Development and Validation of a Predictive Screening Model for Congenital Heart Disease in High-Altitude Children

Current congenital heart disease (CHD) screening protocols adopt plain-standard criteria, yet they predispose to missed diagnoses and misdiagnoses in high-altitude hypoxic environments. This study aimed to develop and validate a predictive model for early CHD screening in plateau-region school-aged children and to evaluate the cost-effectiveness of different screening strategies.

Cross-sectional data from 7315 school-age children undergoing initial school screening in high-altitude areas were analyzed using R 4.2.0 and SPSS 26.0. A logistic regression model predicting CHD detection rate was built via stepwise selection. Model performance was assessed through decision curve analysis (DCA), calibration curves, and ROC analysis. Variable importance was visualized via random forest plots. Cost-effectiveness of three screening strategies was evaluated using a Markov model.

Logistic regression model identified drug use (OR=4.368) and respiratory infections (OR=5.795) were independently associated with CHD diagnosis, followed by age (OR=0.680) and smoking (OR=1.476). The model showed excellent discrimination (AUC=0.867). Random forest analysis confirmed respiratory tract infections as the significant associated factor. Cost-effectiveness analysis identified the three-tier screening model as dominant, offering the lowest cost (¥2604) and highest health benefit (0.060 QALYs), with the favorable incremental cost-utility ratio. Single-stage diagnosis was the least cost-effective (¥15,400).

The three-tier screening strategy, prioritizing children with maternal history of drug use or smoking exposure, younger age, and history of respiratory tract infections, shows a promising cost-effective approach for early CHD detection in high-altitude settings.