Fig. 5: Sensitivity analysis.

a Forest plot illustrating the predictive performance for 48 h ICU mortality of the LGBM-48h algorithm, stratified by ICD-10 diagnosis group assigned on admission. Black filled circles represent AUROC values for each admission diagnosis group based on ICD-10 categories. Error bars indicate 95% confidence intervals derived using the DeLong method. Dotted line indicates area under the receiver operating characteristic curve (AUROC) for the LGBM-48 h algorithm, shaded blue area illustrate 95% confidence intervals (n = 30,171 stay days in the test set). b Forest plot illustrating the predictive performance for 48-hour ICU mortality of the LGBM-48 h algorithm, stratified by ICD-10 diagnosis group assigned after review of all available patient data at the end of the respective ICU stay. Black filled circles represent AUROC values for each final diagnosis group based on ICD-10 categories. Error bars indicate 95% confidence intervals derived using the DeLong method. Dotted line indicates area under the receiver operating characteristic curve (AUROC) for the LGBM-48h algorithm, shaded blue area illustrate 95% confidence intervals (n = 30,171 stay days in the test set). c Receiver operating characteristic curves comparing LGBM algorithms that predict 24 h ICU mortality (LGBM-24h), 48 h ICU mortality (LGBM-48h) and 72 h ICU mortality (LGBM-72 h), respectively. d Receiver operating characteristic curves comparing LGBM algorithms that predict 48 h ICU mortality (LGBM-48 h), overall ICU mortality and in-hospital mortality.