Varied Health Spending Growth Across US States Was Associated With Incomes, Price Levels, And Medicaid Expansion, 2000–19

Health care spending in the US, which has the highest health spending per person in the world, continues to increase rapidly. After inflation is accounted for, national health spending per person in the US doubled between 2000 and 2020 and increased as a share of the economy from 13.3 percent to 19.7 percent. In 2020 total health spending reached $4.1 trillion, or about $12,530 per person.^1,2 Health spending is expected to continue to grow, with total spending estimated to reach $6.8 trillion by 2030—and more by 2050.^3,4

In addition to growing quickly over time, health spending per person varies substantially across the US. In 2014 (the last year with comprehensive estimates), spending per person ranged from $5,982 in Utah to $11,064 in Alaska.⁵ From 2005 to 2014, annualized growth rates ranged from 2.8 percent in Tennessee to 5.3 percent in North Dakota (before adjustment for inflation).⁶ These disparate trajectories are the consequence of state-specific factors that affect the demand for care, including demographics and the underlying health of the population, as well as differences in how health services are provided, financed, and regulated in each state. Factors such as personal income, the share of the population enrolled in Medicare or Medicaid, and health care provider supply have also been linked to differences in health care spending across the US.⁵

In recent years policy changes have occurred at both the federal and state levels that have affected states differently. Many of the changes mandated by the Affordable Care Act (ACA) in 2010 were phased in over time, with components of the law, including Medicaid expansion, taking effect in 2014 or later.^7,8 Variation in the expansion of Medicaid has affected health spending substantially, with smaller increases in the size of the insured population and the federal transfer payments in states that declined to expand Medicaid.⁹ Similarly, state-specific characteristics such as the lifting of individual mandate penalties led to other differences in how the ACA affected consumers’ health care behavior.¹⁰

Beyond the ACA, other changes in health policy are likely to affect health care spending differently across states. Variation in compensation rates, differing approaches to dental health coverage, and varying state legislative approaches to matters of immigrant health care are examples of changes that likely affect health spending in some, but not all, states.^11–15

Although the Centers for Medicare and Medicaid Services (CMS) provides an official accounting of health spending at the national level through 2020 in the annually updated National Health Expenditure Accounts, the State Health Expenditure Accounts extend to only 2014, which is not current enough to enable analysis of the effects of Medicaid expansion and other recent health policy changes.⁶ The Bureau of Economic Analysis produces state-level health expenditure reports through 2020, but they are incompatible with the more-often-relied-upon National Health Expenditure Accounts and State Health Expenditure Accounts and are not disaggregated by payer or type of care.¹⁶ Our work seeks to fill this gap by generating consistent state-specific spending estimates for the period 2015–19 by payer and type of service, and it sheds light on changes in health spending during a period with substantial changes in health policy at the state and federal levels.

Study Data And Methods

State Spending Forecasts

This analysis estimated total health spending and spending by payer and type of service for each US state (and Washington, D.C., hereafter referred to as a state) for the period 2015–19, using, when available, the primary data used in the State Health Expenditure Accounts to construct spending estimates for 1991–2014. Because not all primary data used in the construction of these accounts were available to our team, we used regression methods to predict spending during 2015–19, with the available primary data as our models’ covariates. These data sources (described more fully in section 1.1 of the online appendix)¹⁷ include Medicare fee-for-service claims spending reports and Medicaid total spending reported to CMS, total hospital spending reported by the American Hospital Association, and data from the Census Bureau, among others. Instead of choosing a model specification and set of covariates a priori, we tested more than 380,000 unique iterations of the model. These models included all combinations of available input data, a wide range of model specifications, and distinct ways of aggregating the estimates to ensure that they match the reported 2015–19 national estimates. The “best” model was identified by minimizing out-of-sample root-mean-square error, which ensured that model selection was robust to overfitting. Model uncertainty was estimated using bootstrapping methods to fit models with 1,000 random samples of states. The advantage to this method is that it is relatively simple, is based on linear regressions, and builds primarily from the same building blocks used to construct the State Health Expenditure Accounts. Before modeling spending, we adjusted all estimates for economy-wide national inflation rates to 2020 US dollars, using deflators from the Bureau of Labor Statistics, to ensure comparability across time.¹⁸

