Background: When a patient with prostate cancer needs radiation therapy (RT), their physician submits an order for RT to a prior authorization program, which assesses its concordance with clinical guidelines and evidence-based practices. After the order’s submission, a rule-based clinical decision support system (CDSS) incorporating the latest clinical evidence evaluates whether the order appears appropriate or potentially nonindicated. If potentially nonindicated, a board-certified oncologist will discuss the order with the ordering physician. The discussion concludes with the order being authorized, modified, withdrawn, or recommended for denial. Although the patient’s race is not captured by the program, bias prior to ordering, while ordering, or during the discussion may influence outcomes.
Objectives: To evaluate whether there is an association between a patient’s race and a prostate RT order’s disposition by the CDSS at the beginning of prior authorization, as well as by the overall prior authorization program incorporating the CDSS.
Methods: Orders for prostate RT placed in 2019 were analyzed. All pertained to patients with Medicare Advantage health plans from 1 national organization. Patient race data from the Centers for Medicare & Medicaid Services was appended to the order data. Chi-square tests were used to evaluate univariate associations. Multivariate logistic regression was used to assess whether an association existed between patient race (black vs nonblack) and order disposition after controlling for the patient’s age, urbanicity, the median income in the patient’s home zip code, and the region in which the patient lived.
Results: Black patients accounted for 860 (25%) of the 3436 orders included in the analysis. Among orders pertaining to black patients, 301 of 860 (35.0%) were deemed appropriate by the CDSS, versus 918 of 2576 (35.6%) orders for nonblack patients, an insignificant difference (P = .77). The ultimate approval rates were also similar; 813 of 860 (94.5%) orders for black patients were approved, versus 2401 of 2576 (93.2%) for nonblack patients, an insignificant difference (P = .20). Race had no association with rule-based determinations or final prior authorization dispositions. Black patients had 1.15 adjusted odds (95% CI, 0.93-1.44) of having their order approved by the CDSS, and 1.07 adjusted odds (95% CI, 0.71-1.60) of having their order authorized by the prior authorization program overall, relative to nonblack patients. None of the control variables examined (age, median income in the patient’s home zip code, urbanicity, and region) had a significant association with the disposition of the CDSS or the prior authorization program overall.
Conclusions: Prior authorization was found to produce outcomes that, when combined with retrospective race determination, revealed similar clinical appropriateness of orders for black and nonblack populations. Rule-based CDSSs may be a means of ensuring that patients equally receive guideline-based care considering the latest scientific evidence, and that guidelines are enforced without racial bias. However, while fewer than half of orders were deemed appropriate by the rules-based CDSS, the majority were approved by the overall prior authorization program, which included a physician reviewer. No evidence was found suggesting that combining CDSS with physician review increased bias.