A PRCI-Based Individualized Risk Prediction Model for Potentially Inappropriate Medication in Older Outpatients

Abstract

Objectives: Potentially inappropriate medication (PIM) use remains highly prevalent among older outpatients, yet individualized risk prediction tools are limited. This study aimed to develop and validate a PIM risk prediction model incorporating a novel disease-weighted index. Methods: A multicenter cross-sectional study was conducted using 131,894 outpatient prescriptions from 59 medical institutions in China. Diagnoses were mapped to ICD-10 codes. A PIM Risk-Related Comorbidity Index (PRCI) was developed using multivariable logistic regression and coefficient-weighted scoring. A prediction model was constructed and internally validated using a 6:4 split-sample approach. Discrimination, calibration, and clinical utility were assessed using ROC curves, calibration plots, and decision curve analysis. Results: The proportion of prescriptions containing at least one PIM was 29.0%. Polypharmacy showed a strong dose–response relationship with PIM risk. The PRCI was independently associated with PIM (OR = 1.07 per unit increase). The final model demonstrated good discrimination (AUC = 0.775 in training and 0.773 in validation), satisfactory calibration, and favorable clinical utility. A cutoff probability of 0.304 stratified prescriptions into high- and low-risk groups. Conclusions: The PRCI-based model provides an individualized and clinically applicable tool for identifying older outpatient prescriptions at high risk of containing PIMs, supporting proactive medication safety management.