Cardiovascular diseases (CVD) are among the leading causes of death in the United States.1 CVDs disproportionally affect the socially and economically disadvantaged populations—the underserved racial–ethnic minorities and low-income subpopulations. In fact, CVDs account for one in three deaths each year2,3 with Blacks and Hispanics, and low-income groups being more likely to die from CVDs than Whites, and higher-income groups, respectively.1,4 CVDs are costly to the nation and incur approximately $329 billion in direct and indirect costs each year.4 CVDs and CVD disparities could be reduced by improving the health of the socially and economically disadvantaged populations, as disparities lead to avoidable deaths and unnecessary excess costs and improving the health of disadvantaged populations has been shown to result in economic gains.5

The Affordable Care Act (ACA) was enacted in 2010 with the overarching goal of controlling the rising cost of healthcare, improving the quality and value of health services, and increasing health coverage among uninsured Americans.6 Medicaid expansion, a component of the ACA, was enacted in 2014 to extend Medicaid health coverage among individuals who have income below the 138% federal poverty level.7,8 Medicaid expansion is thought to have great potential for improving the health of the socially and economically disadvantaged populations and for reducing health disparities by increasing their access to affordable and high-quality healthcare that they would otherwise not be able to afford because of lack of insurance.9

In particular, Medicaid expansion resulted in a substantial decrease in the Black–White disparities in uninsured coverage.10 However, there was mixed evidence about the narrowing of the coverage gap between Hispanics and non-Hispanic Whites (Whites hereafter).10–13 Regardless, while the uninsured rates among non-Hispanic Blacks (Blacks thereafter) and Hispanics were reduced, their uninsured rates remained substantially higher compared with Whites.10,13–15 Therefore, given the heterogeneous effect of Medicaid expansion on uninsured rates between Blacks and Hispanics compared with Whites, that CVD is the leading cause of death,2,3 and that Blacks and Hispanics are disproportionately at risk,1,4 it is possible that Medicaid expansion could have a differential impact on CVD mortality among Blacks and Hispanics compared with Whites.

Furthermore, while there were reductions in uninsured rates in both men and women after the adoption of Medicaid expansion,16 some sex–gender disparities in outcomes have persisted. For instance, Medicaid expansion has resulted in an increase in the colorectal cancer screening rates among low-income nonelderly women in Medicaid expansion states, but not among men.17 Therefore, given this differential impact of Medicaid between men and women, and the persistent sex differences in risk factors such as the control of type 2 diabetes and high blood pressure in the past decade,18 it is also plausible that there could be sex–gender differences in the effect of Medicaid expansion on CVD mortality.

The effect of Medicaid expansion on CVD mortality can change over time. In fact, the CVD risk factor distribution has changed over time with significant increases in body mass index, hemoglobin A1c and decreases in total cholesterol and cigarette smoking over the past decade, suggesting that the risk for CVD incidence could vary with time. Additionally, Medicaid enrollment has changed over time since the implementation of Medicaid expansion, initially increasing substantially and then declining in 2018.19 Last, additional states have continued to expand Medicaid to low-income populations since the policy first went into effect in 2014. These include for instance: Pennsylvania (2015), Montana (2016), and Virginia (2019).20

Recent studies have found that Medicaid expansion was associated with a reduction in all-cause mortality among nonelderly adults and this finding was observed as early as in the first year of the policy.21,22 This reduction is thought to have been driven by mortality decreases in states with higher preexpansion uninsured rates and by causes of death (such as CVD risk factors) that are potentially preventable with access to adequate healthcare.21 Reduction in mortality associated with Medicaid expansion was also recently documented for CVD mortality using data through 2016.23 However, this recent finding did not consider how the reduction in mortality would vary by demographic groups over time. Such knowledge has important implications for improving and tailoring efforts among underserved subpopulations to achieve health equity. Furthermore, it is important to continuously monitor the impact of policies, especially as new data become available (now up to 2019), to detect potential variations over time and across subpopulations. Last, the prior study used the “difference-in-difference” (DID) quasi-experimental design,24–26 which assumed the difference in CVD mortality trends in expansion and nonexpansion states would remain constant on average in the postexpansion period in the absence of a policy effect. This parallel trend assumption is essential for causal identification in the DID methodology, and its violation could lead to inaccurate estimates. Fortunately, there have been recent methodologic developments such as the generalized synthetic control,27 which has been shown to be less prone to violations of the parallel trends assumption relative to the traditional synthetic control and the DID methods.28 Therefore, longer-term studies that adjust for nonparallel trends and that assess heterogeneity of policy effects by race and sex subpopulations over time are needed.

Here, we investigated the effect of Medicaid expansion on CVD mortality by race–ethnicity and by sex. To achieve this, we utilized the generalized synthetic control method (SCM) applied to more recent data up to 2019 to estimate (1) the average effect of treatment on the treated (ATT, in terms of mean differences) over all states and over all time points for Blacks, Hispanics, Whites, men, and women and (2) the ATT over all states and for each time point for Blacks, Hispanics, Whites, men, and women. In addition, to specifically test the presence of racial–ethnic and sex–gender disparities over time, we estimated the difference in effect between Blacks or Hispanics versus Whites and that between women versus men.

METHODS Study Population and Data Sources

Data on states that have adopted Medicaid expansion and when each adopted the expansion came from the Kaiser Family Foundation,7 which provides information on key health policy issues related to Medicaid, Medicare, health reform, and health insurance.7

Socio-demographic data came from the behavioral risk factor surveillance system (BRFSS),29 a large, annual, nationally representative telephone survey that collects data on all 50 states in the United States.29

We calculated state-level density of primary care clinicians and cardiologists from county-level data obtained from the Health Resources and Services Administration Area Health Resource File.30 Area Health Resource is an annually maintained repository of data from over 50 data sources that includes information on healthcare professions, health facilities, and hospital utilization.30

State-level CVD mortality rates were obtained from the Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research (CDC WONDER) database,31 a large up-to-date online database resource that manages almost 20 collections of public-use data for US births, deaths, cancer diagnoses, population estimates amongst others.31 The unit of analysis was state–year.

CVD outcome data were partially missing in the Black and Hispanic subpopulations. For confidentiality reasons, in CDC WONDER, statistics representing fewer than ten persons were suppressed and when the death count was less than 20, the corresponding rates were marked as “unreliable”32; we treated these as missing. States with limited outcome data (i.e., missing 3 to 20 years) were excluded from the analysis for the Black and Hispanic subpopulation, otherwise, the states outcome data were imputed. (See eSection 1; https://links.lww.com/EDE/C106, eFigure 1; https://links.lww.com/EDE/C106, eFigure 2; https://links.lww.com/EDE/C106 for details on the missingness pattern).

Variables Outcome

The outcome was CVD deaths that is, deaths due to diseases of the circulatory system (International Classification of Diseases, 10th Revision codes I00 to I99) per 100,000 persons per year among individuals aged 45–64 years, as done elsewhere.23 Examples of CVD included but were not limited to hypertensive diseases, ischemic heart diseases, pulmonary heart diseases, and diseases of pulmonary circulation and cerebrovascular diseases (e.g., stroke). The CDC WONDER database provided mortality rates that were age-adjusted to the 2000 US population and available from 2000 to 2019. Similarly to Khatana et al.,23 we focused on ages 45 to 64 as CVD outcomes are more prevalent among older adults and Medicaid coverage is more common among adults less than 65 years (since they would not be eligible for Medicare).

Potential Effect Modifier

Potential effect modifiers included race (Black, Hispanic, and White) and sex (Male vs. Female).

Exposure

The exposure is whether or not a state adopted Medicaid expansion. As of 2019, 34 states had adopted Medicaid expansion while 16 states had not (See Figure 1). States implemented the policy at different times (e.g., 2014 for California, 2015 for Pennsylvania, and 2016 for Louisiana, see Figure 1). Data spanned the period from 2000 to 2019.

FIGURE 1.:

Analytical data structure of the Medicaid expansion states and control states for the overall study population.

Potential Confounders

We included the following covariates in our analysis: state–year proportion of residents aged 45–64 years who are non-White, males, married, without high school degree, with annual income less than $15,000, who are employed for wages, state–year population size of residents aged 45–64 years, states’ political orientation, state–year density of primary care clinicians per 100,000 residents, state–year density of cardiologists per 100,000 residents. Data for density for primary care clinicians were only available from 2010 to 2016. We therefore imputed the 2000–2009 data using the 2010 density and the 2017–2019 data using the 2016 density data. A similar approach was done for the density of cardiologists which were only available for 2010, 2015, and 2017. Data on other covariates were available from 2000 to 2019. Covariate data for New Jersey were missing in 2019 as the state was unable to collect enough BRFSS data to meet the minimum requirements for inclusion in the 2019 annual aggregate data set.33 We therefore imputed the 2019 covariate data for New Jersey using New Jersey’s 2018 covariate data.

Statistical Analysis Overview

We used a quasi-experimental design method—the generalized SCM,27 an extension to the DID26 and the SCM.34 The generalized SCM has some advantages over the DID and SCM. First, it accommodates multiple units with variable treatment timing (also called staggered adoption).27 This is relevant as Medicaid expansion was adopted by several states at different times. Second, unlike the SCM, which relies on falsification tests for inference, the generalized SCM allows for the estimation of conventional confidence interval (CI) via bootstrapping.27 Third, it has recently been shown28 that, relative to the traditional SCM and the DID methods, the generalized SCM is less prone to violations of the parallel trends assumption.

The Generalized Synthetic Control Model

The generalized SCM assumes that an outcome Yit for a state i at a time t (i.e., annual and state-specific age-adjusted CVD mortality rates) can be modeled using a linear factor model27 as follows:

Yit=δitDit+xit′β+λi′ft+εit

where Dit is an indicator for exposure to Medicaid expansion in the postpolicy period and equals 1 if a state i has been exposed to Medicaid expansion in the postpolicy period and 0 otherwise. δit is the heterogeneous treatment effect for state i at a time t,xit, a vector of observed covariates, β, a vector of unknown parameters, ft, a vector of unobserved common factors, λi, a vector of unknown factor loadings and εit are idiosyncratic errors for state i at a time t and has zero means. The average treatment effect (ATT) over the postpolicy period can then be calculated as the average of the heterogeneous treatment effect as 1Ntr∑i∈itrδit where Ntr is the number of treated states and itr the treated states. More details on implementing the generalized SCM are described in the supplemental materials (eSection 2; https://links.lww.com/EDE/C106). We assumed the effect of Medicaid expansion would start during the year of the adoption as done elsewhere,23,35 and estimated the yearly effect of the policy from the time since adoption.

Heterogeneity Analysis

We estimated the effect of Medicaid expansion on CVD deaths overall and for Blacks, Hispanics, Whites, men, and women.

In addition, we estimated the annual effect since the time of the adoption of Medicaid expansion. To do so, we used a single generalized SCM model for all states since this model can accommodate multiple states with variable treatment timing. However, because the missingness pattern and sample size was different for each race subgroup, we estimated the effect of Medicaid expansion on CVD mortality in each subgroup (e.g., Black subgroup) by running separate generalized SCM models.

Last, to assess the presence of racial–ethnic and sex disparities, we estimated the difference in effect between Blacks or Hispanics versus Whites and that between women versus men. The difference in the effect between subgroups is akin to a triple DID and is a test of the presence of heterogeneity in the effect of Medicaid expansion on CVD across subgroups. (See eSection 3; https://links.lww.com/EDE/C106 for further details).

Sensitivity Analysis

To ensure the robustness of our findings as it relates to the potential influence of the missingness, we conducted a sensitivity analysis by running an analysis of the White subgroup restricted to the nonmissing states used in the Black subgroup and in the Hispanic subgroup and using these estimates for estimating the triple difference.

We implemented data analyses in the R software (version 4.2.1) with the aid of additional analysis packages such as GSYNTH27 for implementing the generalized SCM and the SRVYR package36 for estimating the state-level weighted proportions. All analysis codes can be found at the following link: https://github.com/nianogo/Medicaid-CVD-Disparities.

This study was reported in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology recommendations (See eTable 1; https://links.lww.com/EDE/C106).

This study used publicly available deidentified aggregated data and as such did not qualify as human subject research; therefore, institutional review board approval was not required.

RESULTS Baseline Characteristics

In Medicaid expansion states, there were on average 177 (standard deviation [SD] = 53) CVD deaths per 100,000 persons per year and in the nonexpansion states, there were 154 (SD = 47). The number of cardiologists per 100,000 persons was five (SD = 2) in expansion states and six (2) in nonexpansion states. About 11% of residents aged 45–64 years old made less than $15,000 annually in expansion states compared with 10% in nonexpansion states. Additionally, 62% of residents aged 45–64 years were employed for wages in expansion states compared with 64% in nonexpansion states (Table 1).

