Rural-urban disparities in health outcomes, clinical care, health behaviors, and social determinants of health and an action-oriented, dynamic tool for visualizing them

While rural-urban disparities in health and health outcomes have been demonstrated, because of their impact on (and intervenability to improve) health and health outcomes, we sought to examine cross-sectional and longitudinal inequities in health, clinical care, health behaviors, and social determinants of health (SDOH) between rural and non-rural counties in the pre-pandemic era (2015 to 2019), and to present a Health Equity Dashboard that can be used by policymakers and researchers to facilitate examining such disparities. Therefore, using data obtained from 2015–2022 County Health Rankings datasets, we used analysis of variance to examine differences in 33 county level attributes between rural and non-rural counties, calculated the change in values for each measure between 2015 and 2019, determined whether rural-urban disparities had widened, and used those data to create a Health Equity Dashboard that displays county-level individual measures or compilations of them. We followed STROBE guidelines in writing the manuscript. We found that rural counties overwhelmingly had worse measures of SDOH at the county level. With few exceptions, the measures we examined were getting worse between 2015 and 2019 in all counties, relatively more so in rural counties, resulting in the widening of rural-urban disparities in these measures. When rural-urban gaps narrowed, it tended to be in measures wherein rural counties were outperforming urban ones in the earlier period. In conclusion, our findings highlight the need for policymakers to prioritize rural settings for interventions designed to improve health outcomes, likely through improving health behaviors, clinical care, social and environmental factors, and physical environment attributes. Visualization tools can help guide policymakers and researchers with grounded information, communicate necessary data to engage relevant stakeholders, and track SDOH changes and health outcomes over time.


Introduction
Despite overall improvements in mortality rates in the United States (US) between 2000 and 2019 (before the pandemic), disparities between rural and large metropolitan areas persist, and disparities in overall age-adjusted mortality rates tripled during that period [1].The rural mortality penalty has increased, reducing lifespans in rural, as compared to urban, settings [2], with increasing rural-urban disparities in all-cause mortality having been shown among Medicare beneficiaries dually enrolled in Medicaid [3].Residents of high-poverty rural counties face a particularly steep rural mortality penalty [4].These rural-urban disparities in health outcomes span the age spectrum and disease states: along with differences in underlying health risks and behaviors, socioeconomic factors are associated with higher rates of the five leading causes of death [5], higher infant mortality rates [6], higher rates of cardiovascular disease mortality [7], and COVID-19-related deaths [8] in rural as compared to urban counties.
Socioeconomic variables have been shown to account for much of the mortality [9,10] and self-rated physical health status [11] differences between rural and urban populations.Possibly contributing to those disparities, the high relative use of preventable emergency department visits and hospitalization rates [12] and relatively low cancer screening rates [13] in rural settings suggests an unmet need for high-quality ambulatory care in rural areas.Nevertheless, among older Medicare beneficiaries, at the hospital referral region level primary care seems to be of similar quality and lower cost in rural as contrasted with urban settings after considering the role of local area deprivation [14].This suggests that other factors, like social determinants of health (SDOH), may be contributing to these disparities more than clinical care quality.
SDOH are the non-medical factors that influence health outcomes: they encompass the conditions in which people are born, grow, work, live, and age as well as the wider set of forces and systems shaping the conditions of daily life [15].SDOH can shape individuals' health behaviors, which can then shape health outcomes [16].The impact of SDOH as drivers of rural-urban (and racial) disparities across numerous health indicators in the US calls for a multi-sectoral approach to addressing SDOH in an effort to improve the health of the nation [17].The distribution of economic prosperity among U.S. communities has undergone significant changes in recent decades, resulting in heightened inequality [18].This has led to a growing interest in developing policies and resources that support both "places" and "people," particularly in underserved communities [19,20].These policies recognize that socioeconomic conditions are significant determinants of health and that ameliorating SDOH disparities may improve health at the population level [21].However, formulating an effective policy response requires identifying and targeting areas where interventions are most greatly needed, are achievable, and might have the largest and most sustained impact on health equity.
The aims of this study were to examine cross-sectional and longitudinal inequities in health, clinical care, health behaviors, and SDOH between rural and non-rural populations in 2015 and 2019 (the pre-pandemic era) and to develop and present a Health Equity Dashboard that can be used by policymakers and researchers to visualize and examine disparities across multiple SDOH domains and across time.While prior studies examined rural and urban health differences using data from County Health Rankings [22,23], to our knowledge, this is the first study to examine and visualize these differences across multiple years.

