Multidomain geriatric assessment improves prediction of unplanned inpatient admission after day-case ERCP in older adults

Abstract Identifying older patients at risk of unplanned inpatient admission after planned day-case endoscopic retrograde cholangiopancreatography (ERCP) remains challenging because conventional clinical factors may overlook discharge-related vulnerabilities. We assessed the incremental predictive value of multidomain geriatric assessment in a retrospective cohort of 1,407 patients aged ≥ 65 years undergoing planned day-case ERCP, comprising a development cohort (n = 1,000) and a subsequent temporal validation cohort (n = 407). Adding frailty, functional independence, nutritional status, and cognition to conventional preprocedural factors increased the temporal-validation AUC from 0.730 to 0.830 (difference, 0.100; 95% CI, 0.039–0.164), improved average precision from 0.395 to 0.552, and reduced the Brier score from 0.133 to 0.118. Models additionally incorporating procedural and early-recovery information available before the final discharge decision achieved temporal-validation AUCs of 0.872 for L2-penalized logistic regression and 0.860 for XGBoost, although calibration deteriorated. These findings demonstrate that multidomain geriatric assessment provides incremental predictive value beyond conventional clinical factors in older adults undergoing planned day-case ERCP. Integrating geriatric assessment with procedural and early-recovery information may support individualized risk stratification, perioperative monitoring, and discharge planning.