Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Oct 23.
Published in final edited form as: JCO Clin Cancer Inform. 2026 Feb 5;10:e2500126. doi: 10.1200/CCI-25-00126

Application of the Population Health Institute Model of Health for Identifying Cancer Catchment Area Priorities

Amy Trentham-Dietz 1,2, Thomas P Lawler 1, Ronald E Gangnon 1,2,3, Allison R Dahlke 1, Noelle K LoConte 1,4, Earlise C Ward 1,5, Christine P Muganda 6, Shaneda Warren Andersen 1,2, Marjory L Givens 2,6
PMCID: PMC13688564  NIHMSID: NIHMS2213039  PMID: 41643139

Abstract

Purpose

The University of Wisconsin Population Health Institute (PHI) Model of Health, grounded in models developed over a decade ago, provides a framework for prioritizing health-related investments including setting agendas, implementing policies, and sharing resources for improving community health and health equity. The model includes multiple determinants of health and two broad health outcomes (length and quality of life). We adapted the PHI Model of Health to cancer outcomes.

Methods

Using county-level publicly available data, health factor summary measures were derived in three areas: health infrastructure including health promotion and clinical care, physical environment, and social and economic factors. A composite health factor z-score was calculated as the weighted (40%, 15%, and 45%, respectively) average of the summary measures for each county, and k-means clustering was used to create unequally sized county groups with lower (“healthier”) to higher (“less healthy”) z-scores. We fit age-adjusted negative binomial regression models to estimate rate ratios and 95% confidence intervals (CI) for cancer mortality in relation to county health factor cluster.

Results

Age-adjusted cancer mortality rates increased across the 10 county health factor clusters for all-cancers as well as for lung, colorectal, breast, and prostate cancer. Rate ratios generally increased across the ten health factor clusters for all-cancers combined and for specific cancer types. Compared to counties with the most favorable health factor conditions, the counties with the least favorable conditions had an all-cancer mortality rate ratio of 1.49 (95% CI 1.39-1.60).

Conclusion

The PHI model of health adapted to cancer outcomes provides an approach for linking community-specific conditions to the interventions that hold promise to directly address drivers of the cancer burden.

Keywords: health frameworks, community-level interventions, publicly available data, cancer mortality

INTRODUCTION

National Cancer Institute (NCI)-designated cancer centers must identify a geographic area, referred to as a “catchment area,” that each center serves in its research, clinical care, education, and community outreach and engagement.1 This requirement is intended to focus the center’s mission towards reducing the cancer burden for the catchment area’s population.

Centers have employed different approaches to identify priorities. Often centers use a combination of sources of information in these decisions including cancer incidence and mortality rates, prevalence of risk factors, and cancer screening rates within the catchment area; knowledge about the areas of scientific and clinical expertise and strengths at their centers; and input from community members. Opportunities remain to specify a framework for decision making that takes a systems science approach including consideration of multiple possible interacting, heterogeneous, and dynamic streams of data for different populations and geographic regions within catchment areas.2 Effective cancer prevention and control efforts need to leverage data for intervening at multiple levels—patient, provider, health system, and societal—with multiple complementary approaches.

The University of Wisconsin Population Health Institute (PHI) Model of Health3 (an evolution formerly known as the County Health Rankings Model of Health4) offers promise as a rigorous approach for identifying catchment area priorities for all NCI-designated cancer centers. Supported by the Robert Wood Johnson Foundation, the PHI Model of Health has been used for over two decades to aid health professionals and decision makers in understanding community conditions favorable to living long and well and in identifying evidence-informed approaches for improving community health and health equity. The model draws on a theory of change and recognition of the cross-sector participation needed to catalyze community health improvement. The model draws attention to factors beyond clinical care and individual behavior change that shape community health.3 The PHI Model of Health has had national impact yet has generally been defined and implemented to address broad population health outcomes of quality and length of life.

In this report, we adapt the PHI Model of Health to cancer for counties across the United States and demonstrate the strengths of this approach for identifying priorities in an NCI-designated cancer center catchment area. Specifically, the model is used to identify counties with risk factors that may be driving higher cancer death rates which may then be prioritized as targets for intervention.

METHODS

To adapt the PHI Model of Health to cancer, we modified the health outcomes (“population health and well-being”) and health factors (“community conditions”) while retaining the main model structure including county as the geographic unit. We then compared cancer outcomes with a composite index of the health factors, divided into clusters, to demonstrate variation across >3,000 counties in the United States.

Health Outcomes: County-level cancer mortality rates

The PHI Model of Health includes length and quality of life as health outcomes. For this analysis, we instead used county-level cancer mortality rates from the Centers for Disease Control and Prevention (CDC) State Cancer Profiles website.5 Cancers of interest included all cancers combined and lung, colorectal, female breast, and prostate cancer separately. Data for counties with an average of three or fewer cases per year for a given outcome were suppressed, and consequently these counties were excluded from the analysis. Data from 3,079 out of 3,143 (98.0%) U.S. counties were available for all-cancer mortality.

Health Factors: County-level community conditions

Health factors in the PHI Model of Health were reviewed for cancer relevance; factors were either retained, dropped, or added to the adapted model based on consensus in the study team and data availability for counties nationwide. County-level health factors were included that reflect intervenable targets for change in three categories: health infrastructure, physical environment, and social and economic factors (Table 1). Priority was placed on factors that reflect community conditions affected by policies and laws including healthcare delivery decisions rather than behaviors of individual community members. Weights were based on expert opinion and first assigned to the three categories using a similar top-down approach as for the PHI Model of Health.3 Health factors were generally perceived as having relatively modest (1-2%), medium (4-5%), or higher (8%) contributions to cancer risk within the health infrastructure (weighted 40%), physical environment (weighted 15%), and social and economic factor categories (weighted 45%). Weights and selected measures were similar but not exactly the same as those in the PHI Model of Health as a result of adapting the model for cancer as an outcome.6,7

Table 1.

Cancer-related health factors included in the 2025 University of Wisconsin Population Health Institute Model of Health adapted for cancer

Health Factors Weight Data Source
Health infrastructure 40%
 Access to exercise opportunities (%) 4% ArcGIS Business Analyst and ArcGIS Online; YMCA; and US Census TIGER/Line Shapefiles from CHR&R11
 Food Environment Index 4% USDA Food Environment Atlas and Map the Meal Gap from Feeding America from CHR&R11
 Adult smoking (%) 8% BRFSS from CHR&R11
 Adult obesity (%) 4% BRFSS from CHR&R11
 Excessive drinking (%) 2% BRFSS from CHR&R11
 Uninsured (%) 4% Small area health insurance estimates from CHR&R11
 Primary care physicians (adjusted for county population) 4% Area Health Resource File/American Medical Association from CHR&R11
 Oncologists (adjusted for county population) 5% Centers for Medicare & Medicaid Services10
 Mammography screening (%) 1% Mapping Medicare Disparities Tool from CHR&R11
 Colonoscopy screening (%) 1% CDC State Cancer Profiles5
 Lung cancer screening (%) 1% Mapping Medicare Disparities Tool24
 Prostate cancer screening (%) 1% CDC State Cancer Profiles5
 Cervical cancer screening (%) 1% CDC State Cancer Profiles5
Physical environment 15%
 Air pollution (PM2.5) 4% Environmental Public Health Tracking Network from CHR&R11
 Broadband internet access (%) 4% American Community Survey, 5-year estimates from CHR&R11
 Drinking water violations (no, yes) 1% Safe Drinking Water Information System data from CHR&R11
 Radon (zone 1-4) 1% EPA9
 Severe housing problems (%) 4% Comprehensive Housing Affordability Strategy (CHAS) data from CHR&R11
 Lead (DPI) 1% US Department of Housing and Urban Development8
Social and economic factors 45%
 High school completion (%) 8% American Community Survey, 5-year estimates from CHR&R11
 Some college (%) 8% American Community Survey, 5-year estimates from CHR&R11
 Unemployment (%) 8% Bureau of Labor Statistics from CHR&R11
 Children in poverty (%) 8% Small Area Income and Poverty Estimates; and American Community Survey, 5-year estimates from CHR&R11
 Income inequality 8% American Community Survey, 5-year estimates from CHR&R11
 Social associations (per 100,000) 5% County Business Patterns from CHR&R11
Open in a new tab

Abbreviations: BRFSS – Behavioral Risk Factor Surveillance System; CDC – Centers for Disease Control; CHR&R – County Health Rankings & Roadmaps; EPA – Environmental Protection Agency; GIS – geographic information system; PM – particulate matter; TIGER – Topological Integrated Geographic Encoding and Referencing Line; US – United States; YMCA – Young Men’s Christian Association.

Health infrastructure factors included measures related to health promotion and harm reductions (access to exercise opportunities, the Food Environment Index, smoking, obesity, excessive drinking) and clinical care (percent uninsured under age 65, number of primary care physicians (PCPs) and oncologists adjusted for population size, and screening rates for breast, colorectal, lung, prostate, and cervical cancer). Physical environment factors included measures related to air, water, and land (particulate matter (PM) 2.5, radon levels, drinking water violations, and broadband internet access) and housing (severe housing problems, lead exposure). Finally, social and economic factors included measures for education (high school graduation, completion of some college), income, employment, and wealth (unemployment, children in poverty, income inequality) and social support (social associations).

Screening prevalence for colorectal, prostate, and cervical cancers were obtained from the CDC State Cancer Profiles website.5 Screening for colorectal cancer was defined as having completed a colonoscopy in the past 10 years for adults age 50-75. Lead exposure was determined via the Deteriorated Paint Index, which reflects the number of pre-1980 households with large areas of deteriorated paint (a predictor of exposure to lead dust), obtained from the US Department of Housing and Urban Development website.8 Radon exposure zones were obtained from the Environmental Protection Agency website.9 The number of oncologists for each county was obtained from the Centers for Medicare & Medicaid Services National Downloadable File.10 All other health factors and their data sources were obtained from the PHI website.11

Statistical analysis

County-level z-scores were created for the health factors.12 For beneficial exposures (e.g. cancer screening), values were inverted so that higher values were expected to increase cancer risk. Standardized values that were less than −3 or greater than 3 were truncated to −3 or 3, respectively.4 After standardization, weights were applied to the factors as listed in Table 1. The weighted values for the factors were summed to create a composite z-score for each county, with higher values indicating counties with less favorable health conditions.