Data
We sought to identify cross-sectional and longitudinal inequities in health, clinical care, health behaviors, and SDOH associated with rural status at the county level using data from County Health Rankings [24].For 3,131 counties in the 50 US states and Washington, DC (wherein 325,711,203 people lived in 2019), we collected 33 county level attributes obtained from the 2015-2022 County Health Rankings across five health and SDOH domains: Health Outcomes, Clinical Care, Health Behaviors, Physical Environment, and Social and Economic Factors.We limited measures to those available for two time periods: approximately 2015 and approximately 2019.
Table 1 provides the measure name, definition, orientation, periods of data collection, and year interval, across the five domains.Table 2 shows the original sources from which County Health Rankings obtained these measures.We used the 2013 Urban-Rural Classification Scheme for Counties (based on the 2010 Census) [25] from the Centers for Disease Control and Prevention's National Center for Health Statistics to classify counties as urban (codes 1 and 2), suburban (codes 3 and 4), or rural (codes 5 and 6).

Analysis
Using Analysis of Variance (ANOVA), for both time periods we compared the health and SDOH measures across the rural-urban continuum (there were 436 urban counties, 729 suburban counties, and 1966 rural counties).Further, for each county, we calculated the change in values between 2015 and 2019.Finally, we calculated the ratio of values for rural to urban counties for the first and second data collection period (approximately 2015 and 2019) and provided an indication of whether the gap between the least and most prosperous counties was widening, narrowing, or staying the same.We used SPSS v 28 (released 2022, Armonk, NY: IBM Corporation) for all analyses.
We followed STROBE guidelines in writing the manuscript.

Application and visualization
With the data we collected, we used Microsoft's PowerBI platform to develop a Health Equity Dashboard that could be used by policymakers and researchers to examine disparities in single SDOH measures within domains, to develop their own indices of up to five measures across a single or multiple domains (calculated at the national or state level), and to examine the relationship between index values and county socio-demographic characteristics.

Results
Table 3 compares urban, suburban, and rural measures for 2019 data.For 25 of 33 measures, we found a statistically significant and progressive worsening of values when moving from urban to suburban to rural counties.For two measures (low birth weight and preventable hospitalization rate), rural values were worse than urban, but the pattern was not progressive (and not statistically significant in the case of low birth weight).For chlamydia cases (a behavioral risk factor that estimates the prevalence of unprotected sex), excessive drinking, air quality, severe housing problems, and membership association rates, values improved with increasing rurality.For insufficient sleep, rural values were better than urban values, but there was not a progressive (or a statistically significant) pattern.We conducted the same analysis using 2015 data (Table 4).Findings were similar: for 24 measures, there was a progressive and (save low birth weight) statistically significant worsening of values with increasing rurality; for four measures, rural values were worse than urban values, but there was no progressive pattern; for five measures, values improved with increasing rurality; and for chlamydia cases, rural values were better than urban ones, but there was no progressive pattern.Table 5 shows that diabetes prevalence, preventable hospitalization rate, and deaths due to injury all improved progressively with increasing rurality between the earlier and later period.For measures of years potential life lost, the PCP workforce, uninsurance, chlamydia cases, and children in poverty, rural values improved more than urban ones, though there was no progressive pattern.However, 22 measures worsened in a progressive fashion with increasing rurality and three measures worsened more in rural counties than in urban ones, but without a progressive pattern.
Table 6 shows the ratio of rural to urban values in the earlier and later periods.During that time, the rural-urban gap widened for 15 measures, narrowed for 12 measures, and did not change for 6 measures.Gaps tended to widen in the health outcomes and health behaviors domains and tended to narrow in the clinical care domain.   of index scores with increasing rurality and in counties with higher proportions of Blacks.In the time series comparison, they would find that Mississippi had the highest percentage of adults in fair or poor health (26.66%) while Connecticut had the lowest percentage (13.98%);further, they would discover that between 2015-2019, the percentage of adults in fair or poor health increased in every state, the most in Florida (from 16.44% to 22.82%) and the least in Massachusetts (from 13.26% to 14.59%).Fig 2 shows an example of how the dashboard might be used to differentiate counties with greater relative need, within a state.For instance, when recalculating the previously-defined index at a state level, policymakers or researchers interested in Mississippi-a state in which virtually every county was in the worst health index quintile from a national perspective-could  ) maps, to highlight only rural counties.There, from a national perspective, they might find a high proportion of rural counties in the Midwest, worse index scores within rural counties across the southeastern United States, and that about 11 million people lived in rural counties in the worst index quintile (compared to about 9 million in suburban counties and about 6 million in urban counties).Further, nationally or within a state, they might examine relationships between index scores and, 2013 Rural-Urban Continuum Codes [26], finding, in

