A Prediction Model for Adjusting the Trigger Strategy of Preoperative Magnetically Controlled Capsule Endoscopy in Patients at High Gastrointestinal Risk Undergoing Percutaneous Coronary Intervention

Background Patients scheduled for percutaneous coronary intervention (PCI) often require intensified antithrombotic therapy. Potential high-risk upper gastrointestinal (GI) lesions may influence PCI timing and the selection of antithrombotic strategies. Magnetically controlled capsule endoscopy (MCCE) enables sedation-free upper GI evaluation. However, under inpatient workflow constraints and limited endoscopic resources, a quantitative tool to identify, before testing, patients who are more likely to undergo management modification triggered by MCCE findings is lacking. Methods We retrospectively included a single-center cohort of patients scheduled for PCI who underwent preprocedural MCCE due to high-risk factors for GI injury (n = 2155). The outcome was strategy modification (yes/no), defined as a change in PCI timing and/or antithrombotic-related management attributable to preprocedural MCCE findings. Without incorporating any endoscopic variables, we developed a logistic regression model using only routinely available pretest clinical information and constructed a nomogram. Internal validation was performed using stratified 10-fold cross-validation and bootstrap resampling (150 iterations). Temporal validation was conducted using a time-split approach (first 70% as training; last 30% as testing). Discrimination and calibration were evaluated using the area under the receiver operating characteristic curve (AUC) and Brier score; ROC and calibration curves were plotted. Results Among 2155 patients, 215 (9.98%) experienced strategy modification triggered by MCCE findings. The model achieved a mean AUC of 0.938 (SD 0.029) in 10-fold cross-validation; the apparent AUC was 0.945, and the optimism-corrected AUC after bootstrapping was 0.940. In the time-split test set, the AUC was 0.904 and the Brier score was 0.180. The nomogram enabled bedside individualized estimation of modification probability. Using a predicted probability threshold of ≥ 20% to prioritize MCCE, the negative predictive value was 99.5%, effectively identifying patients with low expected benefit and supporting workflow triage. Conclusions A prediction model based on routinely available preprocedural clinical information can identify patients more likely to undergo PCI and/or antithrombotic strategy modification prompted by preprocedural MCCE, providing quantitative support for inpatient workflow triage and antithrombotic management. External validation in independent cohorts and prospective studies are warranted to evaluate its impact on clinical outcomes and resource utilization.