Evaluating an Electronic Measure of Colorectal Cancer Screening at Indian Health Service Facilities, 2008-2010.

Background: Colorectal cancer (CRC) is a leading cause of cancer mortality in American Indian and Alaska Native (AIAN) people, and incidence rates vary considerably among AIAN populations throughout the United States. Screening has the potential to prevent CRC deaths by detection and treatment of early disease or removal of precancerous polyps. Surveillance of CRC screening is critical to efforts to improve delivery of this preventive service, but existing CRC screening surveillance methods for AIAN are limited. The Government Performance and Results Act (GPRA) CRC screening clinical care measure provides data on CRC screening among AIAN populations.

Purpose: The aim of this study was to evaluate the accuracy of the GPRA measure for CRC screening (sensitivity, specificity, positive predictive value and negative predictive value), determine reasons for CRC screening misclassification (procedures noted as screening when they were actually diagnostic exams), and to suggest opportunities for improving surveillance for CRC screening nationwide for AIAN populations.

Methods: Medical record reviews (paper and electronic) were compared to the GPRA-reported CRC screening status for 1,071 patients receiving care at tribal health facilities. A total of 8 tribal health facilities (2 small, 3 medium, and 3 large) participated in the study from the Pacific Coast, the Southwest, the Southern Plains, and Alaska IHS regions. Screening-eligible patients were identified using queries of the local electronic health record from January 2007 to December 2008, and medical chart reviews were completed at participating facilities from September 2008 to June 2010.

Results: Among 545 patients classified as screened by the GPRA measure, 305 (56%, CI: 52%-60%) had a false positive for screening as compared with medical record review. The overall sensitivity of the GPRA measure for CRC screening was 93% (CI=89%-95%) while specificity was 62% (CI: 59%-66%). The most common reasons for misclassification were for diagnostic or surveillance tests to be recorded as screening (67%), as well as medical record miscoding (18%) due to miscoding, charting errors, screenings performed outside the IHS, testing for a non-screening purpose, and categorization of patients as screened when a test had been ordered but not actually completed.

Conclusions: This study found that the GPRA CRC screening clinical measure overestimates the true screening rate due to the inclusion of diagnostic and surveillance exams, especially colonoscopy, as well as misclassification errors. The results of this study suggest a need to more accurately use the ICD-9 diagnostic code V76.51, which was associated with frequent coding errors. In combination with other programmatic efforts that focus on screening average- risk, asymptomatic American Indian and Alaska Native persons, improving the coding used for CRC screening may help to more accurately detect decreases in AIAN CRC incidence and mortality.