Association of Social Service Spending,Environmental Quality,and Health Behaviors on Health Outcomes

Abstract

Recent studies on state-level spending on social services have shown that states with higher ratios of social to health care spending were associated with better health outcomes. This study extends this work by examining the association of specific elements of social service spending and other determinants of health, such as health behaviors, education, and environmental factors at the metropolitan/city level, on several measures of health outcomes between 2005 and 2014. This study found that several potential determinants of health including exercise, air pollution, smoking, per pupil educational spending, and several types of social service spending were associated with improvements in health outcomes. These health outcomes included age-adjusted mortality, chronic disease prevalence, days of poor health, and obesity rates. The results suggest that a broader strategy beyond health care that includes investments in social services, improved environmental quality, and health behaviors could improve the health of communities.

Previous international studies and a recent study of the states have found that jurisdictions with higher rates of social services spending relative to health care spending are associated with better health outcomes.¹ These social services include public health spending, support services at home, spending on parks and recreation, and public welfare, among others. This line of work follows earlier work on the underlying determinants of health from Marmot and colleagues.² The social determinants of health body of work has explored the impact that health behaviors, housing and other social services, physical/chemical, the environment, transportation, education, socioeconomic position, and spending on health care services have on health outcomes. Though there is a substantial literature outlining the importance of these social determinants on health outcomes, few empirical studies have examined their collective impact on health outcomes. For example, a recent study by McCullough and Leider examined the role of income on the amount of county-level social service spending, but does not consider environmental quality and health behaviors.³ The county health rankings body of work relies on a population health approach that ranks counties with respect to health behaviors, access to health care, social and economic factors, and the physical environment. These data are used to rank counties from healthiest to least healthy but do not show the individual impact of each of these determinants on health outcomes. In contrast, this paper extends these previous analyses by empirically examining the individual impact of these population-based determinants of health, including environmental, education, social and economic, social services, and health behaviors, on several measures of health at the community level.

A challenge for improving health is to encourage cities and counties to focus on improvements in 5 key areas that affect health outcomes: healthy behaviors, community safety, social and economic factors, and environmental exposures. A key aim of this paper is to provide statistical estimates of the impact these key areas have on the health profile of metropolitan areas nationally.

Methods

Metropolitan statistical area (MSA) data on a variety of health outcome, social service, and health care spending was collected for the years 2005 to 2014. Complete information was collected for approximately 90 MSAs during the 10-year period that included more than 70% of the overall MSA-level population. Ozone levels and some poverty measures were missing from some of the MSAs. To compare, the study team ran a regression of a sample of 90 MSAs and the larger full sample of MSAs that excluded both ozone and poverty measures from the regression. The coefficients from the regressions were virtually identical so the team proceeded with the 70% sample.

A number of data sources were used to compile the data. Information on obesity, physical and mental health, general health status, chronic conditions, smoking status, race/ethnicity, education, any exercise, and health insurance was derived from the Behavioral Risk Factor Surveillance System (BRFSS), Selected Metropolitan/Micropolitan Area Risk Trends (SMART) data. SMART is a subset of the BRFSS created to provide estimates at the MSA level.⁴ SMART poverty level measures were obtained from the Current Population Survey, Annual Social and Economic (ASEC) Supplement for the years 2006 to 2015. The ASEC survey is conducted annually in March with the poverty-related information pertaining to the prior calendar year.⁵ Weighted MSA means were calculated using Stata 14 (StataCorp LLC, College Station, TX) survey commands to account for the complex survey designs of both SMART and ASEC surveys. Cancer mortality and overall mortality, both age adjusted using the 2000 US population, were downloaded from CDC WONDER.⁶ Although the cancer mortality data were available at the MSA level, overall mortality data originated from the Compressed Mortality File at the county level. To obtain an MSA mortality rate, the population weighted average of all counties in the MSA was calculated. Unemployment rates, seasonally adjusted, by MSA were obtained from the Bureau of Labor Statistics.⁷ The weighted annual mean concentration (μg/m³) of fine particulate matter and fourth highest daily maximum 8-hour concentration (parts per million) of ozone came from the Environmental Protection Agency air quality trends by city.⁸ Per pupil total spending in the MSA⁹ and total local government expenditures in the state on public welfare, health and hospitals, police and fire, parks and recreation, and sewerage and solid waste¹⁰ were obtained from the Census Bureau. Per capita real gross domestic product (chained 2009 dollars) and per capita personal health care consumption were from the Bureau of Economic Analysis.¹¹ The cigarette taxes per 20-pack were from taxfoundation.org.¹² All variables are measured at the MSA level, except for state-level variables for per capita state spending on parks and recreation, per capita spending on police and fire, per capita personal health care consumption, and per capita public welfare spending, which were not available at the MSA level, and state cigarette taxes.

