BD-KDD: Kidney Disease Diagnosis with 988 Clinical Patient Records

BD-KDD contains 988 patient records for kidney disease diagnosis, featuring 26 clinical and laboratory variables. Developed by Md. Masudul Islam and hosted on Harvard Dataverse, the dataset provides binary classification labels for healthy and diseased states. It was specifically designed to double the sample size of the standard UCI Chronic Kidney Disease benchmark.

Use Cases

• Predicting kidney disease status using laboratory biomarkers like Serum Creatinine (Sc) and Blood Urea (Bu)
• Analyzing the correlation between Hemoglobin (Hemo) levels and renal health classification
• Evaluating the diagnostic importance of Specific Gravity (Sg) in identifying kidney dysfunction

Strengths

• 988 patient records
• 26 clinical variables including 24 laboratory biomarkers
• Balanced class distribution with 481 healthy and 507 kidney disease instances

Limitations

• Retrospective clinical data from a single diagnostic center source
• Geographic origin of the patient population is not explicitly defined
• Lack of longitudinal data for disease progression tracking

Provenance

Source

Md. Masudul Islam, Harvard Dataverse

Collection Method

Retrospective clinical data obtained from diagnostic laboratory records

Freshness

Last updated March 2026.