Variation in Health Care Access and Quality Among US States and High-Income Countries With Universal Health Insurance Coverage | Global Health | JAMA Network Open | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Age-Specific Healthcare Access and Quality (HAQ) Index for US and High-Income Peers With Universal Health Insurance in Selected Years
Age-Specific Healthcare Access and Quality (HAQ) Index for US and High-Income Peers With Universal Health Insurance in Selected Years
Figure 2.  Age-Specific Healthcare Access and Quality (HAQ) Scores for US States in 2016
Age-Specific Healthcare Access and Quality (HAQ) Scores for US States in 2016
Figure 3.  Total Health Insurance Coverage by Age Category for US States in 2016
Total Health Insurance Coverage by Age Category for US States in 2016

Abbreviation: HAQ, healthcare and quality index.

Table 1.  Association Between HAQ Scores and Total Health Insurance Coverage by Age Category in the United States, 2010 and 2016a
Association Between HAQ Scores and Total Health Insurance Coverage by Age Category in the United States, 2010 and 2016a
Table 2.  Summary of Counterfactual Estimates of the Increase in HAQ Score in US States With Universal Health Insurance Coverage, 2010 and 2016a
Summary of Counterfactual Estimates of the Increase in HAQ Score in US States With Universal Health Insurance Coverage, 2010 and 2016a
1.
Herper  M. The most innovative countries in biology and medicine. Forbes. March 23, 2011. Accessed September 14, 2020. https://www.forbes.com/sites/matthewherper/2011/03/23/the-most-innovative-countries-in-biology-and-medicine/
2.
Arrepim. Nobel Prize and other awards charts, graphs and statistics. Accessed September 20, 2020. http://stats.areppim.com/allawards.htm
3.
Nolte  E, McKee  M.  Measuring the health of nations: analysis of mortality amenable to health care.   BMJ. 2003;327(7424):1129. doi:10.1136/bmj.327.7424.1129PubMedGoogle ScholarCrossref
4.
Nolte  E, McKee  M.  Variations in amenable mortality—trends in 16 high-income nations.   Health Policy. 2011;103(1):47-52. doi:10.1016/j.healthpol.2011.08.002PubMedGoogle ScholarCrossref
5.
Barber  RM, Fullman  N, Sorensen  RJD,  et al; GBD 2015 Healthcare Access and Quality Collaborators.  Healthcare Access and Quality index based on mortality from causes amenable to personal health care in 195 countries and territories, 1990-2015: a novel analysis from the Global Burden of Disease Study 2015.   Lancet. 2017;390(10091):231-266. doi:10.1016/S0140-6736(17)30818-8PubMedGoogle ScholarCrossref
6.
Fullman  N, Yearwood  J, Abay  SM,  et al; GBD 2016 Healthcare Access and Quality Collaborators.  Measuring performance on the Healthcare Access and Quality index for 195 countries and territories and selected subnational locations: a systematic analysis from the Global Burden of Disease Study 2016.   Lancet. 2018;391(10136):2236-2271. doi:10.1016/S0140-6736(18)30994-2PubMedGoogle ScholarCrossref
7.
Obama  B.  United States health care reform: progress to date and next steps.   JAMA. 2016;316(5):525-532. doi:10.1001/jama.2016.9797PubMedGoogle ScholarCrossref
8.
United States Department of Labor. Health plans & benefits: Employee Retirement Income Security Act of 1974 (ERISA). Accessed September 14, 2020. https://www.dol.gov/general/topic/health-plans/erisa
9.
Institute of Medicine.  Health Insurance Is a Family Matter. The National Academies Press; 2002.
10.
The Henry J. Kaiser Family Foundation. State health facts: health reform indicators. marketplace enrollment, 2014-2020, and Medicaid expansion enrollment. Accessed on September 22, 2020. https://www.kff.org/health-reform/state-indicator/marketplace-enrollment/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D
11.
Rutstein  DD, Berenberg  W, Chalmers  TC, Child  CG  III, Fishman  AP, Perrin  EB.  Measuring the quality of medical care: a clinical method.   N Engl J Med. 1976;294(11):582-588. doi:10.1056/NEJM197603112941104PubMedGoogle ScholarCrossref
12.
Nolte  E, McKee  M. Does health care save lives? avoidable mortality revisited. The Nuffield Trust. 2004.
13.
GBD 2016 Risk Factors Collaborators.  Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1345-1422. doi:10.1016/S0140-6736(17)32366-8PubMedGoogle ScholarCrossref
14.
GBD 2016 Causes of Death Collaborators.  Global, regional, and national age-sex specific mortality for 264 causes of death, 1980-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1151-1210. doi:10.1016/S0140-6736(17)32152-9PubMedGoogle ScholarCrossref
15.
GBD 2016 Disease and Injury Incidence and Prevalence Collaborators.  Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1211-1259. doi:10.1016/S0140-6736(17)32154-2PubMedGoogle ScholarCrossref
16.
