Poverty and population health – The need for A Paradigm shift to capture the working poor and better inform public health planning

Abstract

Background:

Community-level socioeconomic disparities have a significant impact on an individual’s health and overall well-being. However, current estimates for poverty threshold, which are often used to assess community-level socioeconomic status, do not account for cost-of-living differences or geography variability. The goals of this study were to compare geographic county-level overlap and gaps in access to care for households within poverty and working poor designations.

Methods:

Data were obtained for 21 continental United States (US) states from the United Way’s Asset Limited, Income Constrained, Employed (ALICE) households for 2021. Raw data contained the percentage of households at the federal poverty level, the percentage of households at the ALICE designations (working poor), and the total households at the county level. Local Moran’s I tests for spatial autocorrelation were performed to identify the clustering of poverty and ALICE households. These clusters were overlaid with a 30-min drive time from critical access hospitals’ physical addresses.

Findings:

County-level clusters of ALICE (working poor) households occurred in different areas than the clustering of poverty households. Of particular interest, the extent to which the 30-min drive time to critical care overlapped with clusters of ALICE or poverty changed depending on the state. Overall, clustering in ALICE and poverty overlapped with 30-min drive times to critical care between 46 and 90% of the time. However, the specific states where disparities in access to care were prominent differed between analyses focused on households in poverty versus the working poor.

Interpretations:

Findings highlight a disparity in equitable inclusion of individuals across the spectrum of socioeconomic status. Furthermore, they suggest that current public health programming and benefits which support low socioeconomic populations may be missing a vulnerable sub-population of working families. Future studies are needed to better understand how to address the health disparities facing individuals who are above the poverty threshold but still struggle economically to meet based needs.

1. Introduction

Poverty and socioeconomic disparities are one of the most omnipresent public health issues across the globe (Marmot and Wilkinson, 2005). Community-level socioeconomic disadvantage can have a considerable impact on individuals and has been causally linked to increased risk of disease, disability, and death (Marmot and Wilkinson, 2005; Grant, 2005; Levin, 2017; Braveman et al., 2010; Han et al., 2015). In short, individuals experiencing poverty live shorter lives and experience more hardships (Marmot and Wilkinson, 2005; Petrovic et al., 2018; Stringhini, 2010; Stringhini et al., 2017). As such, the United States (U.S) has prioritized a concerted effort to mitigate health inequities since the 1980’s (Penman-Aguilar et al., 2016). While disparities have been addressed situationally, inequities persist and have grown considerably in certain population sub-groups (Weinstein et al., 2017; Anderson and Institute of Medicine (U.S.), 2012).

The U.S. Census Bureau defines a household as living in poverty if the difference between their income and the poverty threshold for their family size is less than zero (United States Census Bureau, 2023). According to a recent January 2023 U.S. Census Bureau report, 11.6% (n = 37.9 million) of Americans are living below the poverty threshold (Creamer et al., 2022). Unfortunately, this estimate barely scratches the surface of financial hardships in the US. In particular, U.S. Census Bureau estimates for poverty threshold do not account for cost-of-living differences or their variability by geography (United States Census Bureau, 2023). As such, regions with a historical context for disadvantage, including but not limited to racism or environmental exploitation (e.g., Appalachia, Mississippi Delta, Tribal Nation Lands, etc.), have a disproportionate number of communities living in financial hardships (University of Michigan Poverty Solutions, 2020). Furthermore, limited consideration for cost-of-living expenses in poverty thresholds has also resulted in a misrepresentation of financial need with respect to basic necessities such as housing, childcare, food, transportation, healthcare, and more (United for ALICE, 2023a).

To address these shortcomings, United for Asset Limited, Income Constrained, Employed (ALICE) developed a new metric which can be used in combination with the existing federal poverty thresholds to better understand the extent of the population experiencing financial hardships (United for ALICE, 2023a). Not only does ALICE incorporate differences in cost-of-living expenses by geographic location, but it also includes costs for everyday essentials that give a more accurate representation of the financial situation for households (Hoopes et al., 2020). The most current ALICE report, published in 2023, identified 41% (n = 52.5 million) of the 127 million US households as struggling to afford housing, childcare, health care, transportation, food, and more (United for ALICE, 2023b). More specifically, 36.3 million households earned more than federal poverty limits yet made less than the cost of living for their respective county of residence (United for ALICE, 2023b). This was compared to the 16.2 million households which met the federal poverty level in 2021 (United for ALICE, 2023b).