Similar to the State Health Expenditure Accounts, this analysis estimated total state spending as well as payer-specific and type of service–specific spending. The State Health Expenditure Accounts report spending by Medicare (including Medicare Advantage), Medicaid, and private insurers. This analysis used the same payer categories, along with a remainder category that accounted for out-of-pocket spending combined with other public insurers, such as Veterans Affairs and the Indian Health Service. For type of service, this analysis matched the hospital, physician/clinic, nursing facility, home health, dental, pharmaceutical, and other professional spending categories defined and reported in the State Health Expenditure Accounts. The remainder of total spending from these types of services, including durable medical equipment spending, is categorized as “other” spending in this analysis.

Spending Variation Analyses

After total, payer-specific, and type of care–specific health spending were estimated for 2015–19, we completed several analyses to validate the estimates and evaluate spending trends.

First, to validate our spending estimates, we compared them against health services supply and utilization measures, which were observed and not used for our projections. Our hypothesis is that the supply and use of health services should be directly correlated with health spending. These results are in appendix table e6.¹⁷

Second, to assess variation in spending across time and state, we used age standardization, price adjustment, linear regression, and decomposition methods to assess which factors explained the largest amount of spending variation. Age standardization was completed on total and payer-specific spending estimates, using indirect methods and data from the Institute for Health Metrics and Evaluation’s Disease Expenditure Project, and National Health Expenditure Accounts; regional economywide price parity indices were extracted from Bureau of Economic Analysis data and combined with year-to-year inflation factors from the Bureau of Labor Statistics.^2,18–20 Linear regression controlled for time, household income, population density, age-standardized smoking prevalence, and age-standardized physical activity. Shapley decomposition was used to assess how much variation could be attributed to each factor (see appendix figure e5).¹⁷ To illustrate the role that these factors played in explaining disparate spending levels, we estimated per person health spending, setting these factors to the national level for all states. The remaining variation in spending levels reflects variation that cannot be explained by these key factors and highlights differences in the way health care is provided and used.

Third, to assess the impacts of expanding Medicaid eligibility on total health spending and out-of-pocket spending, we used linear regression methods to test whether extending Medicaid coverage was associated with health spending changes. Medicaid expansion and changes to income eligibility thresholds for children, parents, and other adults, as well as the proportion of Medicaid enrollees on managed care, were tested separately to identify which parts of Medicaid expansion were most associated with changes in total spending.

Additional detail on the methodology for all steps of the analysis and equations for the spending forecasts and policy analysis regression are in supplemental methods in the appendix.¹⁶

Limitations

Although these methods are a promising means of estimating state spending on health in the US for 2015–19, there remain some important limitations. First, this analysis uses many, but not all, of the data that were used in the CMS State Health Expenditure Accounts. In particular, data tracking spending on Medicare supplemental insurance (Medigap) and data tracking Medicare Advantage spending more precisely were not readily available, but inclusion of these sources would likely improve fit for the payer models. Similar to the State Health Expenditure Accounts, this analysis missed some categories of spending that were relevant from a policy perspective, such as federal subsidies to states for Medicaid spending, administrative costs associated with running insurance companies, and public health activities. This analysis also assumed (similar to the State Health Expenditure Accounts) that there was no uncertainty coming from the input data. Instead, our estimates of uncertainty reflected uncertainty from the modeling framework. Importantly, modeled estimates should never replace official estimates of spending. The estimates produced in this research depend on prior official accounts from the State Health Expenditure Accounts for model training and selection and can be used alongside other official accounting data for 2015–19 when State Health Expenditure Accounts data are unavailable. Furthermore, although estimates of state health spending for more recent years would provide useful insight into the effects of the COVID-19 pandemic on health spending, pandemic estimates of spending were out of scope of this analysis, as we believed that the expected shift in spending would not be adequately captured by our models.

Some additional limitations relate to the regression analysis of spending estimates. First, because the regression models on supply, utilization, and Medicaid policy variables were modeled independently to avoid multicollinearity and reduced sample size, assessing how these associations are related to each other was not possible. Second, this project investigated only conditional associations, not causal relationships. Further research into the causality of spending growth would be a valuable extension of this analysis.