Radley  DC, McCarthy D, Hayes  SL. 2018 Scorecard on state health system performance. Commonwealth Fund. May 3, 2018. Accessed December 3, 2018. https://www.commonwealthfund.org/publications/fund-reports/2018/may/2018-scorecard-state-health-system-performance
17.
Davern  M, Quinn  BC, Kenney  GM, Blewett  LA.  The American Community Survey and health insurance coverage estimates: possibilities and challenges for health policy researchers.   Health Serv Res. 2009;44(2 Pt 1):593-605. doi:10.1111/j.1475-6773.2008.00921.xPubMedGoogle Scholar
18.
Organization for Economic Cooperation and Development. Total public and primary private health insurance. Accessed September 18, 2020. http://stats.oecd.org/#
19.
GBD 2016 Mortality Collaborators.  Global, regional, and national under-5 mortality, adult mortality, age-specific mortality, and life expectancy, 1970-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1084-1150. doi:10.1016/S0140-6736(17)31833-0PubMedGoogle ScholarCrossref
20.
Nolte  E, McKee  CM.  In amenable mortality—deaths avoidable through health care—progress in the US lags that of three European countries.   Health Aff (Millwood). 2012;31(9):2114-2122. doi:10.1377/hlthaff.2011.0851PubMedGoogle ScholarCrossref
21.
Levy  H, Meltzer  D.  The impact of health insurance on health.   Annu Rev Public Health. 2008;29(1):399-409. doi:10.1146/annurev.publhealth.28.021406.144042PubMedGoogle ScholarCrossref
22.
Dor  A, Umapathi  E. Health insurance and health. In Culyer  AJ, ed.  Encyclopedia of Health Economics. Elsevier; 2014.
23.
Currie  J, Gruber  J.  Saving babies: the efficacy and cost of recent changes in the Medicaid eligibility of pregnant women.   J Polit Econ. 1996; 104: 1263-1296. doi:10.1086/262059Google ScholarCrossref
24.
Currie  J, Gruber  J.  Health insurance eligibility, utilization of Medical care, and child health.   Q J Econ. 1996; 111(2):431–466. doi:10.2307/2946684Google ScholarCrossref
25.
Goldman  DP, Bhattacharya  J, McCaffrey  DF, Duan  N, Leibowitz  AA, Joyce  GF,  et al.  Effect of insurance on mortality in an HIV-positive population in care.   J Am Stat Assoc. 2001; 96(455):883–894. doi:10.1198/016214501753208582Google ScholarCrossref
26.
Bhattacharya  J, Goldman  D, Sood  N.  The link between public and private insurance and HIV-related mortality.   J Health Econ. 2003;22(6):1105-1122. doi:10.1016/j.jhealeco.2003.07.001PubMedGoogle ScholarCrossref
27.
Card  D, Dobkin  C, Maestas  N.  Does Medicare save lives?   Q J Econ. 2009;124(2):597-636. doi:10.1162/qjec.2009.124.2.597PubMedGoogle ScholarCrossref
28.
Andersen  MS.  Effects of Medicare coverage for the chronically ill on health insurance, utilization, and mortality: evidence from coverage expansions affecting people with end-stage renal disease.   J Health Econ. 2018;60:75-89. doi:10.1016/j.jhealeco.2018.06.002PubMedGoogle ScholarCrossref
29.
Chen  A, Oster  E, Williams  H.  Why is infant mortality higher in the United States than in Europe?   Am Econ J Econ Policy. 2016;8(2):89-124. doi:10.1257/pol.20140224PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    2 Comments for this article
    EXPAND ALL
    Important Omission
    Jeoffry Gordon, MD, MPH | Department of Public Health
    Although this is an important and detailed study comparing health outcomes by age and insurance status in US states with other wealthy countries it has missed an important opportunity and data point by not further specifying insurance source, e.g. commercial (employer sponsored vs individual), traditional Medicare, Medicare Advantage, Medicaid, or Veteran's Administration. One could hypothesize that outcome scores might be better for commercially insured or Medicare versus Medicaid, or even better under VA insurance. This type of study is the sort that can discern such differences. Perhaps type of insurance data was not collected in the source survey; perhaps consideration would have been too complex for this paper. Nonetheless, these important distinctions could have a large impact on public policy debates going forward. Remarkably the authors did not note this deficiency under limitations.
    CONFLICT OF INTEREST: None Reported
    READ MORE
    Author Reply: Insurance Type Omitted But Not Overlooked
    Marcia Weaver, PhD | University of Washington
    Thank you, Dr. Gordon, for your careful reading of our article and thoughtful comment. Although our work on private insurance coverage was omitted, the topic was not overlooked. The description of types of insurance reported in the American Community Survey (ACS) was the last sentence deleted from the revised manuscript. ACS respondents are asked about insurance coverage from at least six sources: 1) employer or union, 2) direct purchase from an insurance company, 3) Medicare, 4) Medicaid, Medical Assistance or any kind of government-assistance plan for those with low income or disability, 5) TRICARE military health care, and 6) Veterans Affairs.

