|Home | About | Journals | Submit | Contact Us | Français|
In 2010, the United States (US) passed health insurance reforms aimed at expanding coverage to the uninsured. Yet, disparities persist in access to health care services, even among the insured.
To examine the separate and combined association between having health insurance and/or a usual source of care (USC) and self-reported receipt of health care services.
Two-tailed, chi-square analyses and logistic regression models were used to analyze nationally representative pooled 2002–2007 data from the Medical Expenditure Panel Survey (MEPS).
US adults (≥18 years of age) in the MEPS population who had at least one health care visit and who needed any care, tests, or treatment in the past year (n=62,067).
We assessed the likelihood of an adult reporting unmet medical needs; unmet prescription needs; a problem getting care, tests, or treatment; and delayed care based on whether each individual had health insurance, a USC, both, or neither one.
Among adults who reported a doctor visit and a need for services in the past year, having both health insurance and a USC was associated with the lowest percentage of unmet medical needs, problems and delays in getting care while having neither one was associated with the highest unmet medical needs, problems and delays in care. After adjusting for potentially confounding covariates (age, race, ethnicity, employment, geographic residence, education, household income as a percent of federal poverty level, health status, and marital status ), compared with insured adults who also had a USC, insured adults without a USC were more likely to have problems getting care, tests or treatment (adjusted relative risk [aRR] 1.27; 95% confidence interval [CI] 1.18–1.37); and also had a higher likelihood of experiencing a delay in urgent care (aRR 1.12; 95% CI 1.05–1.20).
Amidst ongoing health care reform, these findings suggest the important role that both health insurance coverage and a usual source of care may play in facilitating individuals’ access to care.
With the recent passage of one of the most monumental health insurance reform policies in US history, energies are now focused on expanding health insurance coverage to millions of Americans. These efforts are important due to the overwhelming evidence that stable health insurance coverage is associated with more consistent access to health care services, which contributes to better outcomes.1–4 Yet, even among insured persons, there are disparities in access to care and the quality of care received.5–7 Insurance coverage is often necessary to access care, but not always sufficient,8–11 especially if insured individuals have no place to obtain care.12,13 Approximately 19% of all US adults are without a usual source of care (USC), and 53% of uninsured adults have no USC.14–16 Safety net services are oversubscribed and the supply of primary care services is widely disparate across the country,17,18 leaving many Americans with few options to find and maintain a stable USC. This is further challenged by the amount of influence payers exert over the choices available to their clients, often creating restrictive provider networks and financial penalties for patients seeking a USC outside the network.19,20
The emphasis on insurance coverage in recent policies coupled with primary care workforce shortages prompted us to revisit the question of how much influence insurance and a USC have on receipt of needed health care services. Past studies have examined health insurance or a USC individually, usually controlling for the other factor. The few studies that have combined the two predictors have usually focused exclusively on a single outcome or a specific population.21–25 The primary objectives of this study were (1) to describe the prevalence of and characteristics associated with having insurance coverage and/or a USC; and (2) to ascertain the separate and combined association between having health insurance and/or a USC and self-reported access to health care services for a nationally representative population of adults who reported having a need for services.
We analyzed data from the Medical Expenditure Panel Survey-Household Component (MEPS-HC).26 The MEPS-HC utilizes a stratified and clustered random sample from National Health Interview Survey households with weights that produce nationally representative estimates for the civilian, non-institutionalized US population.27 MEPS-HC respondents are interviewed 5 times over a 2-year period, with an overlapping panel design; annual public use data files contain data from a single year for 2 consecutive panels. Each year of data constitutes a nationally representative sample, and pooling the data produces average annual estimates. Certain groups (e.g., low income, racial minorities) are oversampled.
We combined data from the 2002 through 2007 annual public use data files, because these 6 years have a common variance structure necessary to ensure compatibility and comparability of our variables within the complex sample design. We first included all adults in the MEPS-HC with positive full-year weights and known information about USC and full-year insurance status (n=134,714, which excluded 1,975 individuals without insurance and/or USC information). This sample is representative of a population of nearly 215.8 million. We then limited the analyses of unmet needs to the adult MEPS-HC subpopulation (≥18 years of age) who had at least one health care visit in the previous 12 months and who reported that they had a need for additional care, tests, or treatment (n=62,067) in order to more accurately assess unmet health care needs.
