Supplementary code for the second revision.

MIMIC-IV extraction was performed using PostgreSQL SQL scripts. eICU-CRD extraction was performed using Google BigQuery SQL scripts. Python scripts were used for preprocessing, imputation support, round-2 sensitivity analyses, and figure generation. R scripts from the first major revision are retained for the multiple-imputation, IPTW, and Cox workflow.

Prone positioning in MIMIC-IV was identified from charted prone-positioning-related records in the local extraction workflow. In eICU-CRD, prone positioning was identified from treatment/nurse-charting entries using prone-positioning labels reviewed during schema inspection.
