Data Sheet 1_Development and validation of interpretable machine learning models for predi

A retrospective study from Zhongshan Hospital, Xiamen University (January 2022–January 2025) developed interpretable machine learning models for stroke prediction. The dataset includes 82 stroke cases and 164 matched controls among non-valvular atrial fibrillation patients with low CHA₂DS₂-VA scores. Data encompasses demographics, comorbidities, laboratory markers, and echocardiographic parameters.

Use Cases

• Training machine learning models for stroke risk stratification based on clinical and biomarker variables mentioned in the description.
• Evaluating feature importance for stroke prediction based on variables like NT-proBNP, CRP, and E/e' ratio.
• Comparing the performance of logistic regression, random forest, and XGBoost algorithms on a clinical dataset.
• Developing personalized anticoagulation strategy tools for low-to-moderate risk atrial fibrillation patients.

Strengths

• Includes 246 total patient records (82 cases, 164 controls) with matched design.
• Models achieved an AUC of 0.906 in the test set, indicating strong predictive performance.
• Feature selection identified seven key predictors, including age, CRP, E/e' ratio, LVEF, NT-proBNP, triglycerides, and white blood cell count.
• Dataset is licensed under CC-BY-4.0, allowing for open reuse and modification.

Limitations

• Row count for the underlying data tables is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data is from a single-center study, which may limit generalizability to other populations.

Provenance

Source

Zhongshan Hospital, Xiamen University.

Collection Method

Single-center, retrospective study.

Time Range

January 2022 to January 2025.

Freshness

Last updated 2026-04 04:24:41; freshness should be verified.

Geography

Likely China, based on the hospital location.