We identified MEPS-HC items that highlighted self-reported access to care and quality of care in the past 12 months among adults, including: (1) unmet medical needs, (2) unmet prescription needs, (3) problems getting care, tests, or treatment, and (4) delayed urgent care. Unmet medical needs were defined as whether the person was unable to receive medical treatment or was delayed in receiving medical treatment. Unmet prescription needs was defined as whether the person was unable to receive prescription treatment or was delayed in receiving prescription treatment. Problems getting care, tests, or treatment was defined by whether or not it was a problem to get care, tests, or treatment that the person or a doctor believed necessary. Delayed urgent care was defined by whether or not the person always got care for an illness, injury or condition as soon as wanted if they had an illness, injury or condition that needed care right away.26
We assessed whether an individual had health insurance coverage and/or a USC at the same point during the year. MEPS ascertains USC status by asking whether there is a particular doctor’s office, clinic, health center, or other place (including emergency departments) where the individual usually goes when he/she is sick or needs advice about his/her health.26 We then created four subgroups, including: (1) insured and had a USC (Yes INS/Yes USC); (2) insured but no USC (Yes INS/No USC); (3) not insured but had a USC (No INS/Yes USC); and (4) not insured and no USC (No INS/No USC).
We used the conceptual model designed by Aday and Andersen to guide the identification of nine additional covariates that might influence access to care, including: age, race, ethnicity, employment, geographic residence, education, household income as a percent of federal poverty level, health status, and marital status.28–30 In two-tailed, chi-square analyses, each covariate was significantly associated with at least one outcome (p<0.10); thus, all covariates were included in logistic regression models.
We determined the prevalence of patterns of insurance coverage and/or a USC for the entire population (n=134,714); we then limited the analysis to the subpopulation of adults (≥ 18 years of age) who had seen a clinician at least once in the previous 12 months and who reported that they had a need for additional care, test, or treatment (n=62,067). We then used chi-square analyses and logistic regression models to examine characteristics associated with having no insurance (versus health insurance) and no USC (versus having a USC). Next, we used chi-square analyses to assess significant socio-demographic differences among the INS/USC subgroups. Finally, we used a series of standard logistic regression models to compare the adjusted relative risks among the INS/USC subgroups for each of the outcomes. Model-building was done first for the unmet medical needs outcome using backward selection. Covariates selected in this model were then used for all remaining outcomes to achieve consistency in our examination, and to facilitate comparisons of results across models. We reported measures of association as risk ratios because when an outcome is common (e.g. prevalence of >10% in a population), odds ratios do not accurately approximate the risk ratio.31
We used SUDAAN Version 10.0.1 for all statistical analyses to account for the complex sampling design of the MEPS; all comparisons utilized a two-tailed p <0.05 to define statistical significance. This study protocol was deemed exempt from review by the Oregon Health and Science University Institutional Review Board because MEPS-HC data are publicly available.
An estimated 77.2% of the US adult population had a USC; and 80.8% had health insurance. Only 68.0% of US adults had both a USC and insurance, while 10.0% had neither one (Table 1). Among adults in the subgroup reporting at least one health care visit and a need for additional services, a higher percentage had health insurance and a USC.
In many cases, demographic patterns were similar when comparing those with no insurance to those with no USC (Table 2). For example, adults younger than 45 were more likely to be in the group without insurance and without a USC, as compared to those between the ages of 45–64, while adults older than 65 were much less likely to be in either group. There were some examples, however, where demographic patterns varied among adults with no insurance versus those with no USC. For example, unemployed adults were more likely to have no insurance (aRR 1.22; 95% CI 1.14–1.31) but less likely to have no USC (aRR 0.86; 95% CI 0.80–0.92), as compared with employed adults. Similarly, compared to those reporting “excellent” health status, those reporting their health status as “good” were more likely to have no insurance (aRR 1.20; 95% CI 1.10–1.30) but less likely to have no USC (aRR 0.75; 95% CI 0.69–0.81).
