OBJECTIVES: Testing the validity of ICD-9-CM codes to identify diabetic patients with stage 3-5 chronic kidney disease (CKD) in claims data. METHODS: We conducted a cross-sectional study using claims from a large commercial health plan. Patients meeting the following criteria were included: 1) diagnosis of type 2 diabetes; 2) ≥12 months of continuous health plan eligibility; and 3) ≥2 non-zero serum creatinine lab values, at least 90 days apart, between January 1, 2004 and June 30, 2011. We identified 12 ICD-9-CM code groups potentially indicative of CKD stage 3-5 and validated them against a “gold standard”, defined as two laboratory claims, at least 90 days apart, with an estimated glomerular filtration rate (eGFR) <60 mL/min. eGFR was calculated using the CKD Epidemiology Collaboration equation and the Modification of Diet and Renal Disease (MDRD) equation. We calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value for the selected codes. Exact binomial 95% confidence intervals (CIs) were derived, and codes with a PPV whose lower CI bound was ≥ 80% were considered valid. RESULTS: The study sample consisted of 383,970 patients. Approximately 16% of the sample (N=61,052) had stage 3-5 CKD based on the “gold standard”. ICD-9-CM codes for chronic renal failure, stage 3-5 (585.3-5), had a PPV of 84.2% (CI: 83.7% - 85.7%).  ICD-9-CM code 585.6, used for end-stage renal disease, had a PPV of 84.7% (CI: 83.6% - 85.7%). ICD-9-CM code 285.21, used to describe anemia in CKD, had a PPV of 85.4% (CI: 84.6% - 86.2%). For the remaining code groups, PPV ranged from 50% to 78%, with CIs of ± 2 percentage points. Similar results were obtained when eGFR was calculated using the MDRD equation. CONCLUSIONS: This cross-sectional validation study suggests that diabetic patients with stage 3-5 CKD can be accurately identified in administrative claims data using selected ICD-9-CM codes.