2005 HSR&D National Meeting Abstract
3089 — Practice- and Patient-Level Factors Predicting Colorectal Cancer Screening Rates
Yano EM (VA Greater Los Angeles HSR&D Center of Excellence)
Soban LM (VA Greater Los Angeles HSR&D Center of Excellence)
Etzioni DA (David Geffen-UCLA School of Medicine)
Parkerton PH (UCLA School of Public Health)
Colorectal cancer (CRC) imparts one of the highest levels of clinically preventable burdens of any disease process. Despite widespread recognition of the importance of early detection through screening, variation in CRC screening performance remains high. While some variation is patient-related (e.g., insurance, knowledge deficits, patient preferences), recent research has pointed to the contribution of practice-level factors. We explored the practice- and patient-level determinants of CRC screening variations to inform improved structuring of primary care (PC) and referral procedures.
We merged patient-level CRC screening data from OQP’s External Peer Review Program (EPRP) (FY01) with patient sociodemographics, comorbid conditions, and utilization data from the National Patient Care Database from 24 months prior to the index visit used for EPRP review. We clustered sampled patients within each VAMC, confirming their PC exposure using Austin visit patterns. We determined practice features (e.g., size, practice autonomy, resource sufficiency) to which individual veterans were exposed using the VA Survey of PC Practices (FY00) and location characteristics (e.g., urban/rural) from the Area Resource File (n=140 VAs). We used a random effects logistic model to evaluate the practice- and patient-level predictors of CRC screening among all eligible veterans, after eliminating cases with documented histories of CRC or IBS (n=37,935 patients).
Adjusting for practice- and patient-level covariates, the strongest predictors of CRC screening performance are higher PC visit rates (p<.0001), smaller facility size (p<.0001), primary care practice autonomy (p<.02), and sufficiency of clinical support arrangements (p<.02), in addition to more advanced patient age (p<.0001), greater income (p<.0001) and male gender (p<.004). Practice characteristics predicted a greater share of the variance than did patient-level predictors.
Primary care programs' configuration and content are strong predictors of CRC screening rates. Research is needed on patients' exposure to different practice types.
PC programs that have experienced resource drains in clinical support and declining autonomy over local primary care operations and referral procedures have significantly lower CRC screening performance. Large urban centers face unanticipated challenges in assuring adequate screening penetrance despite better-than-average access to GI specialists in part because of less organized PC delivery.