    We did not prioritize reporting the results for private insurance for two reasons. First, a key contribution of our research was reporting the relatively slow improvement in recent years in US Healthcare Access and Quality (HAQ) scores among people in working ages 15 to 64, for whom total insurance coverage was more relevant than private. In our initial, unpublished comparison of age-specific HAQ scores across states without state fixed effects for ages 25 to 54, HAQ scores were associated with total insurance coverage (0.091, 95% confidence interval (CI)=0.039, 0.144), but not private. For ages 15 to 24 years, the association with total insurance coverage was larger than private (0.200, 95% CI=0.164, 0.237 vs 0.152, 95% CI=0.164, 0.237). For ages 55 to 64 years, the association was similar for total and private insurance coverage (0.202, 95% CI=0.106, 0.298 vs 0.224, 95% CI=0.151, 0.297).

    Health care access and quality for children is also of concern, but the relatively poor performance of the US health care system for infants is not a recent development. As shown in Figure 1, the US HAQ score for ages 0 to 6 days was the lowest among high-income peers in 1990, 2000, 2010, and 2016. In our initial, unpublished comparison without state fixed effects, HAQ scores were associated with private insurance coverage for ages 0-11 months (0.123, 95% CI=0.054, 0.191) and 2-14 years (0.058, 95% CI=0.029, 0.087), but not with total. NB: The first column in Figure 3 should be labeled “0-11 months.” Infant mortality rates in the US are a topic of ongoing research; for example, see Chen et al's (2016) results about post-neonatal mortality rates.

    Second, total insurance coverage may be reported more accurately than private insurance in the ACS. As explained in Davern et al (2009), the ACS questionnaire does not use specific state names for State Children’s Health Insurance Programs and Medicaid. Evidence from the Current Population Survey suggests that using specific names for these state programs may reduce over-reporting of private insurance coverage and under-reporting of total insurance coverage.

    In closing, we agree that comparing the effects of total and private insurance coverage on health care access and quality would be worthwhile, especially for children. Also, other characteristics of private insurance programs such as benefits package, continuity of coverage, or cost-sharing may promote access and quality, and the evidence may support incorporating them into public insurance programs.
    CONFLICT OF INTEREST: I received funding from the National Institute on Aging of the United States National Institutes of Health during the conduct of the study.
    READ MORE
    Original Investigation
    Health Policy
    June 28, 2021

    Variation in Health Care Access and Quality Among US States and High-Income Countries With Universal Health Insurance Coverage

    Author Affiliations
    • 1Departments of Health Metrics Sciences and Global Health, Institute for Health Metrics and Evaluation, University of Washington, Seattle
    • 2Institute for Health Metrics and Evaluation, University of Washington, Seattle
    • 3Department of Economics, Occidental College, Los Angeles, California
    JAMA Netw Open. 2021;4(6):e2114730. doi:10.1001/jamanetworkopen.2021.14730
    Key Points

    Question  Does personal health care access and quality vary across ages among high-income countries and US states, and does any observed variation associate with insurance coverage?

    Findings  In this cross-sectional study of age-specific Healthcare Access and Quality (HAQ) across the US, Canada, and 3 high-income regions with universal health insurance coverage (western Europe, high-income Asia Pacific, and Australasia), 2016 US national scores were lower than high-income peers with universal health insurance coverage among individuals of working ages between 15 and 64 years. Across US states in 2010 and 2016, age-specific HAQ scores were associated with insurance coverage for some working-age categories.

    Meaning  These findings suggest that personal HAQ is associated with insurance coverage.

    Abstract

    Importance  Based on mortality estimates for 32 causes of death that are amenable to health care, the US health care system did not perform as well as other high-income countries, scoring 88.7 out of 100 on the 2016 age-standardized Healthcare Access and Quality (HAQ) index.

    Objective  To compare US age-specific HAQ scores with those of high-income countries with universal health insurance coverage and compare scores among US states with varying insurance coverage.

    Design, Setting, and Participants  This cross-sectional study used 2016 Global Burden of Diseases, Injuries, and Risk Factor study results for cause-specific mortality with adjustments for behavioral and environmental risks to estimate the age-specific HAQ indices. The US national age-specific HAQ scores were compared with high-income peers (Canada, western Europe, high-income Asia Pacific countries, and Australasia) in 1990, 2000, 2010, and 2016, and the 2016 scores among US states were also analyzed. The Public Use Microdata Sample of the American Community Survey was used to estimate insurance coverage and the median income per person by age and state. Age-specific HAQ scores for each state in 2010 and 2016 were regressed based on models with age fixed effects and age interaction with insurance coverage, median income, and year. Data were analyzed from April to July 2018 and July to September 2020.

    Main Outcomes and Measures  The age-specific HAQ indices were the outcome measures.

    Results  In 1990, US age-specific HAQ scores were similar to peers but increased less from 1990 to 2016 than peer locations for ages 15 years or older. For example, for ages 50 to 54 years, US scores increased from 77.1 to 82.1 while high-income Asia Pacific scores increased from 71.6 to 88.2. In 2016, several states had scores comparable with peers, with large differences in performance across states. For ages 15 years or older, the age-specific HAQ scores were 85 or greater for all ages in 3 states (Connecticut, Massachusetts, and Minnesota) and 75 or less for at least 1 age category in 6 states. In regression analysis estimates with state-fixed effects, insurance coverage coefficients for ages 20 to 24 years were 0.059 (99% CI, 0.006-0.111); 45 to 49 years, 0.088 (99% CI, 0.009-0.167); and 50 to 54 years, 0.101 (99% CI, 0.013-0.189). A 10% increase in insurance coverage was associated with point increases in HAQ scores among the ages of 20 to 24 years (0.59 [99% CI, 0.06-1.11]), 45 to 49 years (0.88 [99% CI, 0.09-1.67]), and 50 to 54 years (1.01 [99% CI, 0.13-1.89]).

    Conclusions and Relevance  In this cross-sectional study, the US age-specific HAQ scores for ages 15 to 64 years were low relative to high-income peer locations with universal health insurance coverage. Among US states, insurance coverage was associated with higher HAQ scores for some ages. Further research with causal models and additional explanations is warranted.

