Development of a predictive model based on clinical indicators for refractory Mycoplasma pneumoniae pneumonia in children: a case-control study.
This study aims to screen indicators for predicting the occurrence of refractory Mycoplasma pneumoniae pneumonia (RMPP) in children, determine the combined factors for predicting RMPP, and provide a basis for the early identification of children with RMPP and the determination of treatment plans.
This study was a retrospective case-control analysis. A total of 522 children with MPP and 28 clinical indicators were included. Clinical feature, hospitalization period, laboratory data, etc., were collected. The risk factors related to RMPP were screened through univariate analysis. A multivariate logistic regression model was established, and stepwise regression was used to screen out independent risk factors. The operating characteristic curve (ROC) of the combined predictor was drawn for predictive efficacy analysis. A visual nomogram model for predicting the probability of RMPP occurrence was constructed and validated.
Differing from other results, there were no statistically significant differences in demographic indicators such as age and gender between the two groups. The multivariate logistic regression analysis showed that duration of fever (OR = 1.407), PLT (OR = 0.997), pleural effusion (OR = 2.084), atelectasis (OR = 3.116), and extrapulmonary complications (OR = 4.251) were independent risk factors for RMPP (P < 0.05). MP antibody titer ≥1:320 (OR = 0.420) is a protective factor. The AUC of the prediction model was 0.870 (95%CI: 0.837, 0.904), the sensitivity of the prediction model was 82.2%, the specificity was 80.5%, and the prediction accuracy was relatively high. The calibration curve, close to the 45° line, exhibited good calibration.
We constructed and validated a visual and user-friendly model for individualized prediction of RMPP risk in children at initial presentation, to support clinical decision-making regarding macrolide therapy. This model provides a tool for identification of high-risk children, which may inform closer monitoring and prompt consideration of adjunctive therapies.
This study was a retrospective case-control analysis. A total of 522 children with MPP and 28 clinical indicators were included. Clinical feature, hospitalization period, laboratory data, etc., were collected. The risk factors related to RMPP were screened through univariate analysis. A multivariate logistic regression model was established, and stepwise regression was used to screen out independent risk factors. The operating characteristic curve (ROC) of the combined predictor was drawn for predictive efficacy analysis. A visual nomogram model for predicting the probability of RMPP occurrence was constructed and validated.
Differing from other results, there were no statistically significant differences in demographic indicators such as age and gender between the two groups. The multivariate logistic regression analysis showed that duration of fever (OR = 1.407), PLT (OR = 0.997), pleural effusion (OR = 2.084), atelectasis (OR = 3.116), and extrapulmonary complications (OR = 4.251) were independent risk factors for RMPP (P < 0.05). MP antibody titer ≥1:320 (OR = 0.420) is a protective factor. The AUC of the prediction model was 0.870 (95%CI: 0.837, 0.904), the sensitivity of the prediction model was 82.2%, the specificity was 80.5%, and the prediction accuracy was relatively high. The calibration curve, close to the 45° line, exhibited good calibration.
We constructed and validated a visual and user-friendly model for individualized prediction of RMPP risk in children at initial presentation, to support clinical decision-making regarding macrolide therapy. This model provides a tool for identification of high-risk children, which may inform closer monitoring and prompt consideration of adjunctive therapies.