K-means clustering of the composite z-scores was performed to group counties with similar health factor values, as described by Pollock et al.13 Clusters were identified numerically with higher clusters indicating less favorable county-level health factors. The total number of clusters was limited to 10 for ease of communication. As previously described, clustering counties with similar health conditions helps to mitigate error when ranking counties based on composite z-scores that have unknown and varying levels of uncertainty.13

Associations between the composite z-score clusters (1-10) and cancer mortality rates at the county level were estimated using negative binomial regression models, with adjustment for county median age and including county population size as an offset. Associations were reported as mortality rate ratios with 95% confidence intervals (CIs) with counties in cluster 1 (most favorable health conditions) serving as the reference category. Further, the age-standardized mortality rates per 100,000 population were calculated for each cluster.

The percentage of missing values for county-level health factors ranged from 0-5.0%. Missing values were imputed using the state-specific median value prior to data standardization and weighting. All analyses were completed using R version 4.3.1, including the ‘MASS’ package for running negative binomial models and the ‘Ckmeans.1d.dp’ package for K-means clustering.14-16

To examine how health conditions varied across health factor clusters, four counties were randomly selected across regions in the US in clusters 1, 4, 7, and 10 using a random number generator. For ease of demonstration, only health factors measured on the same scale (percentages) were plotted.

RESULTS

Across the 3,079 US counties with data, age-adjusted all-cancer mortality rates ranged from 48.9 (95% CI 35.3, 66.6; Mono County, CA) to 387.8 (95% CI 348.9, 430.1; Union County, FL) per 100,000 for 2018-2022. All-cancer age-adjusted mortality rates ranged from 126.2 per 100,000 for health factor cluster 1 (z-scores −1.53 to −0.72) to 165.5 per 100,000 for cluster 10 (RR 1.49, 95% CI 1.39-1.60; z-scores 1.28 to 1.88; Table 2; Appendix Table A1).

Table 2.

Cancer mortality rate ratios according to counties clustered by health factors included in the adapted 2025 University of Wisconsin Population Health Institute Model of Health, US (N = 3,079 counties)

Health Factor
cluster		 Mortality Rate Ratios (95% Confidence Intervals)a
Z-score range N, Countiesb All Cancer Sites Lung Cancer Colorectal Cancer
1 (most healthy) −1.53 to −0.72 194 1 (ref) 1 (ref) 1 (ref)
2 −0.72 to −0.46 442 1.16 (1.13-1.19) 1.26 (1.20-1.32) 1.13 (1.09-1.18)
3 −0.46 to −0.24 448 1.23 (1.20-1.26) 1.42 (1.35-1.49) 1.22 (1.18-1.27)
4 −0.24 to −0.03 540 1.31 (1.27-1.34) 1.60 (1.53-1.67) 1.33 (1.28-1.38)
5 −0.03 to 0.18 419 1.37 (1.33-1.40) 1.71 (1.63-1.79) 1.43 (1.37-1.49)
6 0.18 to 0.40 350 1.44 (1.40-1.48) 1.90 (1.80-2.00) 1.53 (1.46-1.60)
7 0.40 to 0.64 318 1.47 (1.42-1.51) 1.96 (1.86-2.07) 1.67 (1.58-1.76)
8 0.64 to 0.91 209 1.51 (1.46-1.56) 1.99 (1.88-2.12) 1.75 (1.63-1.87)
9 0.91 to 1.25 116 1.56 (1.49-1.62) 2.10 (1.94-2.27) 1.69 (1.54-1.86)
10 (least healthy) 1.28 to 1.88 43 1.49 (1.39-1.60) 1.75 (1.51-2.02) 2.10 (1.72-2.53)
Open in a new tab
a

Data presented as age-adjusted mortality rate ratios (with 95% confidence interval. All cancer mortality rates are for the years 2018-2022.

b

Represents the total number of counties in each cluster for all cancer sites combined. The number of counties for each outcome is: all cancer – 3,079; lung cancer – 2,745; colorectal cancer – 2,180.

Abbreviation: ref, reference category

Rate ratios for all-cancer mortality generally increased across health factor clusters (Fig 1, Table 2, Appendix Table A2). Comparing cluster 10 to 1, the rate ratio for lung cancer (1.75, 95% CI 1.51-2.02) and colorectal cancer (2.10, 95% CI 1.72-2.53) was larger than for breast cancer (1.22, 95% CI 0.90-1.60 among females), and prostate cancer (1.53, 95% CI 1.06-2.13 among males).

