Linezolid Plasma Concentration Predictions for Critically Ill Pediatric Patients

A retrospective cohort of 145 critically ill pediatric patients provided 213 samples for this study. The dataset includes 32 features, such as pharmacokinetic parameters, and the target variable of linezolid plasma concentration. Author Yanping Zhang published the data on figshare under a CC-BY-4.0 license, with a last update in June 2026.

Use Cases

• Training predictive models for drug concentration based on patient features like clearance and weight.
• Evaluating machine learning algorithm performance for clinical pharmacokinetic tasks.
• Conducting feature importance analysis for drug exposure predictors using methods like SHAP.
• Simulating dosing regimen optimization for specific minimum inhibitory concentration targets.

Strengths

• Dataset is based on 145 patient records with 213 samples, providing a clinical foundation.
• Includes 32 features derived from a population pharmacokinetic model and patient data.
• The best-performing LightGBM model achieved an R² of 0.777 on the testing set.
• Clear methodological description of integrating PopPK with machine learning algorithms.

Limitations

• Row count for the underlying data table is unknown, limiting scale assessment.
• Reported model performance metrics are described as optimistic estimates due to potential data leakage.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

figshare, author Yanping Zhang.

Collection Method

Retrospective cohort study of critically ill pediatric patients.

Freshness

Last updated 2026-06-02 05:39:35; freshness should be verified.