Evaluating colorectal cancer care in the Veterans Health Administration: How good are the data?

Objective: Longitudinal observational data sources can be useful for studies of health care use and outcomes among patients treated in the usual care setting. In this study, we sought to determine the quality of a Veterans Health Administration (VHA) observational dataset for ascertaining of colorectal cancer (CRC) diagnoses and selected therapies among veterans. Materials and methods: We determined (1) the completeness of Veterans Affairs Central Cancer Registry (VACCR) CRC case ascertainment by determining the proportion of 7224 probable CRC cases identified from VHA administrative data who also appeared in the VACCR; and (2) the accuracy of VHA administrative data regarding CRC patients' utilization of chemotherapy and radiation therapy (RT) through comparison to gold-standard medical record data on a different sample of patients (N=295). Results: Among probable cases of CRC identified through administrative data, the proportion ascertained by the VACCR was 0.89 (95% confidence interval [CI]: 0.88-0.90). The VACCR-Administrative data were highly accurate in measuring chemotherapy (sensitivity = 0.99, 95% CI: 0.96-1.00; specificity=0.92, 95% CI: 0.86-0.96) and RT (sensitivity = 1.00, 95% CI: 0.96-1.00; specificity = 0.88, 95% CI 0.80-0.94) use among veterans with CRC. Conclusions: The clinically rich patient-level longitudinal VACCR-Administrative dataset is appropriate for empirical studies of chemotherapy and RT-related processes of care for veterans with CRC. (C) 2011 Elsevier Ltd. All rights reserved.