CKD Risk Prediction in Hyperglycemic Adults from a Prospective Community Cohort

A dataset supporting a machine learning model for predicting chronic kidney disease progression in elderly adults with hyperglycemia. The study followed TRIPOD+AI guidelines, using data from four community sites for training and validation. The XGBoost model achieved AUCs of 0.905, 0.809, and 0.837 on training, internal test, and external validation sets.

Use Cases

• Developing CKD risk prediction models based on clinical features like serum creatinine, age, and hemoglobin.
• Validating machine learning models for clinical use with internal and external test sets.
• Conducting feature importance analysis for clinical predictors using SHAP values.
• Building risk stratification tools for community health screening programs.

Strengths

• Model performance is validated with reported AUCs of 0.905, 0.809, and 0.837 across three datasets.
• Risk stratification yielded distinct CKD incidences of 0.7%, 9.9%, and 55.5% across low, medium, and high-risk groups.
• Study design followed TRIPOD+AI guidelines and used data from four community sites for robust validation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the four specific community sites used.

Provenance

Source

figshare

Collection Method

Multicenter prospective community cohort study.

Freshness

Last updated 2026-04-24 05:49:13; freshness should be verified.

Geography

Data collected from four community sites; specific locations are not stated.