Measures

The analysis will be at the MSA level for 2005 to 2014. Because many cities that make community-level policy decisions span several counties, MSAs better reflect the outcomes. The study will examine the impact of traditional measures of access to health care (uninsured rates) on health outcomes as well as a variety of determinants of health including: health behaviors and tobacco taxes, demographics, economics/income, education spending per pupil, environmental quality, and social services spending per capita on education, parks and recreation, and public welfare.

Dependent variables

Health outcome measures at the metropolitan statistical market area

This study has included several MSA-level dependent variables of health outcomes. These outcomes reflect cardiovascular disease risk as well as more general health outcome measures. These measures include: percent obese, average number of days last month not in good physical or mental health, average number of days last month not in good mental health, age-adjusted mortality rate, age-adjusted cancer mortality rate, asthma prevalence, diabetes prevalence, and cardiovascular disease prevalence.

Log regression models were run for these 8 health outcome variables using the following independent (covariates) variables. These variables were categorized into 6 different groups.

Independent variables (covariates)

The study team controls for MSA-level patient demographics that include education, race and ethnicity, and income levels as a percent of poverty. The team is particularly interested in the role that health behaviors, the environment, spending on education, and several measures of social service spending have on average health outcome measure. These covariates are:

Health Behaviors: percent participating in physical activities, smoking prevalence, state cigarette tax

Demographics and Income: indicator variables for level of college education, indicator variables for race and ethnicity, income as a percent of poverty, percent uninsured

Economics: real per capita metropolitan area gross domestic product

Education: education spending per student

Environment: fine particulate matter (PM2.5) – weighted annual mean concentration (μg/m3), ozone concentration

Social Services and Health Care Spending: per capita state spending on health care, per capita state spending on parks and recreation, per capita state spending on police and fire, per capita state spending on public welfare

Data analysis

The study team completed a series of regression analyses using the dependent (log health outcome variables) and independent variables listed. The team explored MSA-level fixed effects models but were limited by the number of years and lack of significant variation in some of the variables over time. Instead, the team estimated log models that included year dummies and a second model that included time period dummies of the pre-recession period (2004–2006), the recession period (2007–2009), and the post-recession period (2010–2014). The results were similar and this study reports the results from the time period dummies. Most of the outcome variables were skewed so the log transformations of the dependent variables were used. The study team also used several lag models (up to 5 years) of the independent variables. Lags were used because spending on social services is not likely to affect health outcomes in the same year. However, spending could affect outcomes in subsequent years. The team also uses lag models to address any concerns over reverse causality. In one version the team used a 2-year lag for the dependent variables and a 1-year lag of the independent variables to account for any potential endogeneity. The results were virtually identical so this study reports results from a 1-year lag and no lag for the dependent variables. Log transformations were used for the per capita health care spending, per capita parks and recreation spending, per capita police and fire spending, and public welfare spending. These spending variables were logged and the remaining independent variables were not. Robust standard errors were estimated using Newey-West standard errors to account for any heteroscedasticity and autocorrelation in the model.

Limitations

The main limitation is the use of state-level per capita spending on the social service variables that were not available at the MSA level. However, in many cases the MSAs examined account for the bulk of health care and social service spending in their states. This is linked to the fact that 92% of households are in urban areas.¹³ The study team did explore MSA-level fixed effects models and the direction of the results were very similar to those that will be presented in Results. However, given the lack of variation within some of the MSAs over time, the statistical significance in some of the models was lower.

Results

The regression results are displayed in Table 1. A summary of the results for each of the 6 categories of predictor categories is provided in this section. Because there are 8 separate regressions, a summary is provided of the independent variables of interest on the 8 outcome measures. Results are presented regarding the association of education spending, environmental quality (fine particulates), tobacco taxes, per capita health care spending, 3 social service measures, and physical activity.

