ISSN 1674-3865  CN 21-1569/R
主管:国家卫生和计划生育委员会
主办:中国医师协会
   辽宁省基础医学研究所
   辽宁中医药大学附属医院

Chinese Pediatrics of Integrated Traditional and Western Medicine ›› 2026, Vol. 18 ›› Issue (1): 71-76.doi: 10.20274/j.cnki.issn.1674-3865.2026.01.014

Previous Articles     Next Articles

Application research on risk factor analysis of severe Mycoplasma pneumoniae pneumonia in children based on random forest model

Sen KONG, Liang ZHANG(), Nan LIAO, Guang TU   

  1. Lichuan People's Hospital, Fuzhou 344600, China
  • Received:2025-07-29 Revised:2025-09-29 Published:2026-02-25 Online:2026-02-25
  • Contact: Liang ZHANG E-mail:987928918@qq.com
  • Supported by:
    First Batch of 2024 Fuzhou Municipal Guiding Science and Technology Projects for Social Development

Abstract:

Objective To identify risk factors for severe pediatric Mycoplasma pneumoniae pneumonia using a random forest model and to provide evidence for clinical recognition and intervention. Methods Clinical data of 198 children hospitalized for Mycoplasma pneumoniae pneumonia in our hospital between January and December 2024 were collected; 133 were classified as non-severe and 65 as severe. Clinical characteristics were compared between the two groups. Univariate Logistic regression analysis and a random forest model were employed to evaluate the impact of each variable on severe Mycoplasma pneumoniae pneumonia. The construction of random forest model included feature importance analysis and ROC curve assessment. Results No statistically significant differences were observed in age or gender distribution between the groups(P>0.05). The severe group showed significantly higher levels of fever duration, WBC, CRP, ESR, PCT, IL-6, LDH, and CK-MB than the non-severe group(P<0.05). The incidence of high fever and wheezing was also significantly higher in the severe group(P<0.05). Univariate logistic regression indicated that high fever, fever duration, wheezing, WBC, CRP, ESR, PCT, IL-6, LDH, and CK-MB were significantly associated with the risk of severe Mycoplasma pneumoniae pneumonia. The random forest model analysis revealed the lowest error rate when the number of decision trees was 45. The descending order of risk factor importance was fever duration, WBC, age, presence of high fever, CK-MB, ESR, CRP, RBC, IL-6, LDH, PCT, presence of wheezing, PLT, and gender. ROC curve analysis demonstrated an AUC of 0.882 of the random forest model, with a sensitivity of 75.68% and a specificity of 93.33%. Conclusion Through univariate Logistic regression and the random forest model, this study identifies fever duration, high fever, wheezing, and WBC as key risk factors for severe pediatric Mycoplasma pneumoniae pneumonia. These findings offer valuable reference for early clinical identification and intervention of severe Mycoplasma pneumoniae pneumonia, helping to improve the outcomes of affected children.

Key words: Severe pneumonia, Mycoplasma pneumoniae pneumonia, Risk factors, Random forest model, Logistic regression analysis, Child

CLC Number: