Table 1.

Definitions of predictor variables in CART and random forest analyses

Variable	Year(s)	Source	Conceptual Reason for Inclusion
% Patients younger than 65	2009-2018	SEER	Women aged 65 years and older may have better access to health insurance as they are mostly eligible for Medicare.
Incidence rate of breast cancer (age-adjusted)			Incidence rate is a measure of the breast cancer burden in the population.
Mammography use (age 40+)	2010	BRFSS	Mammography screening helps detect breast cancer at an early stage (1).
Mammography facilities per 100,000 population	2016	FDA	The availability of mammography could impact the uptake of screening in an area (5-9).
Hospitals per 100,000 population	2015	AHRF	Hospitals could recommend and provide screening services for women.
Community Health Centers per 100,000 population	2014		Community Health Centers could provide screening services or facilitate referrals for screening.
Primary care physicians per 100,000 population	2014		Sufficient primary care physicians are essential for preventive cancer care, and referrals for diagnostic services when necessary (10-12).
Obstetricians-gynecologists per 100,000 population	2015		Obstetrician and gynecologists are more likely to discuss and perform breast cancer screening and more likely to recognize breast cancer than other physicians (22, 23).
Radiologists per 100,000 population	2015		Radiologists are essential for the diagnosis and staging of breast cancer.
Population estimate	2014		An indicator of urbanicity, which may be associated with screening uptake (24).
% Urban Population	2010		An indicator of urbanicity, which may be associated with screening uptake (24).
Per Capita Income	2014		Lower income is associated with late-stage diagnosis of breast cancer (10, 25).
Rural-Urban Continuum Code	2013		An indicator of urbanicity, which may be associated with screening uptake (24)
Urban Influence Code	2013		An indicator of urbanicity, which may be associated with screening uptake (24)
Health Professional Shortage Area - Primary Care	2015		Sufficient availability of primary care physicians is essential for preventive cancer care, and referrals for diagnostic services when necessary (10-12).
% Eligible for Medicare	2014		Medicare-eligible individuals are more likely to undergo all cancer preventive services (26).
Median Household Income	2014		Lower income has been associated with late-stage diagnosis of breast cancer (10, 25).
% People in Poverty	2014		Poverty has been associated with late-stage diagnosis of breast cancer (27).
% Food Stamp or SNAP Recipients	2014		Receipt of SNAP benefits may be predictive of breast cancer tumor size (28).
% Uninsured women (age 18-64)	2014		Women without health insurance are more likely to be diagnosed at late stage for breast cancer compared to those with insurance (29).
% People under 200% of Poverty (age 18-64)	2014		Lower income and poverty are associated with late stage at breast cancer diagnosis (10, 25, 27).
% People with poor or fair health	2014	CHR	Overall poor or fair health may be positively or negatively associated with mammography screening based on either increased healthcare contacts or competing health priorities (30).
Poor physical health days	2014		More poor physical health days may be positively or negatively associated with mammography screening based on either increased healthcare contacts or competing health priorities (30).
Poor mental health days	2014		People with more poor mental health days may have lower priority of screening in the context of managing other medical and life issues (30).
Adult smoking	2014		Smoking may be inversely associated with use of colorectal cancer-screening tests. (31).
Excessive drinking	2014		Alcohol consumption may be associated with breast cancer screening rates (32).
Teen births	2007-2013		Teenage women who bear a child are more likely to have lower socioeconomic status and psychological distress in their later lives (33, 34).
Children in poverty	2014		Child poverty could reflect long term negative consequences of the population along various aspects of social determinants health (35).
Low birthweight	2008-2014		Low birthweight may indicate maternal exposure to various health risks (36).
Adult obesity	2013		High body mass is associated with late-stage breast cancer at diagnosis (37, 38).
Food environment index	2010&2014		The urgency of food insecurity may deprioritize the receipt of preventive screening services (39).
Physical Inactivity	2013		Physical activity may be associated with use of colorectal cancer screening tests (31).
Access to exercise opportunities	2010&2014		A study found that physical activity was also associated with use of colorectal cancer-screening tests (31).
Mammography use (Medicare age 67-69)	2014		Mammography screening helps detect breast cancer at an early stage (1).
Social associations	2014		People with adequate social support had more healthcare access and fostered more productive relationships with their healthcare providers (40).
Violent crime	2012-2014		Homicide rate in the neighborhoods of women’s nearest screening facility is associated with breast cancer late-stage diagnosis (41).
Severe housing problems	2009-2013		People with severe housing problems might have lower priority of screening in the context of managing other acute issues (42).
% Racial/Ethnic Minorities	2012-2016	ACS	Race and ethnicity have been strong predictors of late-stage breast cancer (25, 43).
% Family with own children (age < 18)			Women with more children are less likely to receive follow-up of tests or seek care for symptoms suggestive of breast cancer (44).
% Female-headed households			Women from neighborhoods with greater percentages of female-headed households may be at higher risk of LSBC (45).
% Women with high school degree or higher (age 25+)			Women from neighborhoods with less educated people may be at higher risk of LSBC (45, 46).
% People spoke English less than "very well" (age 18+)			Language may be a barrier to breast cancer screening (46-48).
% Women in management/business/science/arts occupations (age 16+)			Occupation categories are associated with breast cancer stage at diagnosis (49, 50).
% Women in service occupations (age 16+)
% Women in sales and office occupations (age 16+)
% Women in labor intensive occupations (age 16+)
% Renter occupied households			Women living in area with higher rates of renter occupied households are more likely to be diagnosed with LSBC (48).
% People moved residency in the past year			High frequency residential change is potentially a marker for the clinical risk of behavioral and emotional problems (51).
% Women worker drove alone to work (age 16+)			Percentage of women driving alone to work is an indicator of vehicle availability, which is an indicator of spatial access to screening services (3).
% Women worker with >= 30 min travel time to work (age 16+)			Travel time to work may be an indicator of proximity to urban centers where most screening services are located, which in turn may be associated with cancer stage at diagnosis (6-9).
% Women (age 15-50) had a birth in the past 12 months			After a childbirth, mothers experience a transient increased risk of late-stage breast cancer (52).
% Women unemployed among those in labor force (age 16+)			Women living in area with higher rates of unemployment were more likely to be diagnosed with LSBC (48).
Area Deprivation Index	2014	R Package ‘Sociome’	Neighborhood deprivation along various aspects of social determinants of health may be associated with LSBC (4, 5).

SEER = Surveillance, Epidemiology, and End Results; BRFSS = Behavioral Risk Factor Surveillance System; FDA = Food and Drug Administration; AHRF = Area Health Resources Files; CHR: County Health Rankings & Roadmaps; ACS = American Community Survey