Marketplace Health Insurance Ratings: Most Potential Enrollees Have Access To Plans Of Medium Or High Quality

A key goal of the Affordable Care Act (ACA) is to improve access for all to health insurance coverage that is both affordable and of high quality. In 2020 more than eight million consumers in thirty-eight states obtained health coverage on the federally facilitated ACA Marketplace, and another three million obtained coverage on state-based Marketplaces.¹ Although health plan quality ratings were mandated in the ACA, delays in development prevented ratings covering a national set of health plans from being made publicly available until open enrollment began for the 2020 plan year.²

With the development of the ACA Marketplace Quality Rating System, quality ratings ranging from one to five stars now appear alongside listings of plans on HealthCare.gov, with the goal of empowering consumers to make better-informed health insurance decisions while also incentivizing improvement of health plan quality through public reporting. For the first time on the Marketplace, consumers have the opportunity to assess information on the cost and the quality of a plan to make informed enrollment decisions based on a plan’s overall value. These data also allow consumers to assess trade-offs among the breadth of coverage, cost of premiums, and quality of Marketplace plans.

The ACA Marketplace Quality Rating System rates plans from one to five stars by creating a composite score across forty-one quality measures within three domains: medical care, primarily measured by process measures; member experience, assessed by the Consumer Assessment of Healthcare Providers and Systems survey; and plan administration, judged on the basis of customer service, customers’ access to information, and appropriateness of care use.^3,4

With increasing enrollment in the Marketplace, it is unknown whether there is equitable access to plans from high-rated insurers.^5,6 If counties with a higher concentration of impoverished or racial and ethnic minority customers have access only to plans from lower-rated insurers, then underlying disparities in health care access could be exacerbated. Empirical data on whether disparities in access to high-quality insurers exist would be helpful for policy makers as they develop incentives and regulations to improve access to high-rated insurers.

In this study we used recently released ACA Marketplace Quality Rating System data to examine three main questions. First, what is the variation in quality star ratings across US counties, and, related, what are the characteristics of counties with access to plans from low-, medium-, and high-rated insurers? Second, is there a relationship between premiums and quality star ratings? Last, what factors are the predictors of county-level access to high-rated plans?

Study Data And Methods

Data

For plans offered on the federally facilitated ACA Marketplace, quality ratings from the plan year 2020 Nationwide Quality Rating System Public Use File (NatQRS-PUF) were merged with premium and actuarial value information from the plan year 2020 Qualified Health Plan Landscape Medical Individual Market File.⁷ The NatQRS-PUF contained data on thirty-eight states, but we included thirty-five states in our analyses based on the availability of quality ratings and premium data from the Quality Rating System Landscape and Public Use File. State-based Marketplaces using state-based platforms are not included in the plan-level NatQRS-PUF and were excluded from this analysis. State-based Marketplaces using the federal platform were aggregated with states using the federally facilitated Marketplace; this group of states is referred to as the federally facilitated Marketplace for brevity. County-level demographic and market characteristics were obtained from the 2020 Area Health Resources File from the Health Resources and Services Administration.⁸

Quality ratings, first released at the beginning of plan year 2020 open enrollment, were assigned only to plans that had been available in the Marketplace for at least three years and had at least 500 enrollees in the prior year. As we looked at the data, we found that 15,331 plans, or 28 percent of plans on the federally facilitated Marketplace, did not have available quality ratings in the plan year 2020 file. As a result of the COVID-19 pandemic, quality ratings were not released for plan year 2021 and were instead based on plan year 2020 ratings (performance data from 2019). Details of the measures included in the Quality Rating System are in online appendix exhibit 1.⁹ Quality ratings are assigned at the insurer level such that all plan types (health maintenance organization [HMO], preferred provider organization [PPO], and so on) offered by the same insurer have the same rating. Accordingly, in our analyses, references to low-, medium-, or high-rated plans therefore apply to plans offered by low-, medium-, or high-rated insurers.

Variables

Our main dependent variable was the quality rating (“star rating”) of the highest-rated insurer in each county on the federally facilitated ACA Marketplace. Within a rating area, insurers offer plans at the county level; therefore, we chose the county as the main unit of analysis. We chose the highest star rating in each county as our outcome as opposed to the average star rating, as we wanted to assess disparities in access to high-rated plans.

