Study cohort selection and data preprocessing. The data mining process is illustrated as a top-down flow diagram. The entire cohort comprised 262 638 patients, and the identification process returned 34 809 patients for the scope of this study. During data preprocessing, 437 791 samples were generated, which were further split into training, validation, and test sets on a per-patient level. Each extracted resource had to pass validation processes to ensure the raw data were consistent. The validation processes included comparing the total number of available data in FHIR by resource to the downloaded resources and manual spot checks between source systems and the extracted data set.