Development and validation of multi-algorithm machine learning models for predicting phototherapy efficacy in premature infants with hyperbilirubinemia and nursing decision recommendations
Year: 2026
Type: article
Abstract: Objective: To develop multi-algorithm machine learning models for predicting the efficacy of phototherapy in preterm infants with hyperbilirubinemia. Methods: A total of 250 preterm infants who received phototherapy were retrospectively enrolled and randomly divided into training and validation sets at a ratio of 7:3. A reduction of ≥ 20% in serum total bilirubin after 24 h of phototherapy was defined as an effective response. Following variable selection, three predictors—peak bilirubin level, albumin level, and multiple pregnancy—were included in the models. Logistic regression, Lasso, support vector machine, and naive Bayes models were developed and evaluated using the Area under the Receiver Operating Characteristic Curve (AUC), calibration curves, and decision curve analysis. SHAP analysis was also performed. Results: In the validation set, the logistic regression model achieved the highest AUC of 0.905 (95% CI: 0.834–0.975), while the AUCs of the other models ranged from 0.884 to 0.903. All four models demonstrated good calibration and positive net benefits across a relatively wide range of threshold probabilities. SHAP analysis identified the three aforementioned variables as the leading contributors to model predictions. Conclusion: The multi-algorithm models demonstrated good predictive performance, with logistic regression achieving the best overall performance. SHAP analysis provides a quantitative basis for identifying key risk variables and informing stratified nursing decisions. However, the findings require further external validation.
Authors: Qiufen Shu, Xiuping Huang
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7236562154