Table 1.

Regression Results for Determinants of Health

Variable	Obese, BMI ≥30	Mean number of days in last 30 not good physical/mental health resulting in restricted activities	Mean number of days in last 30 not good mental health	Age-adjusted mortality (per 100,000 pop)	Age-adjusted cancer mortality (per 100,000 pop)	Asthma prevalence	Diabetes prevalence	CVD prevalence
Participate in physical activities (%)	−.007^*	−.008^*	−.0006	−.005^*	−.002^*	.005^*	−.13^*	−.01^*
Current smoker (%)	.0008	.011^*	.01^*	.008^*	.012^*	.009^*	.008^*	.018^*
State cigarette tax per 20-pack ($)	.004	−.02^*	.007	.0009	−.008^*	.05^*	−.03^*	−.032^*
Education: some college (%)	.003^**	−.006^*	−.005^*	.001	.0007	.004	.0005	−.005
Education: college graduate (%)	−.003^**	−.012^*	−.011^*	−.002^*	.003^*	−.00009	−.01^*	−.004
Non-Hispanic black (%)	.005^*	−.002^*	−.003^*	.002^*	.002^*	−.005^*	.005^*	−.004^*
Hispanic (%)	.0004	−.003^**	−.001^*	−.002^*	−.001^*	−.002^**	.004^*	−.003^*
Unemployment rate, seasonally adjusted	.004^*	.005^*	.01^*	.0001	.003^*	.004	.007^*	−.005
100%–199% federal poverty level (%)	−.001	−.002	−.00005	−.002^*	−.0005	−.003	−.001	−.003
200%-299% federal poverty level (%)	−.0003	−.001	−.00001	−.002^*	−.002^*	−.002	.003	−.0004
300%-399% federal poverty level (%)	−.002	−.006^*	−.003^**	−.003^*	5.44 e-06	−.005^*	.002	−.001
≥400% federal poverty level (%)	−.003^*	−.002	.006	−.003^*	−.002^*	−.001	−.001	−.004^*
Uninsured (%)	−.001	−.002	−.002	−.002^*	−.002^**	−.008^*	−.007^*	−.0009
Per capita real GDP	−6.5 e-07	1.7 e-07	−5.7 e-07	−5.1 e-06^*	1.04 e-06^*	−1.83 e-06^*	1.58 e-06^*	2.46 e-07
Per pupil total current spending	−.00001^*	−.00002	−2.0 e-06	2.02 e-07	−9.38 e-6^*	−8.32 e-06	−7.53 e-06^*	2.6 e-06
Fine particulate matter (PM2.5) weighted annual mean (μg/m3)	.01^*	−.005^**	.008^*	.006^*	.007^*	.002	.006^*	.002
Ozone – fourth highest daily max 8-hour concentration (ppm)	−1.7^*	−1.09	−.98	.494^*	.04	1.45^**	−3.33^*	−.80
Per capita personal consumption (health care), 1-year lag (logged)	.222^*	.062	−.16^*	−.16^*	.18^*	.014	.08	.21^*
Local govt exp on parks and rec per 1000 state pop, 1-year lag (logged)	.003	−.132^*	−.08^*	−.005	−.04^*	−.10 ^*	−.18^*	−.13^*
Local govt exp on police/fire per 1000 state pop, 1-year lag (logged)	−.076^*	.20^*	.17^*	.03^*	.08^*	.13^*	.23^*	.27^*
Local govt exp on public welfare per 1000 state pop, 1-year lag (logged)	−.006^*	−.01^*	−.01^*	−.01^*	−.004^**	−.004	−.009^*	−.0005

Significantly different from 0, P ≤ .05.

Significantly different from 0, P ≤ .10.

BMI, body mass index; CVD, cardiovascular disease; GDP, gross domestic product; govt, government; ppm, parts per million.

Health behaviors

Health behaviors were almost uniformly important in predicting health outcomes and the summarized results follow.

Increased percent participating in physical activities

Metropolitan areas with a higher percent of residents participating in physical activities were associated with better health outcomes for 7 of the 8 measures examined. For instance, each percentage point increase in the share of metropolitan area residents participating in physical activities was associated with a 0.7% reduction in obesity. Similarly a percentage point increase in physical activity share was associated with a ≤1% reduction in the average number of mean days of poor health per month and in the number of days in poor mental health. Finally, a percentage point increase in the share of residents participating in physical activity was associated with reductions in age-adjusted mortality, age-adjusted cancer mortality, and cardiovascular disease prevalence, and a large (>10%) reduction in diabetes prevalence (Table 1).