FIG 1.

FIG 1.

Open in a new tab

Age-adjusted rate ratios (and 95% confidence intervals) for all-cancer mortality according to health factor county cluster for female, male, Black, and White populations. Clusters range from 1 (most healthy community conditions, reference category) to 10 (least healthy).

Rate ratio estimates for all-cancer, lung, and colorectal cancers were higher for males than for females (Fig 1, Appendix Table A2). For example, comparing cluster 10 to 1, the rate ratio for lung cancer was 1.63 (95% CI 1.26-2.08) for females and 1.97 (95% CI 1.60-2.41) for males.

Confidence intervals were wider and rate ratio estimates were generally greater for the Black than for the White population (Fig 1, Appendix Table A3). For example, comparing cluster 10 to 1, the rate ratio for colorectal cancer was 2.72 (95% CI 1.54-4.42) for Black persons and 2.16 (95% CI 1.44-3.09) for White persons.

To demonstrate how health conditions varied across health factor clusters, four counties were randomly selected across regions in the US in clusters 1, 4, 7, and 10. As could be expected, colon cancer screening rates decreased and the percentage of children in poverty increased from cluster 1 to 10 (Fig 2). Other health factors did not increase or decrease in an orderly manner across clusters. For example, cervical cancer screening was lower in a cluster 1 county (66%) than for counties in cluster 4 and 7 (76% and 81%, respectively); and attending some college was lower in a cluster 7 county (36%) than for a cluster 10 county (51%).

FIG 2.

FIG 2.

Open in a new tab

Histogram of selected health factors for four United States counties. Counties randomly selected within clusters 1 (Ada County, Idaho), 4 (Kent County, Delaware), 7 (Anson County North Carolina), and 10 (McKinley County, New Mexico). Clusters represent community conditions that are most (cluster 1) to least (cluster 10) supportive of health. Counties have a range of population percent in rural areas11 (Ada 5.6%, Anson 77.8%, Kent 26.5%, McKinley 66.5%) and age-adjusted all-cancer mortality rates per 100,0005 (Ada 131.8, 95% CI 127.5-136.2; Anson 191.2, 95% CI 169.4-215.3; Kent 168.1, 95% CI 160.6-175.9; McKinley 143.0, 95% CI 130.4-156.5). CI, confidence interval.

In a sensitivity analysis to examine the impact of choices of factors and weights assigned to the various health factors, different weights were assigned to slightly different health factors arranged in four categories according to the 2014 County Health Rankings Model of Health, including health behaviors (30%), clinical care (20%), social and economic factors (40%), and physical environmental factors (10%; Appendix Table A4). Clustering resulted in fewer counties in clusters 1, 2, and 4 (Appendix Table A5). Comparing cluster 10 to 1, rate ratios for all-cancers (1.60, 95% CI 1.52-1.69), lung cancer (2.06, 95% CI 1.86-2.28), and colorectal cancer (1.79, 95% CI 1.58-2.01) were similar to estimates in the primary analysis (1.49, 1.75, and 2.10, respectively, Table 2).

INTERPRETATION

By applying the PHI Model of Health to cancer mortality, we showed that the model distinguishes between groups of counties based on community conditions that are conducive to larger or smaller cancer mortality burdens. Mortality rates across clusters of counties were as much as two-fold different, demonstrating that the adapted PHI Model of Health was flexible to characterize variation in cancer mortality rates across counties for several different demographic groups and cancer types. The purpose of the model was not to control for confounding factors or model all possible causal factors related to cancer mortality; instead, health factors in the model were restricted to intervenable targets for change. A strength of this approach is its acknowledgement that the factors supportive of health are inextricably interrelated consistent with a systems science framework;2 the PHI Model of Health offers an opportunity to identify intervenable priorities and provides an entry for the improvement of health factors within the local health infrastructure, physical environment, and social and economic ecosystems. Model results are intended to inform conversations and action about priorities for cancer prevention and control efforts and should be complemented with knowledge of other important community conditions, such as rurality, and input of local residents.

This approach complements other techniques used by NCI-designated cancer centers to identify priorities for innovations in research, clinical care, outreach, and education. For many years, institutions have followed the example set by the American Cancer Society and generated lay-friendly reports that resemble Cancer Facts & Figures.17 More recently, dozens of centers are using Cancer InFocus, developed at the University of Kentucky Markey Cancer Center,18 that facilitates exploration of variations in cancer incidence and mortality rates, demographic factors, and several other factors through maps. The Cancer InFocus website provides an interactive platform that complements static reports. Our approach uses much of the same data available on the Cancer InFocus website, with our methods also providing a way to integrate the multiple types of health factors into a single z-score. Numerous indices of health have been developed for similar purposes, including the Area Deprivation Index19 and the Yost Index.20 However, these indices can include factors, such as home values, that lack evidence-based interventions for reducing the burden of diseases such as cancer.

