Geospatial Disparities in the Treatment of Curable Breast Cancer Across the US

Abstract

Importance: Patient factors help explain disparities in breast cancer treatments and outcomes.

Objective: To determine the extent to which geospatial variation in initial breast cancer care can be attributed to region vs patient factors with the aim of guiding quality improvement efforts.

Design, setting, and participants: This was a retrospective population-based cohort study from January 1, 2007, through December 31, 2016, using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database that included 31 571 patients diagnosed with stage I to III breast cancer from 2007 through 2013. Five metrics of care delivery were defined: stage I at diagnosis, chemotherapy receipt, radiation therapy receipt, endocrine therapy (ET) initiation (year 1), and ET continuation (years 3-5). Data analysis was performed from January to June 2021.

Exposures: Stage I diagnosis and treatment with chemotherapy, radiation therapy, or ET.

Main outcomes and measures: For each metric, total variance was attributed proportionally to 4 domains-random, patient factors (eg, age, race and ethnicity, socioeconomic status), region (health service area [HSA]), and unexplained-using hierarchical multivariable modeling.

Results: Of 31 571 total patients (median [IQR] age, 71 [68-75] years), 19 391 (61.4%) had stage I disease at diagnosis. Among eligible patients, 17 297 of 21 190 (81.6%) received radiation therapy, 7204 of 9903 (72.8%) received chemotherapy, 13 115 of 26 855 (48.8%) initiated ET, and 13 944 of 26 855 (52.1%) continued ET. Geospatial density (ie, heat) maps highlight regional performance patterns. For all 5 metrics, region/HSA explained more observed variation (24%-48%) than patient factors (1%-4%); the largest share of variation was unexplained (35%-54%). The metrics with the largest proportion of total variance attributed to region/HSA were ET initiation and continuation (28% and 39%, respectively).

Conclusions and relevance: In this cohort study, there was substantial unexplained geospatial variation in initial breast cancer care. The variance attributed to region/HSA was multifold larger than that explained by patient factors. The importance of patient factors such as race and ethnicity notwithstanding, future quality improvement efforts should focus on reducing unwarranted geospatial variation, especially including optimizing the delivery of ET in low-performing regions.

Figure 1.. Performance by Health Service Area (HSA)

Geospatial heat maps display the unadjusted concordance for 4 metrics of breast cancer in the HSAs represented within the Surveillance, Epidemiology, and End Results–Medicare catchment area (HSAs with <30 patients with breast cancer were excluded). The 4 metrics were stage I at diagnosis (proxy for screening mammography and timely surgery) (A); receipt of chemotherapy (B); receipt of radiation therapy (C); and initiation of endocrine therapy within 1 year of diagnosis (D). The map for continuation of endocrine therapy 3 to 5 years after diagnosis was similar to the map for initiation of endocrine therapy, so it was not displayed. We used the same color-coded absolute 10-point performance thresholds across all 4 maps to permit comparisons within and across maps.

Figure 2.. Total Absolute and Proportional Attributable Variance

For each of 5 breast cancer care metrics, we decomposed the total attributable variance into 4 domains: randomness (assuming that the number of patients within each health service area [HSA] receiving concordant care follows a binomial distribution), measured patient factors (age, race and ethnicity, urbanity, socioeconomic status, education, and comorbidity score), region (HSA level), and unexplained. The figure displays the proportion of the total variance that was attributed to each of these 4 domains.