However, despite the availability of ALICE data for public use and its ability to better understand socioeconomic vulnerability, no public health studies indexed in PubMed have incorporated these data (searched February 6th, 2023, using mesh terms Asset, Limited, Income, Constrained and Employed). This is problematic because, in addition to the low socioeconomic status being associated with an increased risk of morbidity and mortality, low socioeconomic status indicators are also utilized to inform policy and financial decisions aimed at reducing health care disparities (GRAHAM, 2004). Therefore, the specific communities that are identified as most at risk of health disparities, and as such would be the focus of interventions, may change depending on which socioeconomic status indicator is utilized.

The overall goal of this study was to investigate the use of ALICE as a measure of socioeconomic status in identifying at-risk groups for health disparities as compared with the federal poverty level. Specifically, the aims were to 1) identify and compare the distribution of households at the poverty threshold, those outside the poverty threshold that are ALICE, and those who experience either poverty or ALICE, and 2) determine the extent to which households experiencing varying levels of financial hardship overlap with 30-min drive times to critical access hospitals providing inpatient services within rural, medically underserved areas.

2. Methods

2.1. Study measures

United for ALICE was founded by the United Way of Northern New Jersey to better understand the struggles of working families (United for ALICE, 2023a). During their research on the financial hardships of these families, they developed the ALICE threshold, a comprehensive measure of economic hardship that includes adjustments for costs of living and accounts for geographic variation across the country (Hoopes et al., 2020). The ALICE threshold goes beyond the federal poverty level and is able to capture households (individuals and families) that are working but still struggling to afford basic necessities (Hoopes et al., 2020). Detailed information about the ALICE threshold, including the methodology and data sources, has been described elsewhere (Hoopes et al., 2020; Hoopes and Treglia, 2019). Briefly, the ALICE threshold is a measure of the household budget necessary for an individual or family to live and work (United for ALICE, 2023a). It includes costs for housing, food, health care, childcare, transportation, technology, taxes, and a contingency fund equal to 10% of the household budget for emergencies (Hoopes et al., 2020). The ALICE threshold is calculated annually using publicly available data, such as the Bureau of Labor Statistics and the US Census Bureau and is adjusted based on household types and geographic location (Hoopes et al., 2020).

Currently, 22 states and the District of Columbia have partnered with United for ALICE to engage in research and advocacy about the financial hardships of ALICE families (United for ALICE, 2023a). Each partnering state/district have profiles and publicly available data up until 2021 on the United for ALICE’S project webpage (United for ALICE, 2023a). For our study, data was limited to the states within the continental US who are currently partners with United for ALICE. These states include Arkansas, Delaware, Florida, Idaho, Illinois, Indiana, Iowa, Maryland, Michigan, Mississippi, New Jersey, New York, Ohio, Oregon, Pennsylvania, Tennessee, Texas, Virginia, Washington, West Virginia, and Wisconsin.

Critical access hospitals are inpatient facilities that are designed to provide increased access to inpatient and emergency care to underserved individuals in rural communities (Rural Health Information Hub; Joynt et al., 2011). These facilities are an invaluable asset and help provide more equitable access to care to populations with increased disease burdens (Joynt et al., 2011). In order to qualify as a critical access hospital, and be eligible for the cost-based reimbursement for Medicare services, facilities must be located in a state that has established a state rural health plan and be located in a rural, underserved area within the state (Rural Health Information Hub). Eligible hospitals must also have no more than 25 acute care beds, be located more than 35 miles from the nearest hospital, provide emergency services 24/7, and have an average length of stay of 96 h or less for acute care patients (Rural Health Information Hub; Joynt et al., 2011). Of the 21 continental US States with ALICE data, 18 have established a state rural health plan and are eligible to have critical access hospitals. This includes Arkansas, Florida, Idaho, Illinois, Indiana, Iowa, Michigan, Mississippi, New York, Ohio, Oregon, Pennsylvania, Tennessee, Texas, Virginia, Washington, West Virginia, and Wisconsin.