Our main predictors for the star rating of an insurer’s plan were actuarial value and premiums. Insurers calculate premiums on the basis of age, tobacco use, family composition, and the residence of the individual, which are defined for a given geographic unit known as community rating areas. Given that premiums vary by age and health status within each plan, we defined the plan premium as the premium for a twenty-seven-year-old nonsmoker before Advance Premium Tax Credits, in line with prior studies of ACA Marketplace premiums.^10,11 When an insurer offered multiple plan types (HMO, PPO, and so on), these plan types were considered separately, as they could have different actuarial values (also called metal levels). Actuarial values reflect the proportion of covered medical costs paid by an insurer. Actuarial values for the metal levels are platinum (90 percent), gold (80 percent), silver (70 percent), and bronze (60 percent).

Market-level factors included the number of rated insurers in each county, the percentage of county residents who were uninsured, and the percentage of county residents who obtained coverage in the nongroup market through the ACA federally facilitated Marketplace (hereafter referred to as the percentage of Marketplace enrollees) in each county. Health care supply variables included the total number of physicians and of hospital beds per 10,000 residents in each county. Socioeconomic factors included deep poverty, which was defined as the percentage of county residents with income less than half of the federal poverty threshold, and education, which was defined as the percentage of residents in the county with a college degree or higher. The racial composition variable was the percentage of Black residents in the county, and the ethnic composition variable was the percentage of Hispanic residents in the county. These variables were as defined in the Area Health Resources File and derived from census data; they were not altered for this analysis. We focused on Black and Hispanic residents based on sample size and on existing literature on disparities in insurance access for these populations.^12,13 The geographic variable was a dichotomous urban/rural county-level measure from the Medicare core-based statistical area definition.¹⁴

Analyses

We first aggregated plans by county and created a county-level map showing the variations of maximum star ratings across counties in the federally facilitated Marketplace. Then we looked at the characteristics of counties that had access to low-, medium-, and high-rated plans.

We next assessed the relationship between plan premiums and star ratings, using two plan-level linear regressions. In the first regression we assessed the change in premium associated with each unit increase in star rating, with a model in which rating was treated as a continuous predictor, adjusting for actuarial value, plan type, and out-of-pocket maximum. One-star plans were excluded from this regression because these plans were all offered by the same insurer, which was located in West Virginia. In the second regression we treated star rating as a categorical predictor and estimated the average premium at each star rating, adjusted for actuarial value, plan type, and out-of-pocket maximum.

Next we assessed predictors of county-level access to plans from high-rated insurers. We divided counties into three groups: counties with access to only plans from low-rated (one to two stars), those with access to medium-rated (three stars), and those with access to high-rated (four to five stars) insurers. We then assessed the distribution of each explanatory variable across these categories by calculating the median and interquartile range of each variable, stratified by access to low-, medium-, and high-rated insurers. Five small counties without available demographic information were excluded.

We then created a multivariate ordinal logistic regression model to assess the independent relationship of market-level, health care supply, socioeconomic, racial and ethnic composition, and geographic factors with the probability of having access to plans from high-rated insurers. The predictor in this regression was the highest-rated insurer offering plans in the county. To capture potentially nonlinear relationships between variables and star ratings, we categorized predictor variables as quartiles across all counties included in the analysis; the lowest quartile served as the reference group. The first, second, third, and fourth quartiles are referred to throughout as the lowest, low, high, and highest quartiles, respectively, for ease of presentation. The results are presented as adjusted odds ratios and 95% confidence intervals comparing each quartile with the reference group. On the basis of our ordinal logistic regression approach, the model estimates the adjusted odds ratio of access to a five-star plan compared with access to plans rated four stars or lower, or access to four- or five-star plans compared with access to plans rated three stars or lower, and so on for each unit increase of the relevant county-level market, health care supply, socioeconomic, and racial and ethnic composition factor. Our regression model was weighted by the county-level population. Because of concerns of multiple hypothesis testing with the variables in our multivariable model, we applied a Bonferroni correction, and a $p$ value less than 0.0045 was considered significant. Details of the model outputs are in appendix exhibit 2.⁹ All analyses were performed in R, version 3.6.2.

Sensitivity Analyses

We created additional county-level maps of insurer counts and plan premiums, using the second-lowest silver premium in a county for a twenty-seven-year-old nonsmoker as the benchmark premium. We performed a sensitivity analysis of the relationship between plan ratings and premiums by constructing new regression models that adjusted not only for metal level, plan type, and out-of-pocket maximum but also for rating-area fixed and random effects to assess between- and within-rating-area variations. We also performed a sensitivity analysis in which the outcome was the star rating of the second-lowest-cost silver plan in the county to coincide with the benchmark premium plan. However, as every insurer offered a silver plan in each county and the star ratings for all plan types offered by a given insurer are the same across actuarial values, these results were the same as those of our main analyses. Results from this sensitivity analysis therefore are not shown. Given that one- and five-star plans were offered in only one state each during plan year 2019, we reran our multivariable logistic regression excluding all one- and five-star plans to test whether this small subset of plans was skewing results. We also performed a sensitivity analysis with an interaction term between proportion of Black residents and rurality to assess for the interaction of race with rurality on access to high-rated health plans. Finally, to assess for potential biases in the relationship between rurality and plan ratings, we performed a sensitivity analysis interacting the rurality of the county with the insurer count.

Limitations

This study had several limitations. First, the plan landscape and star ratings data sets included only plans in the thirty-eight states that use the federally facilitated ACA Marketplace. These results might not extend to the individual Marketplaces managed by other states. State-based Marketplaces were more likely to have access to five-star insurers and less likely to have only one insurer, but overall the distribution of plans and insurers in these states appears similar to that in states on the federally facilitated Marketplace (appendix exhibit 3).⁹ Second, plans must be offered on the ACA Marketplace for at least three years before receiving a star rating. As such, newer plans are excluded from these analyses, and recent shifts in the characteristics of ACA Marketplace plans might not be captured. In addition, rural counties with few insurers may be disproportionately affected by missing quality ratings as a result of insurer exit and entry or not meeting enrollment thresholds for inclusion in the Quality Rating System.

Third, the Centers for Medicare and Medicaid Services (CMS) insurer file included a one-star insurer in New Mexico, but the plan landscape file did not include any plans from this insurer, and therefore it did not appear in our analyses. We are unaware of any other data discrepancies, but we were dependent on CMS for the completeness and validity of the data we analyzed. Fourth, actuarial values reflect a range of values within a given metal level, but for ease of presentation, we categorized the actuarial value by the metal level. Fifth, we chose to display premiums before application of Advance Premium Tax Credits (“subsidies”) to make our results more generalizable without assumptions of subsidy amounts. Therefore, the effect size of the relationship between plan rating and premium may be higher than that experienced by enrollees after subsidies are applied. Sixth, because individual-level data on plan selections are not publicly available, we could not assess the actual variation in plan enrollment by race or ethnicity. Last, our investigation of racial and ethnic disparities was limited to Hispanic and Black residents.

Study Results

Geographic Variation In Star Ratings Across Counties

A total of 38,562 plans with ratings on the federally facilitated ACA Marketplace were offered in 2,265 counties in thirty-five states. In most counties (1,390, or 61.4 percent) the highest-rated plans were three-star plans. Fewer than one-third of counties (703, or 31.0 percent) had access to plans from insurers with four- or five-star ratings. Fewer than 10 percent (172, or 7.6 percent) had access to only one- or two-star-rated insurers. About half (50.5 percent) of Marketplace enrollees lived in counties where the highest-rated insurer offering plans had a three-star rating, although 46.0 percent had access to at least one insurer offering four- or five-star-rated plans (data not shown).

Counties within a state or region tended to have access to plans from insurers of similar quality (exhibit 1). The only insurer in West Virginia had a one-star rating, and only Virginia had plans from a five-star insurer. The benchmark plan premiums and number of Marketplace insurers per county also varied predominantly across states rather than within states (appendix exhibit 4).⁹

SOURCE Authors’ analysis of data from the plan year 2020 Nationwide Quality Rating System Public Use File, Centers for Medicare and Medicaid Services. NOTE Data reflect states in the federally facilitated Marketplace as well as state-based Marketplaces using the federal platform in 2020.

Characteristics Of Counties By Access To Low-, Medium-, And High-Rated Plans

Plans from high-rated insurers tended to be in more populous counties than those from medium- and low-rated insurers; the median population of a county with access to high-rated plans was 32,500 residents compared with 28,200 and 18,100 residents for medium- and low-rated plans, respectively. The number of insurers offering plans in a county was strongly associated with access to high-rated plans: 65.6 percent of counties with access to high-rated plans had three or more insurers, whereas just 6.6 percent of counties with access only to low-rated plans had three or more insurers. Counties with access to high-rated plans were associated with greater health care supply, with a median of twenty-five physicians per 10,000 residents and sixty-two hospital beds per 10,000 residents compared with a median of eleven physicians and forty-three hospital beds per 10,000 residents in counties with access only to low-rated plans. There did not appear to be a monotonic relationship between star ratings and the percentage of county residents in deep poverty or the percentage of residents with college degrees (exhibit 2).

