[Risk factor analysis and prediction model construction for severe Chlamydia pneumoniae pneumonia in children]

Objective: To explore the risk factors for severe Chlamydia pneumoniae pneumonia (SCPP) in children and construct a predictive model. Methods: The retrospective cohort study included 179 children with Chlamydia pneumoniae pneumonia admitted to the Children's Hospital Affiliated to Zhengzhou University between January 2024 and June 2025. Their general information, clinical symptoms, and main laboratory indicators were collected. The included patients were randomly divided into a training cohort and an internal validation cohort at a ratio of 7:3, and were further categorized into SCPP and non-SCPP groups according to disease severity. In the training cohort, univariate analysis, Lasso regression and multivariate Logistic regression were used to screen the risk factors for SCPP in children, and a nomogram model was constructed. Model performance was validated in both the training and internal validation cohorts using receiver operating characteristic curve, calibration curve, Hosmer-Lemeshow test, decision curve analysis and clinical impact curve. Results: Of the 179 children with Chlamydia pneumoniae pneumonia, there were 106 males and 73 females, including 125 cases in the training cohort and 54 cases in the internal validation cohort. In the training cohort, there were 60 cases in the SCPP group (37 males and 23 females) and 65 cases in the non-SCPP group (39 males and 26 females). In the internal validation cohort, there were 25 cases in the SCPP group and 29 cases in the non-SCPP group. Multivariate Logistic regression analysis showed that co-infection (OR=2.66, 95%CI 1.04-6.79), fever duration (OR=1.94, 95%CI 1.07-3.54), erythrocyte sedimentation rate (OR=2.38, 95%CI 1.30-4.38), interleukin-8 (OR=2.04, 95%CI 1.17-3.55), lactate dehydrogenase (OR=1.84, 95%CI 1.06-3.19) were independent risk factors for SCPP in children (all P<0.05). The areas under the receiver operating characteristic curve were 0.88 (95%CI 0.82-0.94) and 0.83 (95%CI 0.73-0.94) in the training cohort and internal validation cohort, respectively. Calibration curve and Hosmer-Lemeshow test showed high consistency between model predictions and the actual situation (training cohort P=0.787; internal validation cohort P=0.680). Decision curve analysis revealed that the model yielded positive net benefit within the threshold probability range of 8%-90%. Clinical impact curve analysis confirmed that the model could accurately screen intervention subjects. Conclusions: Co-infection, fever duration, erythrocyte sedimentation rate, interleukin-8, and lactate dehydrogenase were identified as independent risk factors for SCPP in children. The constructed SCPP nomogram model for children based on the above risk factors has good predictive accuracy and provides support for clinicians to identify children with SCPP at an early stage.