Table 1_Identification of BMI-related high-risk feature combinations for diabetes among yo

103,693 young adults with normal baseline fasting plasma glucose were tracked for incident diabetes over a median follow-up of 2.99 years. The data, sourced from the Rich Healthcare Group health check-up database in China, was analyzed using Cox regression and machine learning models like XGBoost. SHAP analysis identified BMI as the most influential predictor, revealing a nonlinear risk increase beyond approximately 28 kg/m².

Use Cases

• Training diabetes risk prediction models based on BMI and metabolic features.
• Analyzing nonlinear risk patterns for BMI using restricted cubic spline analysis.
• Investigating joint contribution patterns of risk factors like triglycerides and systolic blood pressure with SHAP.
• Comparing performance of logistic regression and XGBoost models on low-event-rate clinical data.

Strengths

• Large cohort of 103,693 participants with complete BMI data.
• Longitudinal design with a median follow-up of 2.99 years and 266 incident diabetes events.
• Analysis includes both conventional Cox regression and multiple machine learning models (logistic regression, XGBoost).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The low PR-AUC values reported indicate limited ability to identify true positive cases under the very low event rate.

Provenance

Source

Rich Healthcare Group health check-up database in China.

Collection Method

Cohort study from health check-up data.

Freshness

Last updated 2026-05-15 04:21:40; freshness should be verified.

Geography

China.