Risk models to predict chronic kidney disease and its progression: a systematic review

Area of Science:

• Nephrology
• Epidemiology
• Biostatistics

Background:

• Chronic kidney disease (CKD) poses significant risks for cardiovascular and end-stage renal disease.
• Early identification and treatment are crucial for preventing CKD complications.
• Existing risk factors' utility in CKD prediction models needs clarification.

Purpose of the Study:

• To critically assess existing risk models for predicting CKD occurrence and progression.
• To evaluate the suitability of these models for clinical application.

Main Methods:

• Systematic literature search of MEDLINE and Embase (1980-2012).
• Dual review of studies on CKD prediction model development, validation, or impact.
• Data extraction on model characteristics, predictors, performance (discrimination, calibration), and validation.

Main Results:

• Included 30 CKD occurrence and 17 CKD progression risk scores.
• Most models demonstrated acceptable-to-good discrimination in derivation samples (AUC > 0.70).
• Limited external validation (8 occurrence, 5 progression models) showed modest-to-acceptable discrimination; impact studies are lacking.

Conclusions:

• Renal risk score development and clinical application are nascent.
• Existing CKD prediction models possess acceptable discriminatory performance.
• The practical impact of using these models requires further investigation.