TABLE 1. - Characteristics of Medicaid Expansion States and Control States, 2014 State Medicaid Expansion Status Characteristics Overall, N = 50 Medicaid Nonexpansion States, N = 16 Medicaid Expansion States, N = 34 CVD deaths per 100,000 persons aged 45–64 years, mean (SD) 154 (47) 177 (52) 143 (41) Primary care clinicians per 100,000 residents, mean (SD) 77 (13) 68 (7) 81 (13) Cardiologists per 100,000 residents, mean (SD) 6 (2) 5 (2) 6 (2) Percentage of residents aged 45–64 years with annual income less than $15,000, mean (SD) 10 (3) 11 (3) 9 (3) Percentage of residents aged 45–64 years who are males, mean (SD) 42 (3) 42 (3) 43 (3) Percentage of residents aged 45–64 years who are non-White, mean (SD) 21 (13) 23 (12) 21 (13) Percentage of residents aged 45–64 years who are married, mean (SD) 61 (4) 62 (6) 61 (4) Percentage of residents aged 45–64 years without high school degree, mean (SD) 7 (3) 8 (3) 6 (3) Percentage of residents aged 45–64 years who are employed for wages, mean (SD) 64 (7) 62 (8) 65 (6) Percentage of political party affiliation, n (%)  Republican 24 (48%) 15 (94%) 9 (26%)  Democrat 13 (26%) 0 (0%) 13 (38%)  Split 13 (26%) 1 (6%) 12 (35%)

SD indicates standard deviation.

Overall Effect of Medicaid on Cardiovascular Disease Mortality

Overall, Medicaid expansion was associated with a reduction in CVD deaths, although the estimate was somewhat imprecise (Mean difference [MD] −4.29 per 100,000 persons per year, 95% CI = −9.87, 1.29) (Table 2). The estimated overall annual effect was fairly constant and ranging from almost no effect at the time of adoption (MD = 0.40; 95% CI = −3.38, 4.17) to a larger effect in the first year (MD = −5.09; 95% CI = −9.34, −0.85) followed by a reduction of 1.55 CVD deaths per 100,000 persons per year in the 6-year year postimplementation, although the estimate was also somewhat imprecise (95% CI = −9.88, 6.77) (Figure 2, eTable 2; https://links.lww.com/EDE/C106).

TABLE 2. - Overall and Subgroup Effect of the Medicaid Expansion on CVD Mortality Group Adjusted Mean Difference (95% CI) Difference in Mean Difference (95% CI) Overall −4.29 (−9.87, 1.29) −  White −3.18 (−8.30, 1.94) Reference 1  Black −5.36 (−22.63, 11.91) −2.18 (−20.20, 15.83)a Hispanic −4.28 (−30.08, 21.52) −1.10 (−27.40, 25.20)a  Men −5.96 (−15.42, 3.50) Reference 2  Women −3.34 (−8.05, 1.37) 2.62 (−7.95, 13.19)b

aUsing reference 1.

bUsing reference 2.

CI indicates confidence interval.

FIGURE 2.:

Overall annual effect of the Medicaid expansion on CVD deaths per 100,000 persons for overall population and for the Black, Hispanic, White, men, and women subpopulations.

Effect of Medicaid Expansion on Cardiovascular Disease Mortality Among Blacks, Hispanics, and Whites

Overall, among Blacks, Medicaid expansion was associated with a reduction of 5.36 CVD deaths per 100,000 persons per year (95% CI = −22.63, 11.91), although this association estimate was imprecise. Similarly, for Hispanics, Medicaid expansion was associated with a reduction of 4.28 CVD deaths per 100,000 persons per year; however here too, this association was imprecise (MD, 95% CI = −30.08, 21.52). Among Whites, Medicaid expansion was associated with a reduction of 3.18 CVD deaths (95% CI = −8.30, 1.94) per 100,000 persons per year, again imprecise (Table 2).

Among Blacks and Hispanics, the trends in the effect of Medicaid expansion on CVD mortality appeared similar to those in the overall analysis (Figure 2, eTable 2; https://links.lww.com/EDE/C106).

