Analysis of health anxiety trajectories in rabies exposure patients based on machine learning

To identify the dynamic evolution trajectory of health anxiety in patients with initial rabies exposure and to conduct predictive analysis of health anxiety trajectories using machine learning algorithms.

A total of 548 rabies exposure patients from a tertiary hospital in Nanning were selected between January 2024 and October 2025. Health anxiety levels were assessed at 0, 3, 7, 14, and 28 days post-exposure using the Health Anxiety Scale following a 5-dose immunization schedule. A latent variable growth mixture model was employed to identify anxiety development trajectories. Six machine learning algorithms were utilized to establish a high-risk health anxiety identification model, and a comprehensive evaluation of the model's performance was conducted, with the SHAP algorithm used for model interpretation.

All 548 patients completed the survey. The latent variable growth mixture model identified three health anxiety development trajectories: low-level anxiety-stable group, moderate to high-level anxiety-improving group, and high persistent anxiety-hard to relieve group. Multivariate logistic regression analysis indicated that being female, having a college degree or higher, high rabies knowledge, a greater number of comorbidities, level III exposure, and delayed medical treatment were independent risk factors belonging to the 'high persistent anxiety-hard to relieve group.' In contrast, a higher level of health literacy was identified as a protective factor. The XGBoost model demonstrated superior predictive performance on the independent test set, with an area under the receiver operating characteristic curve of 0.881, while SHAP analysis ensured clinical interpretability. SHAP analysis of the XGBoost model revealed that health literacy, Perceived social support, rabies knowledge, Payment method and Animal source were the five most important features predicting the 'high persistent group.

There is heterogeneity in the health anxiety trajectories of patients exposed to rabies. A predictive model based on machine learning can facilitate the early identification of high-risk individuals, providing a basis for implementing precise psychological interventions.