Validity of administrative database coding for kidney disease: a systematic review.

BACKGROUND Information in health administrative databases increasingly guides renal care and policy. STUDY DESIGN Systematic review of observational studies. SETTING & POPULATION Studies describing the validity of codes for acute kidney injury (AKI) and chronic kidney disease (CKD) in administrative databases operating in any jurisdiction. SELECTION CRITERIA After searching 13 medical databases, we included observational studies published from database inception though June 2009 that validated renal diagnostic and procedural codes for AKI or CKD against a reference standard. INDEX TESTS Renal diagnostic or procedural administrative data codes. REFERENCE TESTS Patient chart review, laboratory values, or data from a high-quality patient registry. RESULTS 25 studies of 13 databases in 4 countries were included. Validation of diagnostic and procedural codes for AKI was present in 9 studies, and validation for CKD was present in 19 studies. Sensitivity varied across studies and generally was poor (AKI median, 29%; range, 15%-81%; CKD median, 41%; range, 3%-88%). Positive predictive values often were reasonable, but results also were variable (AKI median, 67%; range, 15%-96%; CKD median, 78%; range, 29%-100%). Defining AKI and CKD by only the use of dialysis generally resulted in better code validity. The study characteristic associated with sensitivity in multivariable meta-regression was whether the reference standard used laboratory values (P < 0.001); sensitivity was 39% lower when laboratory values were used (95% CI, 23%-56%). LIMITATIONS Missing data in primary studies limited some of the analyses that could be done. CONCLUSIONS Administrative database analyses have utility, but must be conducted and interpreted judiciously to avoid bias arising from poor code validity.