“…We used patient demographics (sex (male, female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanics, other), Charlson–Deyo Score (0–2+), facility type (academic, non-academic, other), facility location (New England, Middle Atlantic, South Atlantic, East North Central, East South Central, West North Central, West South Central, Mountain, Pacific), socioeconomic status (rurality (metropolitan, non-metropolitan), education level (rates of patients without high school level ≥21%, 13–20.9%, 7–12.9%, <7%), median income quartiles (<USD 38,000, USD 38,000–USD 47,999, USD 48,000–USD 62,999, >USD 63,000), insurance status (uninsured, private insurance, government insurance), travel distance to treatment facility (<12.5, 12.5–49.9, ≥50 miles)), tumor characteristics (primary site (head, body/tail, other), NCDB analytic stage group (I-IV), pathological grade (I-IV) [ 2 , 28 ], tumor size (<2, 2–4, >4 cm), nodal status (negative, positive), and treatments (radiotherapy (yes, no), and chemotherapy (yes, no)) [ 27 , 30 ]. Travel distance was defined using great circle distance, which calculates the distance between the patient’s residence and the reporting facility, using the geographic centroid of zip codes [ 33 ].…”