Fig. 5: Flowchart from patient selection to time-to-event analysis.

The study included hospitalized patients aged 16 to 99 years from January 1, 2004, to December 31, 2021. Patients registered with ICD-10 codes for Malignant Neoplasms in the DPC were included. The inclusion criteria encompassed patients prescribed any of the three anticancer drug classes and those did not prescribe any anticancer drugs throughout. Exclusion criteria were applied to patients with suspected cancer diagnosis only, those who died within 24 h of hospitalization, patients without clinical records in text form, and those with a recorded outcome within 180 days before the observation start date. Missing data were handled using multiple imputation by chained equation, performed 20 times. Propensity score matching was then conducted, pairing the exposure group with the no treatment group using a one-to-one nearest neighbor matching method. Finally, a time-to-event analysis was performed using a Cox PH model. The model considered the prescription of the anticancer drug class under analysis and the prescription of other anticancer drug classes in the past 180 days as covariates, with the outcome being days to occurrence of the adverse event group under analysis.