Machine Learning Prediction Models for Severe Pediatric Pneumonia in China

A systematic review and meta-analysis of 13 studies evaluating machine learning models for predicting severe Mycoplasma pneumoniae pneumonia in Chinese children. The dataset, authored by Juan Cao and published on figshare in June 2026, synthesizes model performance metrics and methodological characteristics. Reported model discrimination, measured by the area under the receiver operating characteristic curve, ranged from 0.81 to 0.90.

Use Cases

• Benchmarking new predictive models against reported AUC values ranging from 0.81 to 0.90.
• Identifying commonly used predictors for severe pediatric pneumonia, such as age, insulin use, and body mass index.
• Assessing methodological quality and risk of bias in clinical prediction model studies using the PROBAST tool.
• Comparing the performance of different modeling algorithms, such as XGBoost and random forest, for this clinical task.

Strengths

• Includes data from 13 distinct studies, providing a multi-study synthesis.
• Quantitatively synthesizes model discrimination using pooled area under the receiver operating characteristic curve.
• Applies a standardized methodological quality assessment tool (PROBAST).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The data is a meta-analysis of other studies, not primary patient-level data.

Provenance

Source

Juan Cao via figshare.

Collection Method

Systematic literature review and meta-analysis of studies from PubMed, EMBASE, Web of Science, Cochrane Library, CNKI, and Wanfang databases.

Time Range

Literature search covered from database inception to November 2025.

Freshness

Last updated 2026-06-03 05:41:49; freshness should be verified.

Geography

Studies focused on Chinese pediatric populations.