The steps of constructing and validating an algorithm to identify chronic kidney disease patients in medical...

Area of Science:

• Nephrology
• Health Informatics
• Epidemiology

Background:

• Chronic kidney disease (CKD) poses a significant global health challenge, marked by increased mortality, morbidity, and economic costs.
• CKD is often asymptomatic and difficult to detect in medical-administrative databases due to the lack of laboratory results and specific diagnostic codes.
• Accurate identification of CKD cases in large-scale databases is crucial for public health surveillance and resource allocation.

Purpose of the Study:

• To describe the development and validation of an algorithm for identifying chronic kidney disease (CKD) within the French national health insurance information system (SNDS).
• To assess the feasibility of using healthcare claims data to detect CKD cases, particularly those that are asymptomatic or undiagnosed.
• To establish a foundation for improving CKD case detection through expert-driven and machine learning-based approaches.

Main Methods:

• A consortium of experts in nephrology, kidney epidemiology, and healthcare claims databases designed the RENALGO-EXPERT algorithm.
• The algorithm combines various indicators associated with the CKD care pathway to estimate the likelihood of CKD.
• Algorithm performance was evaluated across different populations and databases, assessing its sensitivity and specificity.

Main Results:

• The RENALGO-EXPERT algorithm demonstrated variable performance depending on the specific population and database utilized.
• Sensitivity for CKD detection showed improvement in higher-risk populations.
• Current results indicate that the algorithm's performance is not yet satisfactory for optimal case detection.

Conclusions:

• The RENALGO-EXPERT algorithm represents a step towards identifying CKD in administrative databases but requires further refinement.
• Machine learning methods (RENALGO-IA project) are being explored to enhance case detection and capture subtle CKD signals.
• Continued development is necessary to improve the accuracy and reliability of CKD identification in large healthcare datasets.