Variability of preoperative breast MRI utilization among older women with newly diagnosed early-stage breast cancer.

While magnetic resonance imaging (MRI) is frequently used following breast cancer diagnosis, routine use of breast MRI for preoperative evaluation remains contentious. We identified factors associated with preoperative breast MRI utilization and investigated the variation among physicians. We used the surveillance, epidemiology, and end Results (SEER)-Medicare linked database to analyze the preoperative breast MRI utilization among patients with stage 0, I, or II breast cancer diagnosed between 2002 and 2007. Multilevel logistic regression models were used to identify patient- and physician-level predictors of preoperative MRI utilization. Of 56,743 women with early-stage breast cancer who were treated with surgery and evaluated by a preoperative mammogram and/or ultrasound during the study period, 8.7% (n = 4,913) received preoperative breast MRI. While patient and tumor characteristics did predict preoperative breast MRI utilization, they explained only 15.4% of the variation in utilization rates. Differences in preoperative breast MRI utilization across physicians were large, after controlling patient-level factors and physicians' volumes. Accounting for clustering of patients within individual physicians (n = 3,144), the multilevel logistic regression models explained 36.4% of variation. The median odds ratio of 3.2, corresponding with the median value of the relative odds of receiving preoperative breast MRI between two randomly chosen physicians, indicated a large individual physician effect. Our study found that preoperative breast MRI has been adopted rapidly and variably. Although patient characteristics were associated with preoperative breast MRI utilization, physician practice was a major determinant of whether women received preoperative breast MRI. Future studies should evaluate whether routine use of preoperative breast MRI in newly diagnosed early-stage breast cancer improves clinical outcomes.