Skip to main content
. 2022 Jul 5;35(6):1514–1529. doi: 10.1007/s10278-022-00674-z
Datasets
Design decisions We included only cases with admission CXRs and results of laboratory tests taken within the first 24 h of hospital admission in Tables 1 and 2. Cases were also excluded for missing admission time. Patients aged 16 or under are excluded and only frontal (AP or PA) images are included for the CXR-based and EHR-CXR fusion models
Reasons Our clinical goal is to develop an early assessment algorithm. Admission time is needed to establish the 30-day mortality cut off for this study. Patients under 16 need more privacy protection (very rare and more easily re-identifiable) and their CXR imaging appearance (anatomically) and disease outcome distributions are very different. Not all CXR exam orders include lateral images hence they are not included as an input for the models