Demographic characteristics varied when comparing the INS/USC subgroups (Table 3). For example, Hispanic adults were only 8.2% of the total sample, but represented 25.0% of the group without both insurance and a USC. Of the 15.0% of adults in the subgroups with either insurance alone or a USC alone, a higher percentage were Hispanic, lived in the Southern region of the US, and/or were from lower earning households .
Among US adults in this study population, having health insurance and a USC was associated with the lowest percentage of unmet medical needs, problems and delays in getting care while having neither one was associated with the highest percentage of these three outcomes (Table 4). Unmet prescription needs appeared to be more strongly associated with a lack of health insurance as compared to the other outcomes. The univariate associations between receipt of health care services and having both health insurance and a USC remained strong in multivariate analyses.
When comparing the two middle subgroups (Yes INS/No USC and No INS/Yes USC) to the reference group (Yes INS/Yes USC), having either insurance only or a USC only was associated with higher rates of problems getting care and delayed urgent care as compared with having both insurance and a USC. For unmet medical needs, the insured without a USC were no more or less likely to report differences in access to treatment than the group with both insurance and a USC. And, those with insurance but no USC were actually less likely to report unmet prescription needs (aRR=0.83; 95% CI=0.70–0.99). The subgroup with no insurance and no USC had a higher likelihood of reporting all unmet need outcomes when compared to the reference group with insurance and a USC.
This study addressed the separate and combined association between insurance and/or a USC and access to health care for adults in the US. The uninsured without a USC were at highest risk for not receiving needed services. In many cases, having only insurance or only a USC was associated with higher rates of unmet needs as compared to having both. In addition to the differences noted among the INS/USC subgroups, there were consistent patterns of increased vulnerability among those in all income categories below 400% FPL and among those reporting less than “excellent” health status. Those over age 65 were less likely to experience unmet medical needs, problems getting care and delayed care, than those in the reference age group between 18–24 years of age. This confirms previous work which has found that receipt of unmet health care needs, and many disparities in basic cardiovascular risk measures, decline for those over age 65.32–37 It has been reported that these declines are largely due to many uninsured adults obtaining Medicare at age 65. However, we controlled for insurance status, so there may be an additional protective effect from Medicare insurance as compared to another type of insurance. In addition, perhaps, those in this age group may be more proficient at navigating the health care system to get their needs met or they may be less likely to report an unmet need.
As shown in Table 2, those without insurance were more likely to not have a USC and vice versa; however, Table 3 shows that 15% of adults have either insurance or a USC but not both, highlighting that access to one does not guarantee access to the other. This suggests that covering millions more with stable health insurance will not ensure that everyone has a USC. In fact, some studies propose that expanding eligibility for public insurance programs or mandating individual coverage, without a mechanism to ensure adequate provider capacity, will merely result in more covered Americans with nowhere to go for care.13,38–40 Alternatively, expanding the number of community health centers may bolster the capacity of the safety net in order to improve access to a USC, but leave thousands without insurance to cover necessary prescription medications, tests, referrals and ancillary services.41 Concurrent with health insurance reforms, there is a separate need to focus on bolstering the US health system’s capacity to provide a stable usual source of care.
Further, even individuals with financial access (i.e. insurance) and structural access (i.e. a USC) do not always receive care.5,8,12 For example, in this study, 5.4% to 37.1% of adults in the Yes INS/Yes USC group reported having at least one of the unmet health care needs. This suggests that having insurance and a USC provides potential access, but the degree of synergy between them dictates whether realized access is achieved (Fig. 1).
