# Nomogram to predict severe Mycoplasma pneumoniae pneumonia in children

## TL;DR

This study created a tool to predict severe Mycoplasma pneumonia in children, helping doctors identify high-risk cases early.

## Contribution

A novel nomogram was developed and validated to predict severe Mycoplasma pneumoniae pneumonia in children.

## Key findings

- Fever duration, red blood cell count, and albumin were identified as independent predictors of severe MPP.
- The nomogram showed strong discrimination with AUC values of 0.8574 in training and 0.8147 in validation cohorts.
- Calibration and decision curve analysis confirmed the model's accuracy and clinical utility.

## Abstract

Mycoplasma pneumoniae pneumonia (MPP) is a prevalent community-acquired pneumonia in children, and severe MPP (SMPP) poses a prominent threat to pediatric health with rapid progression, high complication rates, and increased clinical management burden. Clinically, the capacity to identify children at high risk of SMPP remains inadequate. The aim of this study was to develop and validate a nomogram for predicting SMPP in children with MPP.

A total of 475 children with MPP admitted to Xuzhou Children’s Hospital from Jan. 2023 to Dec. 2024 were enrolled, meeting specific inclusion/exclusion criteria. They were categorized into severe MPP (SMPP, n = 151) and non-SMPP (n = 324) groups, then randomly split into training (n = 332) and validation (n = 143) cohorts at a 7:3 ratio. Demographic, clinical, laboratory data and derived inflammatory indicators were collected. LASSO and multivariate logistic regression were used to construct a nomogram, with ROC, calibration curves and DCA for evaluation. The study was ethically approved.

Using LASSO and multivariate logistic regression analyses, fever duration (OR = 1.271, P < 0.0001), red blood cell count (OR = 0.300, P = 0.0069) and albumin (OR = 0.795, P = 0.0002) were identified as independent predictors. The nomogram showed good discrimination (training cohort AUC=0.8574, 95%CI:0.8162–0.8986; validation cohort AUC=0.8147, 95%CI:0.7435–0.8859). The Hosmer-Lemeshow test yielded P = 0.551 in the training set and P = 0.553 in the validation set, and calibration curves in both cohorts confirmed excellent model fit, while DCA verified substantial clinical utility, supporting the nomogram’s clinical value in pediatric SMPP prediction

We developed and validated a practical, user-friendly nomogram for predicting SMPP in children with MPP, which could facilitate early identification and risk stratification of SMPP.