    Introduction

    Despite the contributions of the US to biology and medical science,1,2 the US health care system does not perform as well as most high-income countries according to various measures.3-6 The Healthcare Access and Quality (HAQ) index, created by Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) collaborators, is based on amenable mortality, defined as deaths that should not occur in the presence of timely and effective care.3 According to the age-standardized HAQ index, the US ranked 29 of 195 countries in 2016 with a score of 88.7 (95% uncertainty interval [UI], 88.0-89.4).6 The US health care system serves some populations better than others. When comparing populations across states, the 2016 age-standardized HAQ scores ranged from a high in Minnesota of 92.3 (95% UI, 90.6-93.6) to a low in Mississippi of 81.5 (95% UI, 78.6-84.2).6 Health insurance coverage varies by age and state because state governments can expand benefits and eligibility for programs above the minimum federal requirements,7-10 with the exception of federally provided Medicare for individual aged 65 or older or individuals who are disabled and eligible for Social Security benefits or have end-stage kidney disease. In this study, our objective was to compare US age-specific HAQ scores with those of high-income countries with universal health insurance coverage and compare scores among US states with varying insurance coverage.

    Methods
    Reporting Guidelines

    This cross-sectional study did not require ethical review by the University of Washington Human Subjects Division or informed consent because it used GBD results and US Census Bureau public-use data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

    HAQ Index

    Health care system performance is commonly measured at the population level by mortality among children ages 0 to 5 years and maternal mortality because these mortality rates are low in well-functioning health care systems. Beginning with Rutstein et al11 in 1976, researchers have sought additional measures that include more age categories, more causes, and a broader range of health services. Nolte and McKee’s12 list of amenable causes among ages 0 to 74 years is the most widely used and the basis of the HAQ index. The HAQ index combines mortality data from 32 amenable causes (eTable 1 in the Supplement).5,6 Examples of amenable causes are neonatal disorders, maternal disorders, and 4 vaccine preventable diseases, for which high-quality care affects mortality among children ages 0 to 5 years and maternal mortality; colon, rectum, and breast cancers, for which high-quality care includes screening and early detection as well as treatment; ischemic heart disease, cerebrovascular disease, and diabetes, for which high-quality care includes management of high systolic blood pressure or fasting plasma glucose in primary care; and appendicitis, for which high-quality care includes emergency surgical care.

    The 2016 HAQ index was calculated with risk-standardized, 2016 GBD mortality rates for 24 amenable causes, in which risk-standardization controls for mortality differences across locations and years attributable to environmental and behavioral risk factors as opposed to personal health care.5,6 Risk standardization was based on 2016 GBD comparative risk assessment results.13 It also used 2016 GBD mortality-to-incidence ratios for 8 amenable neoplasms to more robustly reflect differences in access and quality of cancer care (eTable 1 in the Supplement).6 The GBD analysis provides mortality14 and incidence15 estimates by cause and age category, which are 5-year intervals covering ages 5 to 74 years, and 4 narrower categories for ages 0 to 4 years.

    We created age-specific HAQ indices in 1990, 2000, 2010, and 2016 for this analysis that began with age-specific, risk-standardized mortality rates and mortality incidence ratios. The following 2 steps were performed separately for each age category. The cause-specific measures were log transformed and rescaled on a 0 to 100 scale, using the worst (first percentile) and best (99th percentile) mortality results within each age category observed across all countries from 1990 to 2016 to set the minimum and maximum, respectively. With secular improvements in health care access and quality, the worst mortality results for each cause and age category are generally from earlier years and the best are generally from later years.

    Next, to combine the cause-specific scales for each age category, we calculated a weighted mean using the cause weights from the age-standardized 2016 HAQ index. These cause weights are estimated using principal components analysis from the age-standardized cause-specific measures across all health care systems globally (ie, 1 measure for each cause, location, and year). The cause weights reflect patterns of mortality across health care systems. In contrast, an index based on total amenable mortality16 would be weighted by the distribution of mortality in each location, for example, disproportionately weighting ischemic heart disease in the US. By using the same cause weights across all locations, the HAQ index isolates a measure of health care system performance.

    However, some causes are not amenable or relevant for all age categories; for example, the amenable range for diarrheal diseases is 0 to 14 years, and maternal conditions are not relevant for girls ages 0 to 9 years. For each age category, the cause weights are rescaled to sum to 1 (eTable 2 in the Supplement).

    The GBD results for mortality rates are 1000 draws from the posterior distribution of the cause-specific mortality rates for each age, sex, location, and year.14 We used 1000 draws from the GBD mortality results and 1000 draws from the GBD incidence results for neoplasms15 to construct the age-specific HAQ index and report the mean of 1000 draws for each age-specific HAQ index.

    American Community Survey

    To estimate insurance coverage and the median income per person by age category and state, we used the American Community Survey (ACS) Public Use Microdata Sample (PUMS).17 The ACS is an annual survey of a sample of residential addresses selected from every county, and the surveys beginning in 2008 provide the most precise, available estimates of health insurance coverage at the state level and population subgroups at the national level.17 Respondents were asked about insurance coverage at the time of the interview for each member of their household or group quarters.17 The PUMS data covered more than one-third of the overall sample and represented 1% of the US population. To match the 2016 GBD results, we analyzed annual PUMS data for 2010, with samples of 2 769 241 people in households and 64 677 people in group quarters, and for 2016 with samples of 2 778 447 people in households and 124 644 people in group quarters. Sample person weights were applied to observations to ensure representativeness.

    We defined total insurance coverage as coverage from 1 or more of the 6 sources in the ACS and calculated the percentage of the population with insurance coverage in the same 5-year intervals covering ages 5 to 74 years, and 2 categories for children aged 0 to 11 months and 1 to 4 years. ACS data were not disaggregated for children younger than 1 year. A state’s age-specific median income per person represented household income per person for people in households and personal income for individuals in group quarters.

