3113 — An Algorithm for Constructing Longitudinal Data on Colorectal Cancer Screening among Non-Elderly Veterans

Bian J (Atlanta VAMC), Gillespie TW (Atlanta VAMC and Emory University), Halpern MT (American Cancer Society), Jasien C (Atlanta VAMC), Lipscomb J (Atlanta VAMC and Emory University)

Objectives:
Longitudinal data analyses that minimize omitted variable biases may be more attractive than cross-sectional analyses in analyses conducted for causal inferences. This study is to construct a longitudinal dataset for examining colorectal cancer (CRC) screening patterns among veterans receiving care at VAMCs.

Methods:
We used 1999-2006 VHA Medical SAS outpatient data to construct an individual veteran-level longitudinal dataset, consisting of veterans in the 4 index years from 2003-2006. Types of screenings examined were fecal occult blood test (FOBT), sigmoidoscopy, double-contrast barium enema (DCBE), and colonoscopy. To determine individual veteran’s screening status in a given index year based on VA-accepted CRC screening guidelines, we used a combination of procedure and diagnosis codes from a 5-year period (including the index and 4 precious years) to account for variation in recommended CRC screening interval by type of screening. Veterans eligible for screening included men age 52-63 years who had at least one primary care outpatient visit during the index year from 2003-2006. We further excluded those who were diagnosed with colorectal cancer or had a history of surveillance diagnoses in outpatient visits during the 5-year period. Eligible veterans were determined to be screened for the index year if they received FOBT during the index year or received sigmoidoscopy, DCBE, and/or colonoscopy any time during the 5-year period.

Results:
The number of these veterans nationwide eligible for screening from 2003-2006 was 1,130,806, 1,215,721, 1,274,841, and 1,313,306; and the corresponding proportions screened were 36.7%, 39.4%, 39.5%, and 37.8%, respectively. There was substantial variation in the proportion screened across the 21 VISNs, with per-VISN ranges of 27.6-41.9% in 2003 and 29.6-43.9% in 2006. The proportion screened increased with older age.

Implications:
It is feasible to construct s longitudinal dataset to profile CRC screening utilization over time at the individual veteran, VAMC, or VSIN level.

Impacts:
Compared with cross-sectional data, longitudinal data may provide a better approach for assessing VA adherence to colorectal cancer screening guidelines and for investigating individual-level or organizational characteristics influencing screening utilization.