machine-learning modeling approaches were compared by their area under the receiver operating characteristics curve (AUROC), specificity and positive predictive value (PPV) at a threshold achieving 85% sensitivity. Clinical features from radiology reports were processed using cTAKES (clinical text knowledge …