    Statistical Analysis

    We first estimated the age-specific HAQ index using national data on the US and high-income peer countries with universal health insurance coverage in 1990, 2000, 2010, and 2016 to identify relative levels, trends, and differences. Results are reported for the US, 3 high-income GBD regions with universal health insurance coverage (ie, western Europe, high-income Asia Pacific, and Australasia), and Canada. The Organization for Economic Cooperation and Development reports that 91.2% of the population in the US had insurance coverage in 2016, and 100% had it in the 3 GBD regions and Canada.18 The quality of death registration data in these locations is also high.19 We did not include the high-income Latin American region because available evidence showed that less than 100% of the population had insurance coverage in 2016, and countries did not report coverage by age. eTable 3 in the Supplement provides a list of countries in high-income GBD regions, the number of years with complete death registration data, data quality, and the percentage of the population with insurance.

    The mean scores for each US state and the District of Columbia in 2016 were calculated to describe differences in age-specific HAQ scores by state. Details on national cause-specific mortality scores and the range of state scores for each amenable cause are reported in the eAppendix, eFigure 1, and eFigure 2 in the Supplement.

    A multivariable regression analysis was conducted to analyze the association between the age-specific HAQ scores and insurance coverage at the state level in 2010 and 2016. The unit of analysis was the state-age-year with 1632 observations (51 states, 16 age categories, and 2 years). The dependent variable was the state’s mean age-specific HAQ score, and independent variables were the age-specific insurance coverage, age-specific median income per person, fixed effects for each age category, and year. Coefficients for insurance coverage, median income per person, and year were estimated for each age category. Median income per person controlled for the financial resources available, which was similar to limiting the national comparison to high-income countries.

    Pooled ordinary least-squares analyses were reported without and with state-fixed effects. Analyses without state-fixed effects estimated the association between HAQ scores and insurance coverage across states within age categories. If unmeasured state characteristics are associated with age-specific insurance coverage, insurance coefficients may be biased and their standard errors may be underestimated. Analysis with state-fixed effects controlled for unmeasured state characteristics. Results are presented with 99% CIs and based on a 1% level of significance to address multiple testing. We compared the outcome associated with a 1 SD change in total insurance coverage or median income per capita (eTable 4 in the Supplement). Sensitivity analyses were conducted with 2 state-level covariates that measure health care system infrastructure: hospital beds per 1000 population and physicians per 1000 population (eTable 5 and eTable 6 in the Supplement). A counterfactual analysis was conducted to calculate the increase in HAQ scores with universal insurance coverage for every state and each age category using the regression results.

    The age-specific HAQ indices were constructed and national-level analyses were performed with Python statistical software version 3.0. (Python). The state-level analyses were performed with R statistical software version 3.6.3. (R Project for Statistical Computing).

    Results
    International and National Age-Specific HAQ Scores From 1990 to 2016

    In 1990, the age pattern in the age-specific HAQ indices was similar for the US and high-income peers with universal health insurance coverage (Figure 1). Scores were low among children younger than 1 year, high for ages 5 to 9 years and 10 to 14 years and decreased with age beginning at ages 15 to 19 years. For example, for ages 20 to 24 years, the US score (82.8 points) was in the middle among high-income Asia Pacific (72.1 points), western Europe (79.2 points), Australasia (83.1 points), and Canada (83.8 points). Although the range was wider for ages 50 to 54 years, the US score (77.1 points) was in the middle among high-income Asia Pacific (71.6 points), western Europe (73.4 points), Canada (78.1 points), and Australasia (80.3 points). The US score was the lowest observed was ages 0 to 6 days (67.4 points) and highest observed for ages 15 to 19 years (83.7 points), 65 to 69 years (76.1 points), and 70 to 74 years (75.2 points).

    No location had an HAQ score of 100 because the HAQ index was a weighted mean, and no location consistently had the lowest cause-specific mortality result for every cause. The regional results were population-weighted means across countries. Although 1 country may have had a relatively high HAQ score, the regional mean was lower.

    Age-specific HAQ scores increased over time, but the US scores increased less for ages 5 years or older than high-income peer locations from 1990 to 2010 and did not change appreciably from 2010 to 2016. For example, for ages 20 to 24 years, HAQ scores increased 5.0 points (87.8 points) in the US from 1990 to 2016 compared with 19.7 points (91.8 points) in high-income Asia Pacific, 12.4 points (91.6 points) in western Europe, 10.7 points (93.8 points) in Australasia, and 8.8 points (92.6 points) in Canada. For ages 50 to 54 years, the HAQ scores increased 4.9 points (82.1 points) in the US, compared with 16.5 points (88.2 points) in high-income Asia Pacific, 14.0 points (87.4 points) in western Europe, 11.0 points (89.1 points) in Canada, and 12.6 points (92.9 points) in Australasia. Thus, the US ranked the lowest among its peers in every age category in 2016, with the exception ages 65 to 69 years and 70 to 74 years.

    Age-Specific HAQ Scores by US State in 2016

    Although the US national age-specific HAQ scores were less than those at peer locations in 2016, HAQ scores in some states were comparable with peer locations. For example, the age-specific HAQ scores were 85 or greater for individuals ages 15 years or older in 3 states (ie, Connecticut, Massachusetts, and Minnesota), and 84 or greater in 8 states (Figure 2). In contrast, the age-specific HAQ scores were 75 or less for at least 1 age category between the ages of 15 to 64 years in 6 states (ie, Alabama, Arkansas, District of Columbia, Louisiana, Mississippi, and West Virginia). The ranges in age-specific HAQ scores across states were large; for example, it was 9.6 points for ages 15 to 19 years and 13.3 points for ages 60 to 64 years (eFigure 2 in the Supplement).