For our analysis, data on critical access hospital locations, ALICE thresholds, and poverty thresholds were merged at the county level using the 2020 Tiger Line County shape files (U.S. Census Bureau, 2020). First, locations of critical assess hospitals were obtained from the Health Resources & Administration (HRSA) database and geocoded based on exact address location (Health Resources and Services Administration, 2023). Then, data on ALICE thresholds and poverty thresholds in each ALICE state at the county level were obtained from United for Alice and merged with the critical access hospital locations.

2.2. Statistical analysis

Geographic variability in the percent of households living at the 1) federal poverty threshold, 2) at the ALICE threshold, and 3) ALICE or poverty threshold combined were displayed in separate thematic maps. For this study, each measure of socioeconomic status was mapped separately according to Jenks natural breaks classification using five class breaks. Jenks is designed to use the natural breaks in the data to minimize the within-class variance and maximize the variance between classes (Brewer and Pickle, 2002). It is a commonly used classification system for non-normally distributed data, such as socioeconomic data (ESRI. Classification methods).

The study area was broken into two separate sections, the northwest (Washington, Oregon, and Idaho) and the central/eastern (Arkansas, Delaware, Florida, Illinois, Indiana, Iowa, Maryland, Michigan, Mississippi, New Jersey, New York, Ohio, Pennsylvania, Tennessee, Texas, Virginia, West Virginia, and Wisconsin) due to the location of the ALICE states. Tests for spatial autocorrelation (e.g., spatial clustering) were conducted using global statistics (tests clustering across each section) and local statistics (tests for clustering within each section) for each section separately. Moran’s I was conducted to assess global clustering, while local Getis Ord (G*) was used to identify county-level clusters of high and low percent of households at the 1) federal poverty, 2) ALICE, and 3) federal poverty plus ALICE levels. Both tests of spatial autocorrelation incorporated a nearest neighborhood spatial weight (Getis and Aldstadt, 2010). The optimal number of neighbors was calculated for each section separately based on the average number of neighbors for each county, resulting in 4 neighbors in the northwest and 6 neighbors in the central/eastern. Values of Moran’s I range from +1 to −1, with positive values indicating spatial clustering, negative values indicating spatial spreading, and zero indicating complete spatial randomness. Values of Getis-Ord (Gi*) are z-scores, which provide an easy way of detecting clustering of high value (hot spots) or low values (cold spots) at various degrees of statistical significance (90%, 95%, or 99% confidence level) (Environmental Systems Research Institute (ESRI), 2023). For this study, hot spots were clusters of high values with a confidence level of 95% or higher. All statistical analysis was conducted using R statistical software (version 4.2.1).

All thematic and cluster maps were created in ArcGIS Pro 2.9.2 (ESRI, Redlands CA). Cluster maps for each measure of socioeconomic status were overlaid with a critical access hospital 30-min drive time layer created in ArcGIS Online. The rural drive time analysis, under the proximity tools, estimates drive time in minutes from or to a geographic location, with a known latitude and longitude, adjusting for rural routes or unpaved roads (Environmental Systems Research Institute (ESRI)). The drive time analysis excluded states which do not have a state rural health plan (Delaware, New Jersey, and Maryland). The final mapped display provided an in-depth community-level snapshot of counties with a high percentage of the population living in poverty, ALICE (working poor), or either and their access to a critical access hospital.

Lastly, a non-spatial test of agreement, Cohen’s Kappa, was conducted to estimate the extent of agreement between the classifications of counties as hot spots for each measure of socioeconomic status (poverty, ALICE, poverty plus ALICE). This summary statistic provides a national estimate to better understand how public health planning targeting areas of poverty serves communities with a high percentage of the population designated as working poor.

Role of the funding source

The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.

3. Results

The percentage of households at the ALICE threshold, federal poverty level, and both poverty and ALICE thresholds varied geographically throughout the nation (Fig 1). The percentage of households at the ALICE threshold ranged from 18.8 to 75.0 in a county, with pockets of high percentages found throughout the nation, particularly in Idaho, Texas, and West Virginia. The percentage of households at the federal poverty level ranged from zero to 42.1 in a county, with a greater concertation of high percentages found in Texas, Mississippi, and West Virginia. When combined, the percentage of households of poverty and ALICE ranged from 18.8 to 79.9, with pockets of high percentages found in Mississippi, West Virginia, Idaho, and Texas.