The PHI Model of Health is proposed as a framework for identifying cancer priorities because, as mentioned immediately above, it is composed of intervenable targets for health improvement. Several resources are available to identify evidence-based interventions related to the factors in the model.21-23 For example, shared decision making between healthcare providers and patients for cancer screening and the provision of patient navigators when cancer patients are first diagnosed are scientifically supported interventions.23 Health factors in the PHI model are amenable to intervention at both the individual and population levels but call attention to the community conditions (social determinants of health) that require structural change at policy and systems levels, for example, tobacco taxes and expanded access to affordable health insurance. The PHI model is also flexible to identify geographic areas with the highest burden of all cancers combined or for specific cancer sites, and for specific demographic groups. The model can be applied to geographic units other than county for which health factor and outcome data are available.

Several limitations of our approach indicate opportunities for improvement. Health factors weights are nominal and not intended to directly reflect population attributable fractions. Our sensitivity analysis showed that cancer mortality rate ratios were elevated across the health factor clusters under a different set of weights. Using two sets of weights corresponding to the 2014 County Health Rankings and 2025 PHI Models of Health did not result in substantial differences in risk estimates or cluster definitions. However, national data were not available for some key cancer risk factors that ideally could be included such as HPV vaccination. Not all counties had sufficient data, and cancer mortality rates were suppressed for less populous counties and less common cancer types. County was used in this analysis and may represent the appropriate unit for intervention, for example, in a partnership with a county health department or other organizations that are based at the county level. County may not be the ideal geographic area for other partnerships or interventions and may also disguise heterogeneity of cancer risks within its borders. However, this approach provides a framework for straightforward application to geographic units other than county, e.g., census tract, health service area, or province, for which relevant data are available.

The Population Health Institute has accumulated a track record of scholarship and coaching communities to implement positive changes for health and health equity since 2010. The framework described here provides an approach for identifying and linking community-specific conditions to the evidence-based interventions that hold promise to directly address drivers of the local cancer burden. This framework, together with input from community members and the scientific and clinical strengths of a cancer center, provides a data-driven model to prioritize efforts for reducing the cancer burden within cancer center catchment areas.

CONTEXT SUMMARY.

Key Objective:

Can the Population Health Institute Model of Health provide a data framework for informing decision-making to address health-related disparities within cancer center catchment areas through systems, policy, and public health interventions?

Knowledge Generated:

Using the Model framework to cluster counties with similar community conditions describing intervenable health factors, most health factors sorted in a dose-response manner across the clusters, e.g., access to screening tests, exercise opportunities, broadband internet, insurance, and education. The county cluster with the least favorable community conditions had a 49% greater all-cancer mortality rate ratio than the county cluster with the most favorable conditions.

Grant funding:

This work was supported by grant P30CA014250 by the National Institutes of Health/National Cancer Institute.

APPENDIX

Appendix Table A1:

Cancer mortality rates according to counties clustered by health factors included in the adapted 2025 University of Wisconsin Population Health Institute Model of Health, US (N = 3,079 counties)a

Health Factor
Clusterb 		Z-score Range N, Countiesb All cancer Lung Colorectal Breast
(female)		 Prostate
(male)
1 (most healthy) −1.53 to −0.72 194 126.2 23.7 10.5 17.2 17.3
2 −0.72 to −0.46 442 136.8 27.8 11.5 18.2 18.3
3 −0.46 to −0.24 448 143.1 30.4 12.1 18.9 18.2
4 −0.24 to −0.03 540 150.5 34.0 13.0 19.6 18.2
5 −0.03 to 0.18 419 148.7 29.9 13.7 20.0 19.6
6 0.18 to 0.40 350 154.0 33.2 14.3 20.8 19.4
7 0.40 to 0.64 318 165.3 36.8 15.6 21.2 20.5
8 0.64 to 0.91 209 170.1 37.0 16.4 22.4 22.0
9 0.91 to 1.25 116 144.4 26.5 13.6 19.4 18.5
10 (least healthy) 1.28 to 1.88 43 165.5 24.2 19.2 19.1 21.1
Open in a new tab
a

Data presented as the mean of the county-level age-standardized mortality rates for the years 2018-2022 weighted by the inverse of the squared standard error.

b

Represents the total number of counties in each cluster for all cancer sites combined. The number of counties for each outcome is: all cancer – 3,079; lung cancer – 2,745; colorectal cancer – 2,180; breast cancer – 1,778; prostate cancer – 1,612.