The difference in mean difference between the estimated effect of Medicaid expansion in Blacks compared with Whites was −2.18 (95% CI = −20.20, 15.83) and between that in Hispanics compared with Whites it was −1.10; (95% CI = −27.40, 25.20) (Table 2, eFigure 3; https://links.lww.com/EDE/C106).

Effect of Medicaid Expansion on Cardiovascular Disease Mortality Among Women and Men

Overall, among women, Medicaid expansion was associated with a reduction of 3.34 CVD deaths per 100,000 persons per year, although the estimate was a little imprecise (95% CI = −8.05, 1.37). Similarly, for men, Medicaid expansion was associated with a reduction of 5.96 CVD deaths per 100,000 persons per year; however, here too, the association estimate was somewhat imprecise (95% CI = −30.08, 21.52) (Table 2).

Among women and men, the trends in the effect of Medicaid expansion on CVD mortality appeared similar to those in the overall analysis (Figure 2, eTable 2; https://links.lww.com/EDE/C106).

The difference in mean difference between the effect of Medicaid expansion in women compared with men was 2.62 (95% CI = −7.95, 13.19) (Table 2, eFigure 3; https://links.lww.com/EDE/C106).

DISCUSSION

The purpose of this study was to investigate the effect of the Medicaid expansion on CVD mortality by race–ethnicity and by sex–gender and to evaluate whether there existed disparities across these subpopulations over time. Our findings suggest that Medicaid expansion was associated with a reduction in CVD mortality overall and in Whites, Blacks, Hispanics, men, and women in the states that have adopted Medicaid expansion. The estimates were close in magnitude, although much more imprecise for Blacks and Hispanics—potentially due to the smaller sample sizes in these subpopulations. In addition, our study did not find any difference or disparity in the effect of Medicaid on CVD across race–ethnicity and sex–gender subpopulations—likely owing to imprecise estimates. Last, while Medicaid expansion was found to be associated with an overall beneficial effect (although imprecise), the effect appears to be more apparent in the early phase of the postpolicy (i.e., years 1 and 2), but appears to be diminishing in the later phases (i.e., years 5 and 6) of the policy in expansion states, potentially suggestive of a waning or plateauing effect.

Our study did not find any difference or disparity in the effect of Medicaid on CVD across race–ethnicity and sex–gender subpopulations. This, however, did not mean that there were no disparities, but rather that the study was unable to detect heterogeneity—likely owing to imprecise estimates. Nevertheless, persistent disparities have been documented in Medicaid expansion states.37 In fact, results of other studies have suggested that Medicaid expansion increased screening rates for certain cancers among Blacks, but not Hispanics,38 but did not change cancer mortality among Blacks.39 While some of the established disparities in CVDs have been attributed to disparities in treatment and healthcare access,40 our study could suggest that Medicaid expansion may be contributing to closing the disparities in CVD outcomes by improving access to care such as through life-saving cardiovascular interventions including aspirin,41 better diabetes control,41 and coronary artery bypass graft surgery.42

Our findings that Medicaid expansion was generally associated with a reduction in CVD are consistent with a prior study that showed an overall beneficial effect of Medicaid expansion on CVD mortality,23 which could likely be the result of reported increases in the number of visits to primary physicians compared with nonexpansion states.43–45 In fact, prior studies have shown that Medicaid expansion has been associated with an increase in insurance coverage, primary care visits,43–45 a reduction in diabetes hospitalization,46 cancer mortality,47 and end-stage renal disease48 to name a few. Nevertheless, our study differs from the prior study as it estimated the effect of Medicaid expansion by race and sex over time.

The early reduction in CVD mortality associated with Medicaid expansion could be explained by several factors. First, a high number of Americans, especially Blacks, Hispanics, and low-income populations have a cluster of cardiometabolic risk factors (e.g., dyslipidemias, excess adiposity, high blood pressure or impaired fasting glucose) and are considered metabolically unhealthy,49–51 potentially putting a high number of them at increased risk of a serious CVD event. Second, it has been shown that lack of health insurance is associated with fewer primary care visits and higher emergency department visits.52 This suggests that uninsured individuals would seek care very late in the disease progression and sometimes even delay seeking emergency care for serious fatal conditions such as acute myocardial infarctions.52 Third, Medicaid expansion has been associated with increased insurance coverage, primary care visits and fewer emergency department visits, and not having to forgo a physician’s visit as early as in the initial years after the adoption of the policy.45,53,54 This has potentially led to catching CVD early in the disease progression and preventing more serious events such as ischemic heart disease or stroke. For instance, Medicaid expansion has been associated with increased access to life-saving cardiovascular interventions such as aspirin,41 better diabetes control,41 and coronary artery bypass graft surgery.42

