Risk Prediction of Carbapenem-Resistant Pseudomonas aeruginosa Infection in Children

1397 Pseudomonas aeruginosa isolates from children hospitalized between January 2020 and December 2024 were analyzed to build a risk prediction model. The XGBoost model, incorporating 10 key variables, achieved an AUC of 0.848 on an internal test set and 0.715 on a temporal validation cohort of 431 children from 2025-2026. The dataset was created by Lin Qin and is hosted on figshare.

Use Cases

• Training a risk prediction model based on clinical variables like hospital stay length and ventilator use.
• Analyzing risk factors for carbapenem-resistant Pseudomonas aeruginosa infection based on patient demographics and treatment history.
• Validating a machine learning model's generalizability using a temporal validation cohort.
• Interpreting feature importance for clinical prediction using SHAP analysis on variables like CD8+ T cells and prior carbapenem exposure.

Strengths

• Includes a temporal validation cohort of 431 children from 2025-2026 to test model generalizability.
• Model performance is quantified with specific AUC scores (0.848 internal, 0.715 temporal) and 95% confidence intervals.
• The derivation cohort contains 1397 isolates with clear resistance classifications (220 CRPA, 1177 CSPA).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the underlying patient-level data is unknown, which may limit suitability assessment.
• Data is sourced from a single children's hospital, which may limit geographic generalizability.

Provenance

Source

Children’s Hospital Affiliated to Shandong University.

Collection Method

Retrospective cohort study using electronic health records and antimicrobial susceptibility testing.

Time Range

Derivation cohort: January 2020 to December 2024. Validation cohort: January 2025 to March 2026.

Freshness

Last updated 2026-06-01 04:17:33; freshness should be verified.

Geography

Likely Shandong Province, China, based on the hospital affiliation.