Education

Higher per pupil spending on education was associated with improved health outcomes in 3 of the 8 behavior outcome categories, though the impacts were small. For instance, each $100 increase in education spending per student was associated with approximately a 0.1% reduction in obesity prevalence. The same $100 increase in educational spending per student was associated with a 0.9% reduction in age-adjusted cancer mortality and a 0.8% reduction in diabetes prevalence.

Environment

Increased community levels of fine particulates also were associated with worse health outcomes in 5 of the 8 measures. Each 1-unit increase in the mean annual fine particulate concentration was associated with 1% increase in the percent obese. The same increase also was associated with a ≤1% increase in the mean number of monthly poor mental health days (Table 1). Higher levels of fine particulates also were associated with higher rates of age-adjusted mortality. At the same time, a 1-unit increase in mean fine particulate concentration was associated with an increase in cancer mortality and diabetes prevalence.

Results for the ozone levels in communities on health outcomes were mixed. Higher ozone levels were associated with lower rates of obesity and a reduction in the mean number of days in poor physical or mental health. Higher ozone levels also were associated with higher rates of age-adjusted mortality, age-adjusted cancer mortality, and asthma. However, higher ozone levels also were associated with a lower prevalence of diabetes.

Tobacco tax

Higher tobacco taxes also were associated with improved health outcomes in 5 of the 8 measures. A $.01 increase in the state tobacco tax is associated with a 2% reduction in the average number of poor health days per month, and a 0.8% reduction in age-adjusted cancer mortality. The same increase in tobacco taxes also was associated with a reduction in diabetes prevalence and cardiovascular disease prevalence (Table 1).

Social services spending

Higher levels of social services spending on parks and recreation and public welfare also were associated with improved health outcomes. However, in some cases, additional spending on police and fire resulted in improved health outcomes, though the results here were mixed.

Spending on parks and recreation

Providing additional funding for parks and recreation was associated with better health outcomes in 6 of the 8 measures. A 10% increase in per capita spending reduced the average number of restricted activity days related to poor health by 1.3% and the mean number of mental health days by ≤1%. Similarly, the same increase in parks and recreation spending reduced age-adjusted cancer mortality and the prevalence of diabetes, cardiovascular disease, and asthma (Table 1).

Spending on police and fire

Additional per capita spending on police and fire could positively affect residents' sense of safety and therefore improve health outcomes. On the other hand, it could reflect higher crime rates and with it higher levels of stress and anxiety, resulting in poorer health outcomes. Therefore, a priori it is not clear whether such spending would be positively or negatively associated with improved health outcomes. Additional spending on police and fire had mixed results. Each 10% increase in spending per thousand population was associated with a 7.6% reduction in obesity. On the other hand, the same increase was associated with an increase in mean days of poor health, mean days of poor mental health, and age-adjusted mortality. Higher levels (10%) of spending also were associated with increases in age-adjusted cancer mortality, diabetes prevalence, and cardiovascular disease prevalence (Table 1).

Public welfare spending

Additional spending on public welfare was associated with improved health outcomes in 6 of the 8 measures. Each 10% increase in public welfare spending per person was associated with reductions in obesity, the number of monthly poor health days, the number of poor mental health days, age-adjusted mortality, age-adjusted cancer mortality, and diabetes prevalence (Table 1).

Discussion

The analysis presented highlights the important association that environmental quality, social services, education spending, and physical activity assume in better health outcomes. Although the results do not show causation, these associations are important for community policy makers to keep in mind as they seek to improve health outcomes. The results also underscore the fact that health outcomes may be improved through better access to and investment in services outside the health care system. Indeed, although health care as a percent of gross domestic product in the United States is the highest in the world, the United States spends considerably less on social services compared to several European countries. For instance, spending on social services accounts for 9% of gross domestic product in the United States compared to more than 20% in France, Sweden, and Switzerland.¹⁴ This higher combined level of social and health care spending in several European countries may be a factor in their reported better health care outcomes.¹⁵

The results make an argument for additional community-based investments in these activities. Overall, the United States spends substantially more on health care than is invested in social services, controlling pollution, and implementing evidence-based behavior change programs. Subsequent research should drill down further to provide a more complete understanding of the behavior changes that accompany higher spending on social services.

