Predictive Model for Long-Term Coronary Artery Lesion Risk in Kawasaki Disease

Qianzi Ge developed a nomogram prediction model using clinical data from 2,481 pediatric Kawasaki disease patients admitted to the Children’s Hospital of Soochow University between July 2016 and December 2022. The model predicts long-term coronary artery lesion risk one year later based on factors like sex, hospitalization duration, fever duration, hemoglobin, hematocrit, and sodium levels. It achieved an area under the curve of 0.801 in the training dataset and 0.796 in validation.

Use Cases

• Train predictive models for coronary artery lesions based on clinical and laboratory variables like hemoglobin and sodium levels.
• Validate clinical risk scores for Kawasaki disease complications using retrospective patient data.
• Analyze the relationship between demographic factors (e.g., male sex) and long-term cardiovascular outcomes in pediatric patients.

Strengths

• Model performance is reported with specific metrics, including an AUC of 0.801 and sensitivity of 82.5% in the training set.
• The dataset is derived from a substantial cohort of 2,481 pediatric patients.
• The study period is clearly defined as July 2016 to December 2022.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the underlying data is unknown, which may limit suitability assessment.
• Data may reflect geographic and institutional bias inherent to a single hospital source.

Provenance

Source

Children’s Hospital of Soochow University

Collection Method

Retrospective analysis of clinical, laboratory, echocardiographic, and follow-up data.

Time Range

July 2016 to December 2022

Freshness

Last updated 2026-05-28 05:58:09; freshness should be verified.

Geography

Likely Suzhou, Jiangsu, China (based on hospital location).