Study Results

For this research, 368,640 models predicting total health spending for each state were considered. The best model for total spending estimated year-over-year changes and used primary data from Medicaid CMS-64 administrative reports, Medicare spending estimates, the American Hospital Association, the Bureau of Economic Analysis, and the Medical Expenditure Panel Survey (MEPS). Out-of-sample validation showed that for five years of prediction, mean absolute error of health spending per person (which reflects the average amount that our modeled estimates deviated from observed estimates when applied to historic data) was $73 (see appendix table e4).¹⁷ Key covariates in the best payer-specific models included Medicare fee-for-service spending and data from the MEPS–Insurance Component reporting on premiums. For Medicare, Medicaid, private insurance, and out-of-pocket spending, the mean absolute errors were $53, $70, $126, and $135, respectively. For estimating the fraction of spending for each type of service, main covariates included data from the Economic Census, the Census of Annual Retail Trade, and the American Hospital Association. The mean absolute error for five-year out-of-sample predictions was less than or equal to $72 for each type of service (see appendix table e4).¹⁷ Validation regressions showed that there was a significant (all $p$ values <0.002) relationship between health care spending and supply and use of hospitals and hospital beds, admission rate per capita, and inpatient days per capita (see appendix table e6).¹⁷ These data were collected independently and highlight the expected relationships between health spending and the health system infrastructure, workforce, and use of services.

In 2019 the state with the lowest modeled spending per person spent half as much per person as the state with the highest spending. Moreover, variation in spending across states was much higher in recent years than in the early 2000s (see appendix figure e10).¹⁷ Estimates show that spending per person ranged from $7,250 (95% uncertainty interval: 7,190–7,320) in Utah to $14,500 (95% UI: $14,300–$14,710) in Alaska (see exhibit 1; 95% uncertainty intervals not shown). Annualized growth rates in total spending per person from 2013 to 2019 ranged from 1.0 percent (95% UI: 0.8–1.2) in Washington, D.C., to 4.2 percent (95% UI: 4.0–4.4) in South Dakota after adjusting for economywide inflation but without controlling for changes in age patterns, prices, or other factors discussed below (see appendix table e9d).¹⁷ Appendix table e9a contains a table of all projected state estimates of total spending, and appendix figure e10 displays change over time of a measure of variation in spending across states.¹⁷

In 2019, across the states, Medicare spending ranged from 9 percent (95% UI: 6–12) of total health spending in Alaska to 30 percent (95% UI: 26–35) in Florida (see exhibit 2). Medicaid spending ranged from 10 percent (95% UI: 7–15) of total health spending in South Dakota to 26 percent (95% UI: 22–30) in Washington, D.C. Private insurance spending ranged from 25 percent (95% UI: 12–40) of total health spending in West Virginia to 49 percent (95% UI: 42–56) in Washington, D.C. Finally, out-of-pocket spending, along with spending on other public insurance schemes, made up the remainder of spending, with fractions of total spending ranging from 12 percent (95% UI: 6–21) in Washington, D.C., to 42 percent (95% UI: 29–55) in Alaska. In all states, hospital care and physician and clinical services were the largest fractions of spending (per person). Hospital spending ranged from 35 percent (95% UI: 32–38) of total health spending in New Jersey to 51 percent (95% UI: 49–53) in South Dakota, and physician and clinical services spending ranged from 16 percent (95% UI: 16–17) of total health spending in Vermont to 30 percent (95% UI: 27–32) in Alaska (see appendix tables e9b and e9c for a complete set of estimates by payer and type of service by year).¹⁷