Appendix Table A2:

Cancer mortality rate ratios according to counties clustered by health factors included in the adapted 2025 University of Wisconsin Population Health Institute Model of Health by sex, US

Health Factor
Cluster		 Z-score range N,
Countiesb 		Mortality Rate Ratios (95% Confidence Intervals)a
Female
population		 All Cancer Sites Lung Cancer Colorectal Cancer Breast Cancer
1 (most healthy) −1.53 to −0.72 190 1 (ref) 1 (ref) 1 (ref) 1 (ref)
2 −0.72 to −0.46 416 1.13 (1.10-1.16) 1.21 (1.14-1.27) 1.10 (1.05-1.16) 1.08 (1.04-1.12)
3 −0.46 to −0.24 428 1.18 (1.15-1.21) 1.33 (1.27-1.41) 1.17 (1.11-1.22) 1.13 (1.09-1.17)
4 −0.24 to −0.03 514 1.24 (1.21-1.27) 1.46 (1.38-1.53) 1.25 (1.19-1.32) 1.18 (1.13-1.22)
5 −0.03 to 0.18 408 1.28 (1.25-1.32) 1.51 (1.43-1.60) 1.31 (1.24-1.38) 1.20 (1.15-1.25)
6 0.18 to 0.40 345 1.32 (1.29-1.36) 1.63 (1.53-1.72) 1.34 (1.25-1.42) 1.23 (1.17-1.29)
7 0.40 to 0.64 307 1.34 (1.30-1.38) 1.64 (1.53-1.75) 1.49 (1.36-1.62) 1.31 (1.22-1.39)
8 0.64 to 0.91 203 1.37 (1.32-1.42) 1.64 (1.52-1.78) 1.60 (1.42-1.79) 1.31 (1.21-1.43)
9 0.91 to 1.25 112 1.37 (1.31-1.44) 1.66 (1.49-1.85) 1.31 (1.14-1.49) 1.22 (1.10-1.35)
10 (least healthy) 1.28 to 1.88 38 1.30 (1.19-1.42) 1.63 (1.26-2.08) 1.78 (1.23-2.48) 1.22 (0.90-1.60)
Male population All Cancer Sites Lung Cancer Colorectal Cancer Prostate Cancer
1 (most healthy) −1.53 to −0.72 191 1 (ref) 1 (ref) 1 (ref) 1 (ref)
2 −0.72 to −0.46 426 1.18 (1.14-1.21) 1.29 (1.22-1.36) 1.15 (1.09-1.21) 1.13 (1.07-1.18)
3 −0.46 to −0.24 439 1.26 (1.23-1.30) 1.48 (1.40-1.57) 1.27 (1.21-1.33) 1.18 (1.13-1.24)
4 −0.24 to −0.03 525 1.35 (1.32-1.39) 1.71 (1.63-1.81) 1.39 (1.32-1.46) 1.19 (1.13-1.25)
5 −0.03 to 0.18 411 1.42 (1.38-1.46) 1.85 (1.75-1.95) 1.49 (1.42-1.57) 1.26 (1.20-1.33)
6 0.18 to 0.40 348 1.51 (1.47-1.56) 2.11 (1.98-2.23) 1.60 (1.50-1.70) 1.25 (1.17-1.34)
7 0.40 to 0.64 310 1.54 (1.49-1.59) 2.21 (2.07-2.35) 1.83 (1.70-1.97) 1.36 (1.26-1.47)
8 0.64 to 0.91 203 1.57 (1.51-1.64) 2.26 (2.10-2.43) 1.89 (1.71-2.09) 1.45 (1.30-1.61)
9 0.91 to 1.25 113 1.61 (1.53-1.69) 2.34 (2.13-2.58) 1.72 (1.51-1.95) 1.39 (1.20-1.59)
10 (least healthy) 1.28 to 1.88 41 1.54 (1.41-1.68) 1.97 (1.60-2.41) 2.23 (1.62-2.97) 1.53 (1.06-2.13)
Open in a new tab
a

Data presented as age-adjusted mortality rate ratios (with 95% confidence interval. All cancer mortality rates are for the years 2018-2022.

b

Represents the total number of counties in each cluster. The number of counties for each outcome is: all cancer – 3,007 (males), 2,961 (females); lung cancer – 2,466 (males), 2,269 (females); colorectal cancer – 1,585 (males), 1,368 (females); breast cancer – 1,778 (females); prostate cancer – 1,612 (males).

Appendix Table A3:

Cancer mortality rate ratios according to counties clustered by health factors included in the adapted 2025 University of Wisconsin Population Health Institute Model of Health, Black and White population, US

