The impact of guideline treatment nonadherence on survival for colorectal cancer patients: Propensity score calibration via a validation cohort.

308 Background: The quality of cancer care has been the focus of ongoing concern for cancer researchers, providers, and policy makers. The objectives of this study were: 1) to evaluate nonadherence with National Comprehensive Cancer Network treatment guidelines for colorectal cancer (CRC) patients and the impact on survival, and 2) to obtain error-corrected estimates of effect by means of propensity score calibration via a validation cohort.

METHODS: CRC patients identified by the Georgia Comprehensive Cancer Registry for the years 2000-07 were eligible (N = 18,388). Naïve propensity score (PSn) adjustment and PS calibration (PSC) via a validation cohort were utilized to obtain hazard ratio estimates for the impact of guideline treatment nonadherence on 5-year overall survival. The validation cohort contained additional information on comorbidity and payer status which was used to obtain error-corrected estimates of effect by PSC.

RESULTS: Treatment nonadherence conferred a large increased risk of death early in the follow-up period which declined over time (Table 1). Comparison of results from the PSn and PSC models indicated moderate to large bias due to unmeasured confounding in the PSn model (data not shown).

CONCLUSIONS: PSC produced attenuated estimates and had an impact on study conclusions in the latter follow-up period. For CRC patients, health services research into the quality of care received by cancer patients is necessary to continue the improving trend in CRC-related mortality. [Table: see text].