There are three major limitations. First, our study assumes that the effect of external policies or events on CVD mortality at the time of or in the period after the adoption of the Medicaid expansion is on average similar in both the treated and control groups (this is known as the common shock assumption).25 As noted by other studies evaluating the effect of the Medicaid expansion program and the ACA provisions,44,55 the main contemporaneous policies or events at the state level that could have occurred at the time of the Medicaid expansion adoption or thereafter and which could have affected CVD mortality were generally those related to the ACA itself. In fact, around 2014, the ACA had two major relevant provisions,55,56 including expanding the Medicaid program to cover low-income adults (i.e., Medicaid expansion program) and providing individuals or families above the federal poverty level who qualify with subsidies that lower costs for households (i.e., subsidization of the marketplace for private insurance). To adjust for the potential effect of the subsidization provision on the state’s CVD outcome, we adjusted for income and employment status, factors that are related to eligibility for this provision. Additionally, given that expansion states experienced greater difficulty in accessing physician care compared with nonexpansion states,43 we adjusted for density of primary care physicians and cardiologists, especially as there appeared to be more clinicians per 100,000 residents in expansion states. Last, to account for the other determinants of Medicaid expansion and which could be associated with the CVD outcome, we further adjusted for the state’s political orientation. All these adjustments would reduce the effect of external policies and mitigate the potential for violation of the common shock assumption.44 Second, some of our stratified estimates lacked precision, possibly due to the smaller sample sizes of the subpopulations. Relatedly, the pretreatment fit of the generalized SCM model obtained by comparing the observed outcomes to the predicted counterfactuals (synthetic controls) before treatment while near perfect for the overall and for men, women and White subpopulations, was imperfect for the Black and Hispanic subpopulations (Figure 3). This was likely due to the small sizes for these subpopulations related in part to the missingness in the CVD mortality data imposed by CDC WONDER.32 Moreover, because the missingness pattern and sample size for each subgroup were different, we estimated the effect of Medicaid expansion on CVD mortality in each subgroup by running separate generalized SCM models. We then conducted a sensitivity analysis by running an analysis of the White subgroup restricted to the nonmissing states used in the Black subgroup and in the Hispanic subgroup and using these estimates for estimating the triple difference. This yielded overall similar conclusions (eTable 3; https://links.lww.com/EDE/C106) suggesting missingness had a negligible effect on our findings. Third, our study may not be generalizable to populations younger than 45 or older than 64 years or other racial–ethnic groups.

FIGURE 3.:

Pretreatment fit showing the observed CVD deaths per 100,000 persons along with the counterfactual (synthetic control). Vertical line represents the beginning of expansion. The solid line is expansion state, dashed line is the synthetic control.

CONCLUSION

Our findings suggest that Medicaid expansion was associated with an overall reduction in CVD mortality overall and in Whites, Blacks, Hispanics, men, and women subpopulations, although the effects appeared to be less apparent in the later phases of the policy. In addition, our study did not find any difference or disparity in the effect of Medicaid on CVD across race–ethnicity and sex–gender subpopulations—likely owing to imprecise estimates. Future studies should continuously monitor the effect of Medicaid expansion to see whether the effect is sustained over time and across subpopulations.

REFERENCES 1. Arias E. United States life tables, 2010. Natl Vital Stat Rep. 2014;63:1–63. 2. National Center for Health Statistics. Health, United States, 2016: With Chartbook on Long-term Trends in Health; 2017. 3. Centers for Disease Control and Prevention. Heart Disease and Stroke Prevention: Addressing the Nation’s Leading Killers—At A Glance 2011; 2016. 4. Benjamin EJ, Virani SS, Callaway CW, Chang AR, Cheng S, Chiuve SE, et al. Heart Disease and Stroke Statistics—2018 Update: A Report From the American Heart Association; 2018. 5. Schoeni RF, Dow WH, Pamuk ER. The economic value of improving the health of disadvantaged Americans. Am J Prev Med. 2011;40:S67–S72.