Cluster Z-score range N, Countiesb All cancer Lung Colorectal Breast (female) Prostate (male)
Black population
1 (most healthy) −1.53 to −0.72 88 1 (ref) 1 (ref) 1 (ref) 1 (ref) 1 (ref)
2 −0.72 to −0.46 127 1.22 (1.14-1.30) 1.36 (1.19-1.55) 1.19 (1.05-1.36) 1.15 (1.01-1.30) 1.28 (1.09-1.50)
3 −0.46 to −0.24 148 1.28 (1.20-1.37) 1.45 (1.28-1.65) 1.18 (1.04-1.33) 1.12 (1.00-1.26) 1.30 (1.11-1.51)
4 −0.24 to −0.03 192 1.40 (1.31-1.49) 1.62 (1.43-1.83) 1.36 (1.20-1.53) 1.20 (1.07-1.35) 1.42 (1.22-1.66)
5 −0.03 to 0.18 159 1.51 (1.42-1.62) 1.75 (1.54-1.99) 1.47 (1.30-1.66) 1.29 (1.15-1.45) 1.53 (1.31-1.79)
6 0.18 to 0.40 143 1.60 (1.50-1.72) 1.99 (1.73-2.29) 1.46 (1.28-1.67) 1.25 (1.10-1.42) 1.45 (1.21-1.73)
7 0.40 to 0.64 159 1.65 (1.54-1.77) 2.14 (1.86-2.46) 1.76 (1.48-2.08) 1.32 (1.09-1.59) 1.63 (1.27-2.06)
8 0.64 to 0.91 113 1.74 (1.61-1.87) 2.32 (1.98-2.71) 2.03 (1.61-2.52) 1.33 (0.97-1.79) 1.89 (1.38-2.55)
9 0.91 to 1.25 68 1.83 (1.68-2.00) 2.47 (2.05-2.97) 1.44 (1.14-1.80) 1.29 (1.04-1.60) 1.50 (1.08-2.06)
10 (least healthy) 1.28 to 1.88 19 1.99 (1.72-2.30) 2.49 (1.76-3.46) 2.72 (1.54-4.42) 2.01 (1.07-3.42) 2.56 (1.14-4.98)
White population
1 (most healthy) −1.53 to −0.72 193 1 (ref) 1 (ref) 1 (ref) 1 (ref) 1 (ref)
2 −0.72 to −0.46 439 1.15 (1.13-1.18) 1.25 (1.20-1.30) 1.12 (1.08-1.17) 1.07 (1.02-1.11) 1.08 (1.03-1.14)
3 −0.46 to −0.24 444 1.23 (1.20-1.26) 1.44 (1.38-1.50) 1.22 (1.17-1.27) 1.09 (1.05-1.13) 1.11 (1.06-1.17)
4 −0.24 to −0.03 538 1.31 (1.28-1.34) 1.64 (1.57-1.70) 1.32 (1.27-1.37) 1.14 (1.10-1.19) 1.10 (1.04-1.16)
5 −0.03 to 0.18 418 1.38 (1.35-1.41) 1.76 (1.69-1.84) 1.43 (1.37-1.49) 1.20 (1.15-1.26) 1.20 (1.13-1.26)
6 0.18 to 0.40 349 1.45 (1.42-1.49) 1.97 (1.88-2.06) 1.50 (1.42-1.57) 1.22 (1.15-1.28) 1.13 (1.05-1.22)
7 0.40 to 0.64 309 1.49 (1.45-1.53) 2.07 (1.97-2.17) 1.65 (1.55-1.76) 1.31 (1.21-1.42) 1.32 (1.19-1.45)
8 0.64 to 0.91 204 1.55 (1.50-1.60) 2.20 (2.08-2.33) 1.79 (1.65-1.95) 1.34 (1.20-1.50) 1.45 (1.24-1.69)
9 0.91 to 1.25 107 1.55 (1.48-1.62) 2.24 (2.07-2.43) 1.84 (1.61-2.09) 1.32 (1.09-1.59) 1.33 (1.02-1.69)
10 (least healthy) 1.28 to 1.88 31 1.61 (1.46-1.77) 2.51 (2.07-3.01) 2.16 (1.44-3.09) 1.03 (0.26-2.68)c d
Open in a new tab
a

Data presented as age-adjusted mortality rate ratios (with 95% confidence intervals). All cancer mortality rates are for the years 2018-2022.

b

Represents the total number of counties in each cluster for all cancer sites combined. For the White population, the number of counties for each outcome is: all cancer – 1,216; lung cancer – 665; colorectal cancer – 386; breast cancer – 341; prostate cancer – 300. For the Black population, the number of counties for each outcome is: all cancer – 3,032; lung cancer – 2,654; colorectal cancer – 1,970; breast cancer – 1,599; prostate cancer – 1,425.

c

Only 1 county had at least 3 breast cancer deaths among White women in cluster 10.

d

Zero counties in cluster 10 have at least 3 prostate cancer deaths per year on average among White men.

Table A4.

Cancer-related health factors included in the 2014 County Health Rankings Model of Health adapted for cancer

Health Factors Weight Data source
Health behaviors 30%
 Adult smoking (%) 10% BRFSS from CHR&R11
 Adult obesity (%) 10% BRFSS from CHR&R11
 Physical inactivity (%) 3% BRFSS from CHR&R11
 Sun burn risk (UV-B, Wh/m2) 2% National Cancer Institute25
 Excessive drinking (%) 3% BRFSS from CHR&R11
 Hepatitis C mortality (per 100,000) 1% HepVu26
 HIV prevalence (per 100,000) 1% From CHR&R11
Clinical care 20%
 Uninsured (%) 5% Small area health insurance estimates from CHR&R11
 Primary care physicians (ratio) 2% Area Health Resource File/American Medical Association from CHR&R11
 Dentists (ratio) 1% Area Health Resource File/National Provider Identifier Downloadable File from CHR&R11
 Oncologists (per county) 1% Centers for Medicare & Medicaid Services10
 Broadband internet access (%) 1% American Community Survey, 5-year estimates from CHR&R website11
 Mammography screening (%) 2% Mapping Medicare Disparities Tool from CHR&R11
 Colonoscopy screening (%) 2% CDC State Cancer Profiles5
 Lung cancer screening (%) 2% Mapping Medicare Disparities Tool24
 Prostate cancer screening (%) 2% CDC State Cancer Profiles5
 Cervical cancer screening (%) 2% CDC State Cancer Profiles5
Physical environment 10%
 Air pollution (PM2.5) 2% Environmental Public Health Tracking Network from CHR&R11
 