Fig. 1.

County level choropleth maps displaying percentage of households at ALICE, federal poverty level, and poverty plus ALICE.

In the northwest section, global spatial autocorrelation revealed significant clustering of the percentage of ALICE households (Moran’s I = 0.26, p < 0.001) and the percentage of poverty plus ALICE households (Moran’s I = 0.29, p < 0.001) but not the percentage of poverty households (Moran’s I = 0.02, p = 0.33). In the central/eastern section, global spatial autocorrection revealed significant clustering for ALICE (Moran’s I = 0.33, p < 0.001), poverty (Moran’s I = 0.50, p < 0.001), and poverty plus ALICE (Moran’s I = 0.53, p < 0.001). Over the entire study area, local spatial autocorrelation identified 184 counties that were ALICE hot spots, 189 counties that were poverty hot spots and 216 counties that were poverty plus ALICE hot spots. When comparing agreement between county hot spots by socioeconomic measure, there was only fair agreement between ALICE and poverty (k = 0.38, 95% CI: 0.32, 0.42). There was substantial agreement between ALICE and poverty plus ALICE (k = 0.67, 95% CI: 0.62, 0.73), and poverty and poverty plus ALICE (k = 0.70, 95% CI: 0.65, 0.76).

Local county-level cluster results and 30-min drive times from the 624 critical access hospitals across 18 states are displayed for ALICE households in Fig. 2, poverty households in Fig. 3, and poverty plus ALICE households in Fig. 4. Significant clusters of counties with a greater percentage of households at each socioeconomic measure are indicated by deeper shades of red. The 30-min drive time areas are identified by the white enclosed areas around the critical access hospital, and the light grey indicates areas greater than 30 min from a critical access hospital. Overall, the percentage of poverty households outside a 30-min drive to critical access hospitals ranged from 47% in Illinois to 90% in Texas. Among ALICE households and poverty-plus ALICE households, the percentage outside a 30-min drive time to a critical access hospital ranged from 54% in Mississippi to 97% in Idaho.

Fig. 2.

Thirty-minute drive time area from each critical access hospital overlapped on ALICE hot spot maps and the percentage of hot spots inside and outside of 30 min per state.

Fig. 3.

Thirty-minute drive time area from each critical access hospital overlapped on poverty hot spot maps and the percentage of hot spots inside and outside of 30 min per state.

Fig. 4.

Thirty-minute drive time area from each critical access hospital overlapped on poverty plus ALICE hot spot maps and the percentage of hot spots inside and outside of 30 min per state.

4. Discussion

This study is the first to compare geographic distribution in the clustering of the percentage of households at the federal poverty, ALICE (working poor), and poverty plus ALICE thresholds. Findings suggest that communities with a higher percentage of working poor households are distinct and located in different regions of the country than communities with a high percentage of households below the federal poverty limits. Furthermore, the results potentially suggest that public health planning or the creation of critical access hospitals targeting communities with a high degree of poverty would not address health inequities present in communities with a high percentage of working poor households. Nationwide, Cohen’s Kappa statistic revealed that county clusters of high percent of households in poverty were not co-located with county clusters of high percent of working poor households. Overall, results suggest that a more inclusive definition of low socioeconomic status, which includes households of working poor as well as those living below the federal poverty limit, may be needed to more effectively address health disparities across the country.

This study identified significant geographic variation of households at several levels of socioeconomic status in the United States. The heterogeneous spatial distribution of socioeconomic status found in our study was similar to that of previous studies (Curtis et al., 2019; Wang et al., 2012). Curtis et al. examined how the spatial distribution of poverty at the county level in the Midwest varied based on dominant industries, while Wang and colleagues examined the spatial variation of poverty across metropolitan and non-metropolitan areas nationwide (Curtis et al., 2019; Wang et al., 2012). However, these studies only included the federal poverty threshold as the indicator of socioeconomic status. The inclusion of ALICE as another indicator of socioeconomic status in our study adds to previous work by indicating that poverty alone might not be a sufficient indicator of socioeconomic inequality in all areas of the United States. In the northwest area of the United States, which includes Oregon, Washington, and Idaho, the geographic variation of poverty households was not statistically significant, but the variation of ALICE and poverty plus ALICE households was. This is important because policymakers utilize indicators of socioeconomic status and other social determinants of health to inform policy and financial decisions aimed at reducing healthcare disparities (GRAHAM, 2004). Poverty is a prominent indicator of socioeconomic status and is commonly incorporated within public health research (van Domelen, 2007). Our findings indicate that by only including poverty as an indicator of socioeconomic status, policymakers and researchers are not able to properly identify the communities of high need. Altogether, this emphasizes the importance of understanding how socioeconomic status is defined at the community level and its distribution across communities of interest (Small and Newman, 2001; rgen and Blasius, 2003).