Counties with access only to low-rated plans had a lower median percentage of Black and Hispanic residents, but there was not a clear monotonic relationship across counties with access to low-, medium-, and high-rated plans. Counties with access to only to low-rated plans were predominantly rural: 73.3 percent of such counties were rural compared with 56.9 percent of counties with access to high-rated plans.

Association Of Star Ratings And Premiums

On average, high-rated plans tended to have higher monthly premiums (exhibit 3). Excluding plans in West Virginia, which had a single insurer with only one-star plans, each increase in plan quality star rating was associated with a $27.69 (95% CI: $26.28, $29.11, $p<0.001$) increase in average monthly premium (appendix exhibit 5).⁹ This relationship was similar after we adjusted for rating-area fixed and random effects ($24.36 [$p<0.001$] and $24.50 [$p<0.001$], respectively) (appendix exhibit 5).⁹ One-star plans, available exclusively in West Virginia, deviated from this trend with a high average monthly premium of $557 (data not shown).

SOURCE Authors’ analysis of data from the plan year 2020 Nationwide Quality Rating System Public Use File, Centers for Medicare and Medicaid Services. NOTES Adjusted premiums for each star-rating group are from a multivariable linear regression model with controls for actuarial value, plan type, and out-of-pocket maximum. Star rating refers to the plan’s quality star ratings.

Independent Predictors Of Access To High-Rated Plans

Using a multivariable ordinal logistic regression, we assessed the probability of a county having access to plans from high-rated insurers. There were significant but not monotonic associations between access to high-rated plans and median hospital beds per 10,000 residents and between access to high-rated plans and median percentage of residents with a college degree. There were no significant and monotonic associations between access to high-rated plans and other socioeconomic factors, other health care supply factors, or rurality. Among racial and ethnic composition variables, only the percentage of Black residents had a monotonic relationship with access to high-rated plans. Counties in the highest quartile with respect to the percentage of Black residents had four times higher odds of access to plans from high-rated insurers than those in the lowest quartile (adjusted odds ratio: 4.0; $p<0.001$). Among market factor variables, the percentage of Marketplace enrollees and the number of rated insurers in a county had a monotonic relationship with access to high-rated plans. Counties in the highest quartile with respect to the percentage of Marketplace enrollees had seventeen times higher odds of access to high-rated plans than those in the lowest quartile (AOR: 17.7; $p<0.001$). Counties with more insurers also had substantially higher odds of access to high-rated plans, with two-insurer counties having more than fourteen times higher odds as one-insurer counties and counties with three or more insurers having nineteen times higher odds (AORs: 14.6 and 19.0, respectively; $p<0.001$) (exhibit 4).

SOURCE Authors’ analysis of data from the plan year 2020 Nationwide Quality Rating System Public Use File, Centers for Medicare and Medicaid Services, and from the Area Health Resources File, Health Resources and Services Administration, 2020. NOTES Results are from a multivariable ordinal logistic regression model with additional controls for percent of county uninsured, supply of per capita physicians and hospital beds, percent of county that is Hispanic, percent of county in deep poverty, percent of county residents with college degrees, and rurality of county. Model results are weighted by the county population. “Percent of Marketplace enrollees” refers to the percent of county residents who obtained coverage in the nongroup market through the federally facilitated ACA Marketplace. Independent variables were categorized as quartiles across all counties included in the analysis, and the lowest quartile served as the reference group. Lowest, low, high, and highest quartiles refer to the first, second, third, and fourth quartiles, respectively. Based on the ordinal logistic regression approach, the adjusted odds ratios reflect the odds of access to the highest-rated plan in a given county compared with the odds of access to lower-rated plans. For all reference values, the odds ratio is 1. The whiskers represent 95% confidence intervals. All results are statistically significant based on application of the Bonferroni correction ($p<0.0045$).

