Validating a case definition for chronic kidney disease using administrative data

Background. Administrative data are commonly used for surveillance of chronic medical conditions. The purpose of this study was to determine the validity of an algorithm derived from administrative data for identifying chronic kidney disease (CKD) compared to the reference standard of estimated glomerular filtration rate (eGFR). Methods. We identified adults from the province of Alberta with at least two outpatient serum creatinine measurements within a 1-year time period. Validity indices were estimated for CKD using up to 3 years of administrative data (physician billing claims and hospital discharge abstracts) for various case-definition combinations. For each algorithm, the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated against two reference standard definitions of CKD (two eGFR measurements < 60 mL/min/1.73m(2) or mean eGFR < 30 mL/min/1.73m(2)). Results. A total of 321 293 eligible subjects were identified. Irrespective of the algorithm, sensitivities for defining CKD (eGFR < 60 mL/min/1.73m(2)) using administrative codes were low. A case-definition algorithm employing two physician claims or one hospitalization within a 2-year period had sensitivity of 19.4%, specificity of 97.2%, PPV of 60.1% and NPV of 84.8% for detecting CKD. Estimates of sensitivity were higher when < 30 mL/min/1.73m(2) was used as the reference standard, although PPVs were lower and consistently less than 50%. Conclusion. These results, using eGFR as a reference standard, suggest that administrative data have insufficient sensitivity and PPV for CKD surveillance, although they may be useful when highly specific algorithms are required for research purposes.