Complete diagnostic evaluation, or CDE (i.e., a colonoscopy or combined barium enema X-ray and flexible sigmoidoscopy) is recommended for individuals who have an abnormal screening fecal occult blood test result. Accurate measures of CDE use are needed in colorectal cancer (CRC) screening programs. This study compares the sensitivity and specificity of different methods for measuring CDE recommendation and performance. We identified 17 primary-care practices with 120 patients who had a positive fecal occult blood test result in a CRC screening program operated by a managed-care organization. Approaches used to measure CDE recommendation and performance included external chart audit (ECA) only; internal chart audit (ICA) only; administrative data review (ADR) of electronic claims data; ICA plus ADR; and ECA plus ADR (the "gold standard"). Sensitivity and specificity of each method were assessed relative to CDE recommendation and performance as measured by ECA plus ADR. For CDE recommendation, sensitivity measures were ECA only, 0.926; ICA only, 0.790; ADR only, 0.617; and ICA plus ADR, 0.901. The specificity of each method for CDE recommendation was no less than 0.95. In terms of CDE performance, sensitivity measures were ECA only, 0.877; ICA only, 0.790; ADR only, 0.877; and ICA plus ADR, 0.965. The specificity of each method for CDE performance was 1.0. The ICA-plus-ADR method was a highly sensitive and specific measure of CDE use. This method should be considered in situations that involve primary-care physician follow-up of patients with abnormal CRC screening test results.