Six non–health system factors were tested as potential factors explaining variation in health spending per person, both in the State Health Expenditure Accounts estimates for 2000–14 and in the estimated spending for 2015–19 (see appendix figure e5).¹⁷ Income explained the largest fraction of variation in health spending per person, at 25.3 percent (95% UI: 25.2–25.4). Moreover, a 10 percent increase in mean state income was associated with a 4.8 percent (95% UI: 4.3–5.9) increase in health spending (see appendix table e6).¹⁷ Regional price parity also explained a great deal of spending variation, at 21.7 percent (95% UI: 21.1–22.3). Two main behavioral risk factors were tested (physical activity levels and smoking rates, lagged fifteen years), collectively explaining 8.0 percent (95% UI: 7.7–8.2) of variation in health spending, whereas 12.8 percent (95% UI: 11.8–13.6) of the variation was explained by a time trend, reflecting growth year over year that occurred consistently across all states. Different age and sex profiles explained 4.6 percent (95% UI: 4.4–4.8) of the variation in health spending, with population density explaining 4.8 percent (95% UI: 4.7–5.0). Collectively, these controls explained 77.2 percent (95% UI: 76.7–77.6) of the variation in state health spending per person during the period 2000–19, with 22.8 percent (95% UI: 22.4–23.3) left unexplained (appendix figure e5).¹⁷Exhibit 1 shows that when states are standardized across these factors, the spread of per person spending across states is smaller, although still sizable, ranging from $8,950 (Tennessee) to $14,820 (Alaska).

The expansion of Medicaid access through income eligibility changes and Medicaid expansion under the ACA had a complex relationship with state spending growth. Exhibit 3 shows that after state spending is standardized by age, sex, price, and other key factors, the annualized growth rates for total health spending per person from 2013 to 2019 (labeled “Aggregate” in exhibit 3) were similar for early Medicaid expansion and nonexpansion states, with the median for the states that expanded before 2016 being slightly lower. In contrast, median Medicaid-specific standardized spending per person growth from 2013 to 2019 was much larger for early Medicaid expansion states (4.2 percent) than for nonexpansion states (2.6 percent). For the same period, standardized Medicare spending growth was comparable for early expansion and nonexpansion states, but standardized private insurance spending growth and out-of-pocket and other insurance spending growth were higher for nonexpansion states, at 3.2 percent and 0.9 percent versus 2.7 percent and 0.1 percent, respectively. Medicaid expansion timelines and categorization for exhibit 3 are described in appendix table e5.¹⁷

SOURCE Authors’ estimates of standardized health spending and the Centers for Medicare and Medicaid Services’ State Health Expenditure Accounts. Authors’ estimates generated from multiple sources, as listed in appendix table e1 (see note 17 in text). NOTES This exhibit shows the annualized rate of change for total standardized health spending per person and payer-specific standardized health spending per person. The box-and-whisker plots show the interquartile range (the boundaries of the box), the median (the line within the box), the range of the data (the whiskers), and outlier points (the dots).

Exhibit 4 shows the factors tested for associations with total and out-of-pocket spending, where the reported coefficient represents the percentage increase in spending associated with a 10 percent increase in the percentage of Medicaid beneficiaries in managed care or when Medicaid raised or lowered income eligibility thresholds in a given state year. Medicaid expansion was statistically associated with a 1 percent (95% UI: 1–3) increase in estimated total health spending. In addition, Medicaid income eligibility thresholds for adults were statistically associated with increases in estimated health spending per person, whereas those for children were associated with less spending. The Medicaid income eligibility threshold for pregnant women was associated with less out-of-pocket spending.

Discussion

The state with the highest estimated spending per person in 2019 had nearly double that of the lowest state.

This study showed that health care spending estimates for the period 1991–2019 varied dramatically across US states and that although spending predictions in all states increased consistently during this period, differences between states increased across time. The ranking of the highest- and lowest-spending states also changed in the same period. Some of this variation can be explained by different demographic patterns between states over time, and age-standardized estimates can account for demographic shifts and allow for a study of potential drivers of the remaining variation. However, even after we performed standardization against main controls, such as income and consumer prices, the state with the highest estimated spending per person in 2019 had nearly double the standardized spending of the lowest state (exhibit 1).

Income explained the largest fraction of variation in health spending per person, at 25.3 percent.