    Association Between Health Insurance Coverage and Age-Specific HAQ Index in the US 2010 and 2016

    In the multivariable analysis of state-level data, the age-specific HAQ score was positively associated with insurance coverage for most working ages between 15 and 64 years (Table 1). For example, in pooled estimates without state-fixed effects, the coefficients were large and statistically significant for working ages, with 2 exceptions at ages 25 to 29 years and 55 to 59 years. A 10% absolute increase in insurance coverage was associated with an increase of 1.43 (99% CI, 0.48-2.37) points in HAQ score for ages 20 to 24 years, 2.95 (99% CI, 1.51-4.39) points for ages 45 to 49 years, and 2.51 points (99% CI, 0.90-4.11) for ages 50 to 54 years. In pooled estimates with state-fixed effects, the coefficients were smaller and statistically significant for 3 categories of working-age individuals: ages 20 to 24 years (0.059 [99% CI, 0.006-0.111]), 45 to 49 years (0.088 [99% CI, 0.009-0.167]), and 50 to 54 years (0.101 [99% CI, 0.013-0.189]). For these groups, a 10% absolute increase in insurance coverage was associated with an increase in HAQ scores of 0.59 (99% CI, 0.06-1.11) points for ages 20 to 24 years, 0.88 (99% CI, 0.09-1.67) points for ages 45 to 49 years, and 1.01 (99% CI, 0.13-1.89) points for ages 50 to 54 years.

    The increases in HAQ scores associated with universal health insurance coverage were calculated as the product of the insurance gap and the regression coefficients (Figure 3). In 2016, using pooled estimates without state-fixed effects, universal health insurance coverage would increase HAQ scores by at least 3 points for ages 40 to 44 in 30 of 50 states and the District of Columbia (Table 2; eTable 7 in the Supplement). Using pooled estimates with state-fixed effects, HAQ scores would increase by at least 1 point for ages 20 to 24 years in 13 states, ages 45 to 49 years in 21 states, ages 50 to 54 years in 23 states, and ages 60 to 64 years in 9 states.

    Discussion

    This cross-sectional study found that when health care system performance was measured at the population level with an age-specific HAQ index, the US national scores increased from 1990 to 2010 across age categories. However, the increase was less than those of high-income peers with universal health insurance coverage and did not change between 2010 and 2016. Using state-level age-specific HAQ scores, we identified several states with scores in 2016 that were comparable with high-income peers and large differences in performance across states. Analyzing these state-level results, we found that the age-specific HAQ scores were associated with insurance coverage in some working-age categories.

    Our comparison of national results for the age-specific HAQ indices is consistent with previous research on amenable mortality in the US and adds insights into the US performance during working ages. The absence of an increase in HAQ scores between 2010 and 2016 is consistent with the Commonwealth Fund’s report16 that the total mortality rate from amenable causes declined at a slower rate from 2010 and 2011 to 2012 and 2013 than from 2004 and 2005 to 2010 and 2011 and increased from 83.7 per 100 000 population in 2012 and 2013 to 84.3 in 2013 and 2014. Nolte and McKee20 reported that the age-standardized amenable mortality rate for ages 0 to 64 years was high in the US relative to France, Germany, and the United Kingdom from 1999 to 2006 and 2007. Mortality rates for ages 65 to 74 years were similar in the US, Germany, and the United Kingdom in 2006 and 2007 and higher than in France.

    In the international comparisons, the US health care system performed poorly for working ages in recent years, and this finding was consistent with lower insurance coverage for these age groups in the US. In the comparison among states, total insurance coverage and the median income per person were associated with increased HAQ scores for working-aged individuals; increases in coverage and income were associated with offsetting a negative time trend between 2010 and 2016. However, total insurance coverage may be confounded with other state-level insurance policies, such as the benefits package, continuity of coverage, and cost-sharing, or state-level employment policies, such as paid leave for medical care. The association between the HAQ scores and total insurance coverage suggests that further research on the effects of other state-level insurance and employment policies is warranted.

    Other potential reasons why health care system performance would differ across age categories or states in the US were addressed in our estimates. For example, environmental, occupational, and behavioral risk factors differed across ages, states, and years, and we controlled for these risks in our analysis by calculating the age-specific HAQ index with risk-standardized mortality rates. Also, health infrastructure, such as hospitals or physicians per 1000 population, differed across states and years but not ages. In estimates that included these state-level variables, the coefficient estimates for insurance coverage were similar or larger than estimates without state-level variables. These state-level variables did not improve model fit in our sample with 2 years of data and estimates with state fixed effects.

    Other possible factors that could contribute to health care system performance differences, which we did not address in this study, are extended office hours or location of health facilities close to populations with inflexible work hours or without sick leave. However, an analysis of these factors would require local-level data, and measures of insurance coverage by 5-year age categories are unavailable at the local level.

    To our knowledge, this is the first study to explore the association between health insurance coverage and US HAQ scores across multiple age categories at the state level. Although our analysis focuses on health care system performance, it may have implications for research on the effects of insurance coverage on health. Although the effects of health insurance on health are unclear, a review by Levy and Metzler21 concluded that health insurance improves health for populations such as infants, children, and people with AIDS or high blood pressure. A review by Dor and Umapathi22 recommended further research on which patient populations would benefit from insurance coverage. There is evidence that insurance coverage reduces mortality for infants,23 young children,24 people with HIV,25,26 hospital patients admitted from an emergency department,27 and people with end-stage kidney disease.28 Health insurance may be more likely to affect amenable causes and benefit populations without access to high-quality care for those causes. Future researchers could test the outcomes associated with insurance coverage in the US with appropriately weighted mortality or disability-adjusted life-years for amenable causes in age categories with the largest ranges in HAQ scores.