 Radon (zone 1-4) 2% EPA9
 Lead (DPI) 1% US Department of Housing and Urban Development8
 Access to exercise opportunities (%) 2.5% ArcGIS Business Analyst and ArcGIS Online; YMCA; and US Census TIGER/Line Shapefiles from CHR&R11
 Food Environment Index 2.5% USDA Food Environment Atlas and Map the Meal Gap from Feeding America from CHR&R11
Social and economic factors 40%
 High school completion (%) 5% American Community Survey, 5-year estimates from CHR&R11
 Some college (%) 5% American Community Survey, 5-year estimates from CHR&R11
 Unemployment (%) 10% Bureau of Labor Statistics from CHR&R11
 Children in poverty (%) 7.5% Small Area Income and Poverty Estimates; and American Community Survey, 5-year estimates from CHR&R11
 Income inequality 2.5% American Community Survey, 5-year estimates from CHR&R11
 Social associations (per 100,000) 5% County Business Patterns from CHR&R11
 Black/White residential segregation (Index of Dissimilarity) 5% From CHR&R11
Open in a new tab

Abbreviations: CDC – Centers for Disease Control; CHR – County Health Rankings; CMS – Centers for Medicare & Medicaid Services; DPI – deteriorated paint index; EPA – Environmental Protection Agency; HIV – human immunodeficiency virus; PM – particulate matter

Appendix Table A5.

Cancer mortality rate ratios according to counties clustered by health factors included in the adapted 2014 County Health Rankings Model of Health, US (N = 3,079 counties)a

Health Factor
Clusterb 		Z-score range N,
Countiesc 		All cancer Lung Colorectal Breast (female) Prostate (male)
1 (most healthy) −1.40 to −0.82 109 1 (ref) 1 (ref) 1 (ref) 1 (ref) 1 (ref)
2 −0.81 to −0.54 278 1.14 (1.10-1.17) 1.22 (1.16-1.29) 1.10 (1.05-1.15) 1.08 (1.03-1.12) 1.10 (1.04-1.16)
3 −0.54 to −0.32 464 1.26 (1.22-1.29) 1.46 (1.38-1.54) 1.22 (1.17-1.27) 1.15 (1.10-1.20) 1.19 (1.13-1.26)
4 −0.32 to −0.12 460 1.33 (1.29-1.37) 1.64 (1.56-1.73) 1.31 (1.25-1.37) 1.15 (1.11-1.20) 1.20 (1.14-1.27)
5 −0.12 to 0.07 493 1.39 (1.35-1.43) 1.76 (1.67-1.85) 1.40 (1.35-1.47) 1.22 (1.17-1.27) 1.25 (1.18-1.32)
6 0.08 to 0.27 419 1.48 (1.44-1.52) 1.96 (1.85-2.06) 1.54 (1.47-1.61) 1.29 (1.23-1.35) 1.31 (1.23-1.39)
7 0.27 to 0.48 351 1.52 (1.48-1.57) 2.09 (1.98-2.21) 1.61 (1.53-1.70) 1.26 (1.19-1.33) 1.25 (1.16-1.35)
8 0.48 to 0.72 266 1.53 (1.48-1.58) 2.09 (1.97-2.22) 1.65 (1.56-1.76) 1.27 (1.19-1.36) 1.39 (1.28-1.52)
9 0.72 to 1.03 159 1.63 (1.57-1.70) 2.28 (2.12-2.44) 1.93 (1.77-2.09) 1.42 (1.28-1.58) 1.55 (1.35-1.78)
10 (least healthy) 1.04 to 1.76 80 1.60 (1.52-1.69) 2.06 (1.86-2.28) 1.79 (1.58-2.01) 1.28 (1.11-1.46) 1.34 (1.11-1.61)
Open in a new tab
a

Data presented as age-adjusted mortality rate ratios (with 95% confidence interval. All cancer mortality rates are for the years 2018-2022.

b

Clusters are arranged in order from most to least healthy characteristics.

c

Represents the total number of counties in each cluster for all cancer sites combined. The number of counties for each outcome is: all cancer – 3,079; lung cancer – 2,745; colorectal cancer – 2,180; breast cancer – 1,778; prostate cancer – 1,612.

Footnotes

Earlier versions of this paper were presented at the Catchment Area Data Excellence (CADEx) conference in January 2025 (Coronado, CA) and the American Society of Preventive Oncology (ASPO) conference in April 2025 (Philadelphia, PA).

REFERENCES

RESOURCES