Many existing studies have connected income, prices, and behavioral risk factors to health spending variation and growth.^21–24 In particular, prior work by Louise Sheiner in 2014 found significant relationships between state-specific spending on Medicare and population age, insurance rate, prevalence of diabetes, and percentage of the population that is Black, along with Medicare pricing. Sheiner also found that four-fifths of the variation in Medicare spend could be attributed to these variables in 2008.²¹ Our analysis produced similar findings for spending across all insurer types. The results of the decomposition of spending estimates the period 2000–19 show that income and consumer prices explained the most variation of all controls tested and that factors generally outside the purview of the health sector explained four-fifths of the variation in predicted health spending per person, a finding similar to that of Sheiner, although her work looked at Medicare spending in 2008.²¹

The fact that income was the factor that explained the most spending variation highlights the known association between wealth and access to health care services throughout the US. Income and regional consumer prices combine in such a way that wealthy states with the highest prices consistently had the highest estimated health spending per person. At the same time, those living in states with lower mean income had lower spending despite having a generally higher need for health care in poor regions because of systematically worse health.²⁵

Our study also found strong associations between estimated spending and behavioral risk factors, such as lower levels of physical activity and higher smoking rates, consistent with existing literature and highlighting that prevention of disease through healthier living may be a means to drive down health spending.²¹ It is worth noting that smoking and physical activity together explained less than one-tenth of estimated health spending variation, offering some evidence that health care resources are not necessarily targeted to people whose behavior may put them in greater need of intervention.

Prior research has also identified relationships between the supply and the use of services, especially hospital services, and spending growth.^21,23,26,27 Our study also supports this association: Regressions showed significant positive relationships between spending estimates and multiple measures of health system capacity and use (see appendix table e6 and figure e6).¹⁷ As one example, this analysis found that a greater supply of health care workers was associated with increased state-level health care spending per person year over year. Earlier work found similar relationships, but with substantial evidence suggesting that general increases in the supply of the clinical health care workforce improve overall health care efficiency through increases in the effectiveness of medical care service provision.²⁸

Although much of the variation in state spending estimates can be explained by demographic and economic characteristics outside the purview of policy makers, this analysis also showed that early Medicaid expansion states had comparable estimated growth in total health spending than states that have not expanded Medicaid. In the expansion states, predicted Medicaid spending drove the increases, whereas the median growth rate for out-of-pocket spending in these states was lower than in states that expanded Medicaid. This finding is consistent with trends emerging on the overall spending impact of the ACA, which show that increased state and national spending is associated with increased access to care.⁸ Increased insurance coverage has been shown to increase access to care at the state level, and Medicaid expansion was associated with gains particularly among the most financially disadvantaged segments of the population.^8,29 Low levels of access to care are associated with higher hospitalization rates and costs; thus, it is plausible that the increased access to care enjoyed by beneficiaries in Medicaid expansion states could lead to a shift in cost away from high-price hospital services in those markets.²⁹

It is worth noting that our study found a significant negative association between increases in state spending estimates and increased Medicaid income eligibility thresholds over time for children (exhibit 4). States that have enhanced care access for children (irrespective of the state’s participation in specifically ACA-related Medicaid expansion initiatives) do not seem to systematically exhibit significant overall differences in long-term spending because of these policies, yet the benefits of increased child health care access are known to be substantial.³⁰

Without asserting causality of health care spending growth, this study helped unpack factors underlying state spending variation by investigating differences in spending estimates across payer types and identifying the impact of system-level variation. After we assessed spending variation associated with the usual and best-understood spending drivers, one-fifth of the variation in health spending remained unexplained. This finding calls for a reexamination of possible factors underlying the often piecemeal, state-specific approach to addressing the rising cost and uneven quality of health care.

Conclusion

Even as health care spending becomes an increasingly dominant portion of the US economy, health outcomes are falling short of national expectations for what this spending should deliver. To reach goals of health spending containment alongside improvement in population health, policy makers must seek data-driven solutions calibrated by accurate assessments of changes in US spending, yet both generating and interpreting credible assessments of spending changes are complex undertakings. Official state-specific spending estimates that date back to 2014 are not recent enough to enable analyses that consider newer policies or changing state characteristics. In addition, structural, legal, economic, and demographic features vary across US states and types of payers, resulting in substantial variation in health spending across states and payers that is challenging to interpret. Our study offers state-level spending estimates that are compatible with existing State Health Expenditure Accounts estimates through 2014 and report the increasing range of health spending by state for a more recent and actionable period. Moving forward, it seems critical as well as feasible to update comprehensive estimates of state-specific health spending and to continue analysis of why spending levels vary so dramatically across the fifty US states and Washington, D.C.

