Background: When a physician determines that a patient with breast cancer needs radiation therapy (RT), they submit an RT order to a prior authorization program which assesses guideline-concordance. An up-to-date, evidence-informed, rule-based clinical decision support system (CDSS) evaluates whether the order is appropriate or potentially non-indicated. If potentially non-indicated, a board-certified oncologist discusses the order with the ordering physician. After discussion, the order is authorized, modified, withdrawn, or recommended for denial. Although the patient’s race is not captured by the program, bias prior to and during ordering, or during the discussion, may influence outcomes. This study evaluates the association between patient race and order determination by both the CDSS and the overall prior authorization program. Methods: We assessed orders pertaining to female patients who were enrolled in a Medicare Advantage plan from one national organization and had their breast RT order placed in calendar year 2019. Centers for Medicare & Medicaid Services patient race data was appended to the order data, as the orders did not include race. We used multivariate logistic regression to examine the association between patient race (Black, White, Other) and order determination, controlling for age, the urbanicity and median income of the patient’s ZIP code, and whether the patient had breast magnetic resonance imaging (MRI) prior to the RT order. Results: We identified 3,043 orders for inclusion in analysis. Black patients accounted for 590 (19%) and White patients for 2,371 (78%) of the orders analyzed. There was no association between race and CDSS determinations or prior authorization determinations. The CDSS was significantly less likely to deem orders appropriate for older patients (OR: 0.97; CI: 0.96-0.98), urban patients (OR: 0.75; CI: 0.61-0.90), and patients with a history of breast MRI (OR: 0.66; CI: 0.54-0.81). Orders for patients from communities with median income > $60,000 were more likely determined appropriate (OR: 1.21; CI: 1.02-1.45). Orders were less likely to be approved by prior authorization if they pertained to older patients (OR: 0.96; CI: 0.94-0.98). Conclusions: Prior authorization made similar determinations of the clinical appropriateness of orders regardless of patient race. No evidence was found to suggest that the prior authorization process was a source of bias.
HSR22-160 Table: Multivariate Logistic Regression Results