3.3 External validation
After applying the eligibility criteria, the external validation dataset included 372 cases with serious GI complications and 254,179 controls without any GI complications. These control groups did not undergo a control-selection matching process (Fig 1). The Hosmer–Lemeshow test for the external validation dataset showed an AUROC of 0.77 (95% CI, 0.75–0.80), indicating acceptable discrimination (Table 3). The calibration curves of the external validation dataset showed nearly perfect calibration (calibration slope β =1.08), and the risk of GI complications increased as the prediction scores increased (Fig S1).
Based on a total risk score cut-off of 16, patients were classified into high-risk or low-risk groups. There were 237 (4.6%) and 134 (20.5%) cases of serious GI complications in the high- and low-risk groups, respectively, with a statistically significant difference (p<0.01, Table 4).
The top ten cases predicted based on our model are presented in Table S3. For age factors from the risk model, 85 years or older could not be applied due to data limitations (HIRA-APS 2019).