Figure 1 illustrates the difference between the potential for care and the reality of care. An individual can have insurance, which provides potential financial access; but if he or she cannot find a primary care physician within the insurance network, care cannot be realized. Alternatively, if structural access is available but the care is unaffordable (i.e. lack of financial access), then the care is not realized. In either case, potential access is not real access until both financing and delivery are coordinated and consistently available. Figure 1 demonstrates the important overlap that must exist to make access a reality—only at the confluence of the two circles. Thus, unmet need does not mean having no access; rather, unmet need is the difference between having potential and realized access (the outer, non-overlapping areas of each circle in Figure 1). This model might help to explain why over one-third of adults with insurance and a USC reported at least one unmet need in this study. If someone has financial and structural access (potentially) but the circles in Figure 1 have little or no overlap, then they are likely to experience unmet needs.
There are exceptions. Some individuals who lack potential access or who have minimal overlap do realize care. For example, a patient’s USC might agree to continue seeing her even though she has lost insurance and no longer has financial access. Or, a patient who has changed insurance carriers may pay out-of-pocket to continue receiving services from a trusted USC outside of his network (although this is rare and usually results in patient’s reducing contact with their USC).19
“Patient-centered medical homes” might help to improve synergistic relationships between the financing and delivery of primary health care, especially if payment mechanisms are transformed to facilitate comprehensive and integrated care.42–49 Innovators have made headway in defining and demonstrating medical home models.44,45,50–53 However, the number of US medical school graduates entering primary care professions has been in a rapid decline, which casts doubt on the US health care system’s ability to provide a basic USC, whether a medical home or not, to all newly insured persons.54 Thus, it remains to be seen whether medical home efforts to improve quality and performance will increase capacity and ensure continuous access to a USC.
Secondary data analyses are limited to existing data. For example, MEPS-HC data are available through 2007, so we were not able to ascertain the effects of the recent economic downturn. We reported on cross-sectional measures for both insurance and a USC because we did not have data to capture longitudinal USC status; therefore, we could not capture the effects of duration of insurance and/or continuity with a USC.55–57 We also did not account for the type of USC provider, which might contribute to subtle differences that were not measured in this study.58 Also, we did not include a full analysis of the specific reasons people reported a lack of a USC; however, upon review of the main reasons we found the top reported to be that respondents were seldom or never sick and the second most common was that the cost of medical care was too high. Since we used a subgroup that included people who had seen a clinician at least once in the past year and who reported that they had a need for care, it is unlikely that these reasons for a lack of USC would change our results. As with all studies that rely on self-report, response bias remains a possibility.
Although the MEPS-HC is representative of the civilian, non-institutionalized US population, the observational nature of the data limits causal inferences. We aimed to achieve consistency in our examination; thus, we included the same covariates across all logistic regression models. We secondarily assessed associations with other covariates, but did not build individual models for a comprehensive examination of each covariate.
Finally, we recognize that every state has unique insurance programs, and the availability of services varies widely by region. While we could not account for the willingness of providers to care for underserved populations or the availability of safety net services in every region, the multivariable analyses did include a MEPS-HC geographic region variable, which would be considered a crude proxy for some of these variations.
Unless policy reforms are well-coordinated, the best we can expect is that an incremental two-pronged approach will continue—investment in safety net services for some and expansions in health insurance coverage for others. This study suggests that neither one of these two approaches to expand access displaces the need for the other; thus, future policies should focus on ensuring that those with insurance coverage also have a usual source of care. Further, having both health insurance and a USC are necessary but not always sufficient to ensure receipt of services. Greater synergy between these financing and delivery aspects of care will move us even closer to achieving optimal access to care.
We also wish to acknowledge the thousands of individuals who participated in the MEPS.
Funding Sources This project was directly supported by grants 1 K08 HS16181 and 1 R01 HS018569 from the Agency for Healthcare Research and Quality (AHRQ) and the Oregon Health & Science University Department of Medicine. This publication received indirect support, from the Oregon Clinical and Translational Research Institute (OCTRI), grant number UL1 RR024140 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research.These funding agencies had no involvement in the design and conduct of the study; analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. AHRQ collects and manages the Medical Expenditure Panel Survey.
Presentations An abstract with similar findings was presented at the North American Primary Care Research Group Annual Meeting in Seattle, Washington, November 2010.
Conflicts of Interest None disclosed.