ACKNOWLEDGMENTS

This work benefited from the contributions of Ian Pollock, Meera Beauchamp, and Theresa McHugh at the Institute for Health Metrics and Evaluation. The authors additionally thank Larry Levitt from the Henry J. Kaiser Family Foundation, Joe Thompson from the Arkansas Center for Health Improvement, and David Radley from the Commonwealth Fund for their review and advisement. Funding for this work was provided by an award from Gates Ventures and by the Peterson Center on Healthcare, Grant No. 21016 (for both, Joseph Dieleman, principal investigator).

NOTES

• 1 Hartman M, Martin AB, Washington B, Catlin AThe National Health Expenditure Accounts Team. National health care spending in 2020: growth driven by federal spending in response to the COVID-19 pandemic. Health Aff (Millwood). 2022;41(1):13–25. Go to the article, Google Scholar
• 2 Centers for Medicare and Medicaid Services. National health expenditure data: historical [Internet]. Baltimore (MD): CMS; [cited 2022 Jun 3]. Available from: https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/nationalhealthexpenddata/nationalhealthaccountshistorical Google Scholar
• 3 Keehan SP, Cuckler GA, Poisal JA, Sisko AM, Smith SD, Madison AJet al. National health expenditure projections, 2019–28: expected rebound in prices drives rising spending growth. Health Aff (Millwood). 2020;39(4):704–14. Go to the article, Google Scholar
• 4 Micah AE, Cogswell IE, Cunningham B, Ezoe S, Harle AC, Maddison ERet al. Tracking development assistance for health and for COVID-19: a review of development assistance, government, out-of-pocket, and other private spending on health for 204 countries and territories, 1990–2050. Lancet. 2021;398(10308):1317–43. Google Scholar
• 5 Lassman D, Sisko AM, Catlin A, Barron MC, Benson J, Cuckler GAet al. Health spending by state 1991–2014: measuring per capita spending by payers and programs. Health Aff (Millwood). 2017;36(7):1318–27. Go to the article, Google Scholar
• 6 Centers for Medicare and Medicaid Services. Health expenditures by state of residence, 1991–2014 [Internet]. Baltimore (MD): CMS; [last modified: 2021 Dec 1; cited 2022 Jun 24]. Available from: https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/NationalHealthExpendData/NationalHealthAccountsStateHealthAccountsResidence Google Scholar
• 7 McMorrow S, Kenney GM, Long SK, Goin DE. Medicaid expansions from 1997 to 2009 increased coverage and improved access and mental health outcomes for low-income parents. Health Serv Res. 2016;51(4):1347–67. Crossref, Medline, Google Scholar
• 8 Hero JO, Sinaiko AD, Peltz A, Kingsdale J, Galbraith AA. In New England, partisan differences in ACA Marketplace participation and potential financial harm. Health Aff (Millwood). 2021;40(9):1420–9. Go to the article, Google Scholar
• 9 Mazurenko O, Balio CP, Agarwal R, Carroll AE, Menachemi N. The effects of Medicaid expansion under the ACA: a systematic review. Health Aff (Millwood). 2018;37(6):944–50. Go to the article, Google Scholar
• 10 Fung V, Liang CY, Shi J, Seo V, Overhage L, Dow WHet al. Potential effects of eliminating the individual mandate penalty in California. Health Aff (Millwood). 2019;38(1):147–54. Go to the article, Google Scholar
• 11 Dondero M, Altman CE. Immigrant policies as health policies: state immigrant policy climates and health provider visits among U.S. immigrants. SSM Popul Health. 2020;10:100559. Crossref, Medline, Google Scholar
• 12 Fisher-Owens SA, Soobader MJ, Gansky SA, Isong IA, Weintraub JA, Platt LJet al. Geography matters: state-level variation in children’s oral health care access and oral health status. Public Health. 2016;134:54–63. Crossref, Google Scholar