Discussion
We found that rural counties overwhelmingly had worse measures of health and SDOH at the county level.With few exceptions, many of the measures we examined were getting worse between 2015 and 2019 in all counties; in addition, measures generally got relatively worse in rural counties, resulting in the widening of rural-urban disparities in these measures during this period.In the health behaviors domain, while chlamydia cases were lower and increasing at a slower rate in rural settings, the rural advantage in excessive drinking is diminishing.The good news for rural dwellers is in the physical environment realm, where air quality is better and housing problems are fewer.While the membership association rate is higher in rural settings, that advantage is diminishing as well.While our findings may not be unexpected, the opportunity to examine numerous SDOH measures together, across time, and through index development may offer policymakers and researchers an opportunity to consider where best to focus efforts and which factors to focus on, across the country or within a state.Further, the ability to consider the potential market-as represented by population distributions and numbers of counties-might inform policymakers or researchers interested in health equity on the overall impact proposed programs might have.
Our study has several limitations.First, we used one coding system to categorize counties into rural, suburban, or urban.
There are multiple systems to designate counties and places within a rural-urban continuum [27], and different ways to interpret what "rural health" means [28].Findings may be different when using different rural-urban continuum classification systems.Second, our results are derived from data in two relatively close time periods; studies of different time periods may have different results.Importantly, we evaluated periods before the COVID-19 pandemic; reports suggest that economic and health inequities have increased since COVID-19 began [29].Therefore, our results might underestimate current inequities.Third, measures are not adjusted for local demographic factors that may impact measure values.For example, Blacks are more likely than Whites to have diabetes [30], lower life expectancy [31], and low birth weight babies [32].To the extent that racial disparities are conflated with the rural-urban disparities we found, our analysis is limited.However, while demographic factors may be partially explanatory [33], they offer policymakers no pragmatic solutions: changing the demographic makeup of a county cannot be a reasonable policy platform.Finally, our findings are associative and not causative.
Despite these limitations, our findings highlight the need for policymakers to prioritize rural settings for interventions designed to improve health outcomes, likely through improving health behaviors, clinical care, social and environmental factors, and physical environment attributes.Timely, accurate, and high-quality data are a critical component of public health decision making [34].Data visualization tools can help the effective delivery and translation of data, thereby engaging key stakeholders and prompting action [35,36].By leveraging these tools, policymakers can make more informed decisions that are grounded in objective evidence, ultimately leading to better outcomes for all stakeholders.As all policy decisions have population health implications [37], interventions should be evaluated for return on investment to population health and reduction of rural-urban disparities, as well as any other policy goals.Tools like the Health Equity Dashboard (publicly available at aka.ms/healthequity) can facilitate those evaluations.Hopefully, by guiding policymakers with grounded information in a way that can be personalized to a community's interest and consumed, shared, and tracked visually, over time, policies can be developed and focused to measurably improve population health in areas where the greatest health inequities exist and those with the greatest unmet social needs reside.