    Limitations

    This study had limitations. Amenable mortality, as defined by Nolte and McGee3 ends at age 74 years, which was adopted for the HAQ index. Since 2003, advances in health care for older individuals may have extended amenable mortality to ages 75 years or older for some causes. Similarly, amenable mortality combines results for both sexes, which limits the understanding of potential sex-specific differences in access and quality; future research could use GBD mortality results for each sex to calculate sex-specific and age-specific HAQ indices.

    The HAQ index was calculated with only the principal component analysis (PCA) weights from the age-standardized HAQ. In the future, it would be possible to estimate the age-specific HAQ indices with both PCA and arithmetic mean to determine whether the method of combining the cause-specific mortality rates altered the comparison among US states and between the US and peer locations. Furthermore, the international comparison was not included, and socioeconomic characteristics such as education, racial and ethnic discrimination, and immigration status, were not controlled for in the regression analysis. Our estimates controlled for age-specific median income per person but did not compare HAQ scores between income categories across states. Chen et al29 show differences in mortality among children aged 2 to 12 months across US regions among disadvantaged groups, but not among advantaged groups. The ACS sample size is designed for state-level analyses or national-level subgroup analyses but not for state-level subgroup analyses. Although statistically significant associations were found, our cross-sectional study was limited by its design and did not determine causation.

    Conclusions

    In this cross-sectional study, the 2016 US age-specific HAQ scores for working individuals ages 15 to 64 years were low compared with high-income global peers with universal health insurance coverage. Among US states, insurance coverage was associated with higher HAQ scores for some working ages between 15 and 64 years. Further research with causal models and additional explanations is warranted.

    Back to top
    Article Information

    Accepted for Publication: April 25, 2021.

    Published: June 28, 2021. doi:10.1001/jamanetworkopen.2021.14730

    Correction: This article was corrected on August 6, 2021, to fix an error in a label on Figure 3.

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Weaver MR et al. JAMA Network Open.

    Corresponding Author: Marcia R. Weaver, PhD, Departments of Health Metrics Sciences and Global Health, Institute for Health Metrics and Evaluation, University of Washington, 3980 15th Ave NE, Hans Rosling Center, Seattle, WA 98105 (mweaver@uw.edu).

    Author Contributions: Dr Weaver had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Weaver, Barber, Yearwood, Murray.

    Acquisition, analysis, or interpretation of data: Nandakumar, Joffe, Fullman, Singh, Sparks, Yearwood, Lozano, Ngo.

    Drafting of the manuscript: Weaver, Nandakumar.

    Critical revision of the manuscript for important intellectual content: Joffe, Barber, Fullman, Singh, Sparks, Yearwood, Lozano, Murray, Ngo.

    Statistical analysis: Weaver, Nandakumar, Joffe, Barber, Singh, Sparks, Yearwood.

    Obtained funding: Weaver.

    Administrative, technical, or material support: Nandakumar, Sparks.

    Supervision: Weaver, Murray, Ngo.

    Conflict of Interest Disclosures: Dr Weaver reported receiving grants from the National Institute on Aging of the United States National Institutes of Health during the conduct of the study. No other disclosures were reported.

    Funding/Support: Drs Weaver and Murray were funded by grant P30AG047845 from the National Institute on Aging.

    Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

    References
    1.
    Herper  M. The most innovative countries in biology and medicine. Forbes. March 23, 2011. Accessed September 14, 2020. https://www.forbes.com/sites/matthewherper/2011/03/23/the-most-innovative-countries-in-biology-and-medicine/
    2.
    Arrepim. Nobel Prize and other awards charts, graphs and statistics. Accessed September 20, 2020. http://stats.areppim.com/allawards.htm
    3.
    Nolte  E, McKee  M.  Measuring the health of nations: analysis of mortality amenable to health care.   BMJ. 2003;327(7424):1129. doi:10.1136/bmj.327.7424.1129PubMedGoogle ScholarCrossref
    4.
    Nolte  E, McKee  M.  Variations in amenable mortality—trends in 16 high-income nations.   Health Policy. 2011;103(1):47-52. doi:10.1016/j.healthpol.2011.08.002PubMedGoogle ScholarCrossref
    5.
    Barber  RM, Fullman  N, Sorensen  RJD,  et al; GBD 2015 Healthcare Access and Quality Collaborators.  Healthcare Access and Quality index based on mortality from causes amenable to personal health care in 195 countries and territories, 1990-2015: a novel analysis from the Global Burden of Disease Study 2015.   Lancet. 2017;390(10091):231-266. doi:10.1016/S0140-6736(17)30818-8PubMedGoogle ScholarCrossref
    6.
    Fullman  N, Yearwood  J, Abay  SM,  et al; GBD 2016 Healthcare Access and Quality Collaborators.  Measuring performance on the Healthcare Access and Quality index for 195 countries and territories and selected subnational locations: a systematic analysis from the Global Burden of Disease Study 2016.   Lancet. 2018;391(10136):2236-2271. doi:10.1016/S0140-6736(18)30994-2PubMedGoogle ScholarCrossref
    7.
    Obama  B.  United States health care reform: progress to date and next steps.   JAMA. 2016;316(5):525-532. doi:10.1001/jama.2016.9797PubMedGoogle ScholarCrossref
    8.
    United States Department of Labor. Health plans & benefits: Employee Retirement Income Security Act of 1974 (ERISA). Accessed September 14, 2020. https://www.dol.gov/general/topic/health-plans/erisa
    9.
    Institute of Medicine.  Health Insurance Is a Family Matter. The National Academies Press; 2002.
    10.
    The Henry J. Kaiser Family Foundation. State health facts: health reform indicators. marketplace enrollment, 2014-2020, and Medicaid expansion enrollment. Accessed on September 22, 2020. https://www.kff.org/health-reform/state-indicator/marketplace-enrollment/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D
    11.
    Rutstein  DD, Berenberg  W, Chalmers  TC, Child  CG  III, Fishman  AP, Perrin  EB.  Measuring the quality of medical care: a clinical method.   N Engl J Med. 1976;294(11):582-588. doi:10.1056/NEJM197603112941104PubMedGoogle ScholarCrossref
    12.
    Nolte  E, McKee  M. Does health care save lives? avoidable mortality revisited. The Nuffield Trust. 2004.
    13.
    GBD 2016 Risk Factors Collaborators.  Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1345-1422. doi:10.1016/S0140-6736(17)32366-8PubMedGoogle ScholarCrossref
    14.
    GBD 2016 Causes of Death Collaborators.  Global, regional, and national age-sex specific mortality for 264 causes of death, 1980-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1151-1210. doi:10.1016/S0140-6736(17)32152-9PubMedGoogle ScholarCrossref
    15.
    GBD 2016 Disease and Injury Incidence and Prevalence Collaborators.  Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1211-1259. doi:10.1016/S0140-6736(17)32154-2PubMedGoogle ScholarCrossref
    16.
    Radley  DC, McCarthy D, Hayes  SL. 2018 Scorecard on state health system performance. Commonwealth Fund. May 3, 2018. Accessed December 3, 2018. https://www.commonwealthfund.org/publications/fund-reports/2018/may/2018-scorecard-state-health-system-performance
    17.
    Davern  M, Quinn  BC, Kenney  GM, Blewett  LA.  The American Community Survey and health insurance coverage estimates: possibilities and challenges for health policy researchers.   Health Serv Res. 2009;44(2 Pt 1):593-605. doi:10.1111/j.1475-6773.2008.00921.xPubMedGoogle Scholar
    18.
    Organization for Economic Cooperation and Development. Total public and primary private health insurance. Accessed September 18, 2020. http://stats.oecd.org/#
    19.
    GBD 2016 Mortality Collaborators.  Global, regional, and national under-5 mortality, adult mortality, age-specific mortality, and life expectancy, 1970-2016: a systematic analysis for the Global Burden of Disease Study 2016.   Lancet. 2017;390(10100):1084-1150. doi:10.1016/S0140-6736(17)31833-0PubMedGoogle ScholarCrossref
    20.
    Nolte  E, McKee  CM.  In amenable mortality—deaths avoidable through health care—progress in the US lags that of three European countries.   Health Aff (Millwood). 2012;31(9):2114-2122. doi:10.1377/hlthaff.2011.0851PubMedGoogle ScholarCrossref
    21.
    Levy  H, Meltzer  D.  The impact of health insurance on health.   Annu Rev Public Health. 2008;29(1):399-409. doi:10.1146/annurev.publhealth.28.021406.144042PubMedGoogle ScholarCrossref
    22.
    Dor  A, Umapathi  E. Health insurance and health. In Culyer  AJ, ed.  Encyclopedia of Health Economics. Elsevier; 2014.
    23.
    Currie  J, Gruber  J.  Saving babies: the efficacy and cost of recent changes in the Medicaid eligibility of pregnant women.   J Polit Econ. 1996; 104: 1263-1296. doi:10.1086/262059Google ScholarCrossref
    24.
    Currie  J, Gruber  J.  Health insurance eligibility, utilization of Medical care, and child health.   Q J Econ. 1996; 111(2):431–466. doi:10.2307/2946684Google ScholarCrossref
    25.
    Goldman  DP, Bhattacharya  J, McCaffrey  DF, Duan  N, Leibowitz  AA, Joyce  GF,  et al.  Effect of insurance on mortality in an HIV-positive population in care.   J Am Stat Assoc. 2001; 96(455):883–894. doi:10.1198/016214501753208582Google ScholarCrossref
    26.
    Bhattacharya  J, Goldman  D, Sood  N.  The link between public and private insurance and HIV-related mortality.   J Health Econ. 2003;22(6):1105-1122. doi:10.1016/j.jhealeco.2003.07.001PubMedGoogle ScholarCrossref
    27.
    Card  D, Dobkin  C, Maestas  N.  Does Medicare save lives?   Q J Econ. 2009;124(2):597-636. doi:10.1162/qjec.2009.124.2.597PubMedGoogle ScholarCrossref
    28.
    Andersen  MS.  Effects of Medicare coverage for the chronically ill on health insurance, utilization, and mortality: evidence from coverage expansions affecting people with end-stage renal disease.   J Health Econ. 2018;60:75-89. doi:10.1016/j.jhealeco.2018.06.002PubMedGoogle ScholarCrossref
    29.
    Chen  A, Oster  E, Williams  H.  Why is infant mortality higher in the United States than in Europe?   Am Econ J Econ Policy. 2016;8(2):89-124. doi:10.1257/pol.20140224PubMedGoogle ScholarCrossref
    ×