Validating a case definition for chronic kidney disease using administrative data.

Area of Science:

• Nephrology
• Health Informatics
• Public Health Surveillance

Background:

• Administrative data are frequently utilized for monitoring chronic medical conditions.
• The study addresses the common practice of using administrative data for disease surveillance.

Purpose of the Study:

• To evaluate the accuracy of an algorithm derived from administrative data for identifying chronic kidney disease (CKD).
• To compare the validity of administrative data algorithms against the reference standard of estimated glomerular filtration rate (eGFR).

Main Methods:

• Adults from Alberta with at least two serum creatinine measurements within a year were identified.
• Validity indices (sensitivity, specificity, PPV, NPV) were calculated for CKD using administrative data (physician claims, hospital abstracts) over three years.
• Algorithms were validated against two CKD definitions based on eGFR (<60 mL/min/1.73m² or mean eGFR <30 mL/min/1.73m²).

Main Results:

• A total of 321,293 subjects were analyzed.
• Sensitivities for CKD detection using administrative codes were generally low.
• An algorithm using two physician claims or one hospitalization within two years showed 19.4% sensitivity and 97.2% specificity for CKD (eGFR <60).
• Positive predictive values (PPVs) were consistently below 50% when using a mean eGFR <30 mL/min/1.73m² as the reference standard.

Conclusions:

• Administrative data exhibit insufficient sensitivity and PPV for effective CKD surveillance.
• These data may be valuable for research requiring highly specific case-finding algorithms.
• The findings highlight limitations in using administrative data for public health surveillance of CKD.