Fig 1
Fig 1 shows an example of how nationally oriented policymakers and researchers might use the Health Equity Dashboard (publicly available at aka.ms/healthequity).The user could create a map of nationally calculated index values (in this example, composed of five equally weighted, 2019 measures (life expectancy, percentage of adults with obesity, uninsurance rate, income inequality, and air quality)) at the county level (Fig 1, top), explore the distribution of index values (in quintiles) across the rural-urban continuum (Fig 1, middle), and examine a measure's mean value at the state level, over time (Fig 1, bottom).Such users would discover that Los Alamos County (rural) in New Mexico had the best index score in the nation and Brooks County, Texas (also rural) had the worst index score.They would discover a worsening

Fig 1 .
Fig 1.A national map view of the Health Equity Dashboard, showing: the 2019 distribution in quintiles of a national index formed from five measures (life expectancy, percentage of the adult population that is obese, percentage of the population that is uninsured, income inequality, and air quality) and the percentage of rural, suburban, and urban counties with index values in the best to worst quintiles (top); the county-level distribution of that index across urban, suburban, and rural county designations (middle); and a comparison of the 2015 and 2019 values of one measure (percentage of the population in fair or poor health) at the state level (bottom).The Health Equity Dashboard tool is publicly available at: aka.ms/healthequity.The base layers for the maps are Shapefiles from the US Census TIGER file repository.https://doi.org/10.1371/journal.pgph.0002420.g001

Fig 2 .Fig 3 .
Fig 2. A state-level view of the Health Equity Dashboard, showing, for Mississippi: the 2019 distribution in quintiles of a Mississippi-specific generated index formed from five measures (life expectancy, percentage of the adult population that is obese, percentage of the population that is uninsured, income inequality, and air quality) and the percentage of rural, suburban, and urban counties with index values in the best to worst quintiles (top); the county-level distribution of that index across urban, suburban, and rural county designations (middle); and a comparison of the 2015 and 2019 values of one measure (life expectancy at birth) at the state level (bottom).The Health Equity Dashboard tool is publicly available at: aka.ms/healthequity.The base layers for the maps are Shapefiles from the US Census TIGER file repository.https://doi.org/10.1371/journal.pgph.0002420.g002

Table 2 . Measures collected, with domain and original source of data that was compiled in County Health Reports.
https://doi.org/10.1371/journal.pgph.0002420.t002

Table 3 . Results for the later data collection period (around 2019), across the rural-urban continuum.
All ANOVA differences were statistically significant at p<0.001 except those marked with * which are p<0.05 and those marked # which are not statistically significant.Values in italics did not follow a progressive worsening of measure value with increasing rurality.Values in bold indicate measures where there was improvement in measure values with increasing rurality. https://doi.org/10.1371/journal.pgph.0002420.t003

Table 4 . Results for the earlier data collection period (around 2015), across the rural-urban continuum.
All ANOVA differences were statistically significant at p<0.001 except those marked # which are not statistically significant.Values in italics did not follow a progressive worsening of measure value with increasing rurality.Values in bold indicate measures where there was improvement in measure values with increasing rurality.
Mississippi.Further, they would discover that the prevalence of worst quintile index scores was highest in rural counties which also were much more likely to be experiencing economic distress.Finally, they could see that, between 2015 and 2019, life expectancy at birth decreased

Children in poverty -4.04 -4.68 -4.67 Deaths due to injury 12.48 11.69 8.03
Values in italics did not follow a progressive worsening of measure value when moving to increasing rurality.Values in bold indicate measures where there was improvement in measure values with increasing rurality.https://doi.org/10.1371/journal.pgph.0002420.t005from 73.13 to 72.15 years across counties, increased for 15 Mississippi counties, ranged from 73.00 in Prentiss County to 79.21 in Lamar County in 2019, and ranged from 71.92 in Attala County to 78.97 in Rankin County in 2015.Finally, should a policymaker or researcher want to examine and compare index or measure values only for rural counties, they could select 'rural' on national (Fig 3, top) or state (Fig 3, middle