• 13 Mandal M, Edelstein BL, Ma S, Minkovitz CS. Changes in state policies related to oral health in the United States, 2002–2009. J Public Health Dent. 2014;74(4):266–75. Crossref, Google Scholar
• 14 Pollack Porter KM, Lindberg R, McInnis-Simoncelli A. Considering health and health disparities during state policy formulation: examining Washington State Health Impact Reviews. BMC Public Health. 2019;19(1):862. Google Scholar
• 15 Roberts ET, Hatfield LA, McWilliams JM, Chernew ME, Done N, Gerovich Set al. Changes in hospital utilization three years into Maryland’s Global Budget Program for rural hospitals. Health Aff (Millwood). 2018;37(4):644–53. Go to the article, Google Scholar
• 16 Bureau of Economic Analysis. Regional data: GDP and personal income. SAPCE2 per capita personal consumption expenditures (PCE) by major type of product 1997–2020 [Internet]. Suitland (MD): BEA; [last updated 2021 Oct 8; cited 2022 Jun 3]. Available via query from: https://apps.bea.gov/iTable/index_regional.cfm Google Scholar
• 17 To access the appendix, click on the Details tab of the article online.
• 18 Bureau of Labor Statistics. Consumer Price Index for All Urban Consumers: all items in U.S. city average, all urban consumers, not seasonally adjusted, 1997–2022 [Internet]. Washington (DC): BLS; [cited 2022 Jun 3]. Available via query from: https://www.bls.gov/data/#prices Google Scholar
• 19 Neuman P, Cubanski J, Damico A. Medicare per capita spending by age and service: new data highlights oldest beneficiaries. Health Aff (Millwood). 2015;34(2):335–9. Go to the article, Google Scholar
• 20 Bureau of Economic Analysis. Regional data: GDP and personal income. Regional price parities by state, 2008–2019 [Internet]. Suitland (MD): BEA; [cited 2022 Jun 3]. Available via query from: https://apps.bea.gov/iTable/index_regional.cfm Google Scholar
• 21 Sheiner L. Why the geographic variation in health care spending cannot tell us much about the efficiency or quality of our health care system. Brookings Pap Econ Act. 2014;2014(2):1–72. Google Scholar
• 22 Herring B, Trish E. Explaining the growth in US health care spending using state-level variation in income, insurance, and provider market dynamics. Inquiry. 2015;52:0046958015618971. Google Scholar
• 23 Di Matteo L. The macro determinants of health expenditure in the United States and Canada: assessing the impact of income age distribution and time. Health Policy. 2005;71(1):23–42. Crossref, Google Scholar
• 24 Rettenmaier AJ, Wang Z. Regional variations in medical spending and utilization: a longitudinal analysis of US Medicare population. Health Econ. 2012;21(2):67–82. Crossref, Google Scholar
• 25 Egen O, Beatty K, Blackley DJ, Brown K, Wykoff R. Health and social conditions of the poorest versus wealthiest counties in the United States. Am J Public Health. 2017;107(1):130–5. Crossref, Google Scholar
• 26 Gottlieb DJ, Zhou W, Song Y, Andrews KG, Skinner JS, Sutherland JM. Prices don’t drive regional Medicare spending variations. Health Aff (Millwood). 2010;29(3):537–43. Go to the article, Google Scholar
• 27 Gearhart R, Michieka N. A non-parametric investigation of supply side factors and healthcare efficiency in the U.S. J Prod Anal. 2020;54(1):59–74. Crossref, Google Scholar
• 28 Weaver MR, Nandakumar V, Joffe J, Barber RM, Fullman N, Singh Aet al. Variation in health care access and quality among US states and high-income countries with universal health insurance coverage. JAMA Netw Open. 2021;4(6):e2114730. Crossref, Google Scholar
• 29 Bindman AB, Grumbach K, Osmond D, Komaromy M, Vranizan K, Lurie Net al. Preventable hospitalizations and access to health care. JAMA. 1995;274(4):305–11. Crossref, Medline, Google Scholar
• 30 Institute of Medicine. Health insurance is a family matter. Washington (DC): National Academies Press; 2002. Google Scholar