Prediction models for severe hand-foot-mouth disease risk
10.19485/j.cnki.issn2096-5087.2026.03.016
- Author:
XU Jing
;
ZHANG Junhong
;
FENG Huifen
- Publication Type:Journal Article
- Keywords:
hand-foot-mouth disease;
severe;
random forest model;
linear support vector machine model;
logistic regression model;
prediction
- From:
Journal of Preventive Medicine
2026;38(3):296-301
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct and compare risk prediction models for severe hand-foot-mouth disease (HFMD), analyze the influencing factors of severe HFMD, so as to provide a reference for early identification and intervention of severe HFMD.
Methods:HFMD patients who were hospitalized in a hospital in Zhengzhou City from November 2018 to December 2021 were selected as study subjects. Demographic information, clinical symptoms, and laboratory test indicators were collected. The subjects were randomly divided into a training set and a validation set at a 7∶3 ratio. According to the Guidelines for Diagnosis and Treatment of Hand-Foot-Mouth Disease (2018 Edition), the patients were divided into a mild HFMD group and a severe HFMD group. Variables with statistically significant differences between the two groups were identified as predictive variables. Random forest model, linear support vector machine (LSVM) model, and logistic regression model were constructed. The predictive performance of the three models was evaluated using accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC).
Results:A total of 1 650 HFMD patients were included, including 1 094 males (66.30%) and 556 females (33.70%). The mean age was (1.97±1.17) years. There were 1 293 cases in the mild HFMD group (78.36%) and 357 cases in the severe HFMD group (21.64%). The accuracy of random forest model, LSVM model, and logistic regression model in predicting severe HFMD were 95.76%, 88.28%, and 88.08%, respectively. The sensitivity were 94.55%, 59.09%, and 58.18%, respectively. The specificity were 96.10%, 96.66%, and 96.66%, respectively. The AUC values were 0.981, 0.895, and 0.899, respectively. The random forest model results showed that the top three important predictors for severe HFMD were neutrophil ratio, alanine aminotransferase, and platelet count.
Conclusions:The constructed random forest model demonstrates good predictive performance and can be applied in predicting severe HFMD. It is recommended to pay attention to changes in neutrophil ratio, alanine aminotransferase, and platelet count in HFMD patients, which can assist in early identification and intervention of severe HFMD.
- Full text:2026110910124757574重症手足口病风险预测模型研究.pdf