Sensitivity Analyses

The trend of higher premiums for high-rated plans remained largely intact when we stratified by actuarial values, although there was slight variation from the pattern for gold and platinum plans (appendix exhibit 6).⁹ Results of ordinal logistic regression were unchanged when we excluded all one- and five-star plans and when we conducted analyses at the rating-area level (appendix exhibits 7 and 8).⁹ Both urban and rural counties with a greater percentage of Black residents were associated with higher odds of access to high-rated plans (appendix exhibit 9).⁹ Last, we found a significant association between rurality and the number of insurers where access to more insurers was associated with improved access to high-rated plans in rural areas more than in urban areas ($p<0.001$) (appendix exhibit 10).⁹

Discussion

In this analysis of recently released ACA Marketplace data, we found that residents in most counties had access to plans from medium- and high-rated insurers in plan year 2020. Counties that were more populous, had a greater number of insurers, and had a greater percentage of Marketplace enrollees were more likely to have access to plans from high-rated insurers. High-rated plans were associated with higher premiums. Overall, we found no disparities in access to high-rated plans in counties with a higher percentage of Black or Hispanic residents. Taken together, these results suggest that medium- and high-rated plans are widely available on the federally facilitated ACA Marketplace, but consumers will need to consider the potential trade-offs of quality versus premium cost.

Counties with multiple insurers had higher odds of access to high-rated plans, suggesting that increasing the number of insurer options available to consumers in the ACA Marketplace could play a valuable role in improving the quality of Marketplace plans. Given that in plan year 2021 an additional thirty insurers entered the ACA Marketplace and 1,207 counties (38 percent) gained one insurer, access to high-rated plans might increase.¹⁵

Our study suggests a few important implications for policy makers. First, the release of the quality star ratings provides an important opportunity to monitor the quality of plans offered on the ACA Marketplace. The current iteration of the ACA Marketplace Quality Rating System is primarily based on process measures of clinical quality, patient satisfaction, and administrative measures of efficiency. Given the predominance of three-star plans, more sensitive measures of quality may be needed if star ratings are to truly aid consumers in selecting plans of higher quality. Opportunities to incorporate metrics on clinical outcomes, access, and affordability from the consumer perspective in the Quality Rating System should therefore be considered. Potential affordability metrics may include rates of claims denials or financial toxicity to consumers in the form of catastrophic health expenditures and medical bankruptcy.

Second, there is a trade-off between plan affordability and quality. Research is needed on whether new consumers are more likely to select plans based on quality rating or premium cost and whether consumers switch plans during open enrollment based on quality rating or premium cost. To incentivize plan selection based on quality ratings, policy makers could employ choice architecture by displaying high-rated plans first on HealthCare.gov.

Third, efforts should be made by policy makers to ensure that consumers have access to an adequate supply of insurance offerings to be able to choose plans based on quality. This could be accomplished by offering financial incentives for insurers entering a new market or imposing penalties on insurers exiting a market. Last, given the predominance of high-rated insurance plans in more populous counties, policy makers should carefully monitor the quality of insurance offerings in less populous or rural counties.

Conclusion

In the first analysis of quality star ratings for plans on the federally facilitated ACA Marketplace, we found that most potential Marketplace enrollees had access to plans of medium or high quality in plan year 2020. However, there may be trade-offs between higher-rated plans and premium costs to Marketplace enrollees. Overall, counties with three or more insurers and those with the highest percentage of county residents obtaining individual coverage through the ACA Marketplace were most likely to have access to plans from high-rated insurers. Importantly, there does not appear to be any evidence of reduced access to high-rated plans among Black and Hispanic consumers. Policy makers should explore additional options to encourage insurers to enter the ACA Marketplace, which remains an important source of nongroup insurance coverage.

ACKNOWLEDGMENTS

Thomas Tsai receives funding for work unrelated to this manuscript from the Commonwealth Fund, Arnold Ventures, the Massachusetts Consortium on Pathogen Readiness underwritten by the Massachusetts Life Sciences Center, and the William F. Milton Fund of Harvard University. E. John Orav receives funding for work unrelated to this manuscript from the Department of Health and Human Services Office of the Assistant Secretary for Planning and Evaluation and the National Institute on Aging, National Institutes of Health. Arnold Epstein receives funding for work unrelated to this manuscript from the New England Journal of Medicine and the National Institute on Aging, National Institutes of Health. Views expressed in this article represent the views of the authors and do not represent the official views of the US government.