Our study found that the geographic variation of households at the ALICE threshold was significantly different from households at the federal poverty level. Not only was there just a fair amount of agreement between all ALICE and poverty county-level hot spots across the country, but some states, such as Florida, only had clusters of one indicator of socioeconomic status. Understanding where clusters of households of each socioeconomic measure are located is essential in ensuring effective policy and outreach efforts. Our results indicate that if policy and intervention efforts are only focused on areas of high poverty, as measured by the federal poverty level, they will not be sufficient in providing assistance to ALICE (working poor) households. Even though ALICE households are above the poverty threshold, they struggle economically to meet basic needs, especially regarding health and childcare (Hoopes and Treglia, 2019). ALICE families often delay medical care due to costs, lack of insurance, and/or transportation issues (United for ALICE, 2023c). Unfortunately, these households are ineligible for many public assistance programs, which are based on the federal poverty threshold (Hoopes and Treglia, 2019). Consequently, ALICE families do not make enough money to address their needs but make too much to be eligible for public assistance programs.

The conclusions from this study provide data to support a broader classification of socioeconomic status, which includes both ALICE and poverty thresholds. Figs. 2-4 display the percentage of hot spots within each classification of socioeconomic status within and outside a 30-min drive time from a critical access hospital. We found that there is a wide variation in the percentage of households who lack adequate access to care across the country for each classification of socioeconomic status. Adequate access to healthcare is a major concern for numerous communities, especially those with a large number of households that have low socioeconomic status. Critical access hospitals are a safety net for these communities and are able to increase access to inpatient and emergency care and help to improve the overall health of the population (Joynt et al., 2011). As a result, policymakers need to be able properly identify and assess areas of need in order to decide on locations of future healthcare centers, including critical access hospitals (Hong et al., 2022).

There are several strengths and limitations to this study. To the best of our knowledge, this is the first public health study that utilizes ALICE threshold data. This study provides a novel contribution to aid public health practitioners by highlighting a need to critically examine what low socioeconomic status means for the populations they serve. Additionally, we provide an example of how a broader inclusion of households just over the federal poverty limit can decrease disparities between access to care and communities experiencing low socioeconomic status. However, our study is vulnerable to the modifiable areal unit problem, which postulates that the results of a study are biased based on the arbitrary choice of spatial aggregation unit (Tuson et al., 2019). We chose counties as the geographic unit of analysis, but our conclusions may not apply to every community, particularly those based on different geographic units. Ultimately, further population-level research is warranted to understand the wide variation of access to care for each classification of socioeconomic status within each state. Future research should investigate how the poverty plus ALICE threshold is related to the incidence of chronic or infectious diseases and how this may impact other types of healthcare services (e.g., childcare services) for communities with a larger number of working poor households.

5. Conclusion

Our study examined the geographic distribution of households at the ALICE (working poor) and federal poverty in 21 states across the country. We found different geographic patterns for each indicator of low socioeconomic status. Importantly, the different geographic patterns resulted in unique clusters of community-level poverty and working poor households within 30-min drive times from critical access hospitals. As a result, policymakers should consider the inclusion of ALICE as a measure of socioeconomic status to more accurately identify households that are economically vulnerable. These findings highlight an increased need to evaluate how socioeconomic status is defined and operationalized within public health programming and policymaking. While this study was conducted in certain states, it can serve as a model for all areas of the country to better understand how current public health programming efforts are meeting the needs of the working poor. These households potentially experience greater health and social vulnerability than households in poverty due to their restricted access to public assistance programming, and more effort needs to be made to address their needs.

Data availability

All data are publicly available and links have been provided in the manuscript.