County-Level Social Vulnerability and Breast, Cervical, and Colorectal Cancer Screening Rates in the US, 2018

Key Points Question What is the association between social vulnerability and breast, cervical, or colorectal cancer screening rates at the county level in the US? Findings In this cross-sectional study of 3141 counties in the US, the US Preventive Services Task Force guideline-concordant screening rates of each cancer showed regional disparities. US counties with higher social vulnerability—measured as a composite index score—had significantly lower odds of receiving the recommended cancer screenings. Meaning These findings suggest that geographically targeted public health interventions could be further informed and improved by a composite measure reflecting the multidimensional nature of area-level social determinants of health.

eFigure 1. Maps of the US county-level Rural-Urban Continuum Codes (RUCC), percentage of uninsured population and access to primary physicians eMethods. Model equations and details eTable. Association of SVI and three cancer screening rates using 2018 PLACES data eResults. Additional analysis for PLACES cancer screening rates and SVI eFigure 2. Association between individual SDoH and cervical cancer screening rates (%) eFigure 3. Association between individual SDoH and breast cancer screening rates (%) eFigure 4. Association between individual SDoH and colorectal cancer screening rates (%) eFigure 5. Association between SDoH and cervical cancer screening rates (%) with all SDoH in the model eFigure 6. Association between SDoH and breast cancer screening rates (%) with all SDoH in the model eFigure 7. Association between SDoH and colorectal cancer screening rates (%) with all SDoH in the model eFigure 8. Maps identifying US counties that are currently not meeting 2023 Healthy People target for the three cancer screening rates eFigure 9. US County-level maps of three cancer screening rates using national average as cutoff points This supplemental material has been provided by the authors to give readers additional information about their work.

eMethods. Model equations and details.
We used a Bayesian mixed-effect Beta model to evaluate the association between county-level cancer screening rates by type and SVI categories (i.e., Q1: 0-<0.2, Q2: 0.2-<0.4, Q3: 0.4-<0.6, Q4: 0.6-<0.8, Q5: 0.8-1). Because all three screening rates were reported in percentages bounded between 0 and 1, we chose a Beta distribution when modeling the rate for the th county. The model (Model 1) can be presented as: We included a fixed effect to account for the unmeasured state-level factors (e.g., policy difference), and a fixed effect of the SVI categories denoted by SVI . Note that we used the bold font for these coefficient parameters to highlight that they were vectors -for example, there were 4 parameters in , with one each from SVI Q2, Q3, Q4 and Q5, where SVI Q1 was used as the reference category. We included a county-level random effect to account for the effect from any additional unmeasured county-level factor on the cancer screening rates, and assumed a normal distribution for it. The residuals were assumed to have a beta distribution with parameters and . Our primary interest was the coefficient , which was presented as the odds ratio relative to the reference group Q1. We used the 95% posterior credible intervals (95% CrI) to assess the statistical significance of SVI .
We considered the following model as Model 2: where we further included the urban/rural indicator variable as the adjustment variables.
In Model 3, we additionally adjusted for the percentage of uninsured (Uninsured) and access to primary care (PrimaryCare), both were scaled to have mean 0 and SD 1.
We presented the results from all three models. ©

eResults. Additional Analysis For PLACES Cancer Screening Rates and Social Vulnerability Index
Many contextual social determinants of health (SDoH) have been demonstrated as risk factors to many health related outcomes and behaviors. Individuals SDoHs, such as % of population living under poverty and % of population unemployed, have been extensively studied and established as having adverse effect to population health. This analysis used county-level 15 SDoH variables from 3,142 counties, and quantified their associations with three cancer screening rates (cervical cancer screening, breast cancer screening, and colorectal cancer screening). The cancer screening rates were extracted from 2018 PLACES project, which were model-based estimates using BRFSS database. The 15 SDoH variables selected as they were used in the construction of social vulnerabilities index (SVI) by CDC. The purpose of this analysis to investigate the association between individual SDoH and the three cancer screening rates at the US county level.
For each cancer screening outcome, we divided the particular SDoH under study to quintiles Q1 -Q5, with Q1 as the lowest quintile and the reference group in the analysis. We used a linear regression model, where we adjusted the eligible population size given the relevancy of the type of cancer screening, and included the State as a fixed effect.
Supplementary Figures 7-9 present the estimated association between each of the cancer screening rates and individual SDoH. For each outcome, most of the individual SDoH presented statistically significant association with the cancer screening rates. Specifically, % of population living with crowded housing, disabled, living in poverty, as minority, with no high school diploma, with no vehicles, unemployed, and single parent house, and per capita income presented consistent and strong association with all three cancer screening outcomes.
Supplementary Figures 10-12 present the estimated association between each of the cancer screening rates, and the SDoH, when all SDoH were included simultaneously in the model. It was clear that effects of the SDoH now were highly attenuated, and some no longer showed any statistical significance. More importantly, some SDoH showed an opposite association direction compared to those in Supplementary Figures 5-7. For example, the % of minority showed a positive and statistically significant association with cervical cancer screening rate in Supplementary Figure 5, where counties with higher % of minority had decreased screening rates; however, this association was flipped in Supplementary Figure 8, where counties with higher % of minority had increased screening rates.
The reason for such opposite direction in estimated effects of the SDoH is speculated due to the high correlation between these contextual SDoH. Therefore, using a composite score, such as the SVI constructed using these SDoHs, may be better in capturing the difference of the county-level cancer screening rates due to the SDoHs.