As the health policy debate pivots toward improved efficiency and better health outcomes, these results suggest the need to broaden the focus on not just the health sector, but to include these important underlying determinants of health. Previous estimates have shown that up to 50% of potentially avoidable mortality is linked to health behaviors, environmental, and other social issues.¹⁶ Community-level approaches are an important approach for improving health outcomes.

The type of social service spending also matters. Additional spending on parks and recreation services had a notable impact on improving health outcomes. This, along with higher shares of residents with any exercise, could result in measurable improvements in health outcomes over time. Communities seeking to improve health outcomes have several programmatic changes they could make to social service spending and access to health care services to improve population health. Ideally, a comprehensive approach that integrates efforts to increase physical activity, improve access to care, and invest in social services, such as parks and recreation and better environmental controls, would be designed to achieve synergies among them.

Footnotes

Author Disclosure Statement

The authors declare that there are no conflicts of interest. The authors received the following financial support: The Aetna Foundation provided research support.

References

Bradley

, Canavan

, Rogan

, et al. Variation in health outcomes: the role of spending on social services, public health, and health care, 2000–09. Health Aff (Millwood), 2016; 35:760–768.

Marmot

, Rose

, Shipley

, Hamilton

PJS

. Employment grade and coronary heart disease in British civil servants. J Epidemiol Community Health, 1978; 32:244–249.

McCullough

, Leider

. Association between county wealth, health and social security spending and health outcomes. Am J Prev Med, 2017; 53:592–598.

National Center for Chronic Disease Prevention and Health Promotion—Division of Population Health. Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System (BRFSS) SMART: City and County Data. www.cdc.gov/brfss/smart/Smart_data.htm Accessed January 9, 2017 .

United States Census Bureau, Bureau of Labor Statistics. Current Population Survey, Annual Social and Economic (ASEC) Supplement. http://thedataweb.rm.census.gov/ftp/cps_ftp.html#cpsmarch. Accessed January 19, 2017 .

Centers for Disease Control and Prevention. CDC Wide-Ranging Online Data for Epidemiologic Research (WONDER). http://wonder.cdc.gov Accessed February 24, 2017 .

United States Department of Labor, Bureau of Labor Statistics. Local area unemployment statistics (LAUS). www.bls.gov/data/#unemployment Accessed June 30, 2016 .

United States Environmental Protection Agency. Air Quality Trends by City, 2000–2015. www.epa.gov/air-trends/air-quality-cities-and-counties Accessed August 4, 2016 .

United States Census Bureau. Federal, state, and local governments public elementary-secondary education finance data. www.census.gov/programs-surveys/school-finances.html Accessed March 9, 2017 .

10.

United States Census Bureau. State by level/type of government summary file. www2.census.gov/govs/local Accessed February 24, 2017 .

11.

United States Department of Commerce, Bureau of Economic Analysis. Gross domestic product (GDP) for metropolitan areas. Personal consumption expenditures (PCE) by state. https://bea.gov/regional/downloadzip.cfm Accessed June 30, 2016 .

12.

Tax Foundation Center for State Tax Policy. Cigarette and tobacco taxes. https://taxfoundation.org/state-tax/cigarette-and-tobacco-taxes Accessed February 24, 2017 .

13.

Hawk

. Expenditures of urban and rural households in 2011. 2013. www.bls.gov/opub/btn/volume-2/expenditures-of-urban-and-rural-households-in-2011.htm#_edn1 Accessed March 9, 2017 .

14.

Squires

, Anderson

. U.S. health care from a global perspective. New York: The Commonwealth Fund, 2015.

15.

Brookings Institution. Re-balancing Medical and Social Spending to Promote Health: Increasing State Flexibility to Improve Health Through Housing. www.brookings.edu/blog/up-front/2017/02/15/re-balancing-medical-and-social-spending-to-promote-health-increasing-state-flexibility-to-improve-health-through-housing Accessed March 22, 2017 .

16.

McGinnis

, Foege

. Actual causes of death in the United States. JAMA, 1993; 270:2207–2212.