NOTES

• 1 Centers for Medicare and Medicaid Services. Health insurance exchanges 2020 open enrollment report [Internet]. Baltimore (MD): CMS; 2020 Apr 1 [cited 2022 Jan 3]. Available from: https://www.cms.gov/files/document/4120-health-insurance-exchanges-2020-open-enrollment-report-final.pdf Google Scholar
• 2 Centers for Medicare and Medicaid Services. Quality rating information bulletin [Internet]. Baltimore (MD): CMS; 2019 Aug 15 [cited 2022 Jan 3]. Available from: https://www.cms.gov/CCIIO/Resources/Regulations-and-Guidance/Downloads/QualityRatingInformationBulletinforPlanYear2020.pdf Google Scholar
• 3 Centers for Medicare and Medicaid Services. Quality Rating System and Qualified Health Plan Enrollee Experience Survey: technical guidance for 2020 [Internet]. Baltimore (MD): CMS; 2019 Oct [cited 2022 Jan 3]. Available from: https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/QualityInitiativesGenInfo/Downloads/QRS-and-QHP-Enrollee-Survey-Technical-Guidance-for-2020-508.pdf Google Scholar
• 4 Centers for Medicare and Medicaid Services. Fact sheet: health insurance exchange quality ratings system 101 [Internet]. Baltimore (MD): CMS; 2019 Aug 15 [cited 2022 Jan 3]. Available from: https://www.cms.gov/newsroom/fact-sheets/health-insurance-exchange-quality-ratings-system-101 Google Scholar
• 5 Corallo B, Mehta A. Analysis of recent national trends in Medicaid and CHIP enrollment [Internet]. San Francisco (CA): Henry J. Kaiser Family Foundation; 2021 Oct 29 [cited 2022 Jan 3]. Available from: https://www.kff.org/coronavirus-covid-19/issue-brief/analysis-of-recent-national-trends-in-medicaid-and-chip-enrollment/ Google Scholar
• 6 Keith K. Record high ACA enrollment at 31 million Americans. Health Affairs Blog [blog on the Internet]. 2021 Jun 7 [cited 2022 Jan 3]. Available from: https://www.healthaffairs.org/do/10.1377/forefront.20210607.367870/full/ Google Scholar
• 7 Centers for Medicare and Medicaid Services. Health insurance exchange Public Use Files (Exchange PUFs) [Internet]. Baltimore (MD): CMS; 2020 [cited 2022 Jan 3]. Available from: https://www.cms.gov/CCIIO/Resources/Data-Resources/marketplace-puf Google Scholar
• 8 Health Resources and Services Administration. Area Health Resources File [Internet]. Rockville (MD): HRSA; 2021 Jul 31 [cited 2022 Jan 3]. Available from: https://data.hrsa.gov/topics/health-workforce/ahrf Google Scholar
• 9 To access the appendix, click on the Details tab of the article online.
• 10 Boozary AS, Feyman Y, Reinhardt UE, Jha AK. The association between hospital concentration and insurance premiums in ACA Marketplaces. Health Aff (Millwood). 2019;38(4):668–74. Go to the article, Google Scholar
• 11 Jacobs PD, Banthin JS, Trachtman S. Insurer competition in federally run Marketplaces is associated with lower premiums. Health Aff (Millwood). 2015;34(12):2027–35. Go to the article, Google Scholar
• 12 Lipton BJ, Decker SL, Sommers BD. The Affordable Care Act appears to have narrowed racial and ethnic disparities in insurance coverage and access to care among young adults. Med Care Res Rev. 2019;76(1):32–55. Crossref, Medline, Google Scholar
• 13 Courtemanche C, Marton J, Ukert B, Yelowitz A, Zapata D, Fazlul I. The three-year impact of the Affordable Care Act on disparities in insurance coverage. Health Serv Res. 2019;54 Suppl 1(Suppl 1):307–16. Crossref, Medline, Google Scholar
• 14 Long JC, Delamater PL, Holmes GM. Which definition of rurality should I use?: The relative performance of 8 federal rural definitions in identifying rural-urban disparities. Med Care. 2021;59(Suppl 5):S413–9. Crossref, Medline, Google Scholar
• 15 McDermott D, Cox C. Insurer participation on the ACA Marketplaces, 2014–2021 [Internet]. San Francisco (CA): Henry J. Kaiser Family Foundation; 2020 Nov 23 [cited 2022 Jan 3]. Available from: https://www.kff.org/private-insurance/issue-brief/insurer-participation-on-the-aca-marketplaces-2014-2021/ Google Scholar