0
We're unable to sign you in at this time. Please try again in a few minutes.
Retry
We were able to sign you in, but your subscription(s) could not be found. Please try again in a few minutes.
Retry
There may be a problem with your account. Please contact the AMA Service Center to resolve this issue.
Contact the AMA Service Center:
Telephone: 1 (800) 262-2350 or 1 (312) 670-7827  *   Email: subscriptions@jamanetwork.com
Error Message ......
Caring for the Uninsured and Underinsured |

Ability to Obtain Medical Care for the Uninsured:  How Much Does It Vary Across Communities? FREE

Peter J. Cunningham, PhD; Peter Kemper, PhD
[+] Author Affiliations

From the Center for Studying Health System Change, Washington, DC.


JAMA. 1998;280(10):921-927. doi:10.1001/jama.280.10.921.
Text Size: A A A
Published online

Context.— Communities differ in the way that medical care for medically indigent persons is organized and delivered, which is likely to result in differences across communities in the ability of uninsured persons to obtain medical care. Changes in the health care system, many of which are driven locally, may further exacerbate these differences.

Objective.— To examine the extent of variation across US communities in the ability of uninsured persons to obtain medical care and the extent to which health status and other characteristics of the uninsured population account for these differences.

Design.— Analysis of the 1996-1997 Community Tracking Study Household Survey.

Setting.— A nationally representative sample of the US civilian, noninstitutionalized population residing in 60 randomly selected communities. Larger sample sizes were obtained for 12 of these communities, which were randomly selected to represent metropolitan areas in the United States with more than 200000 persons.

Participants.— A total of 60446 individuals and 7200 uninsured persons.

Main Outcome Measures.— The percentage of persons who either did not obtain needed medical care in the previous year or postponed receiving needed medical care in the previous year.

Results.— Differences between communities with the highest percentage of uninsured persons reporting difficulty obtaining care and communities with the lowest percentage were more than 2-fold (41.4% vs 18.5%, P<.05). Little of the variation across communities is accounted for by differences in health status or sociodemographic characteristics of the uninsured population. The pattern of variation across communities in the ability of uninsured persons to obtain medical care is not correlated with variations in the ability of privately insured persons to obtain care (Pearson r, 0.04). Simulation results indicate that expanding private or public insurance coverage would not only increase the ability of uninsured persons to obtain medical care but would also reduce the variation across communities.

Conclusions.— If people are uninsured, where they live is an important factor in determining the difficulty they have in obtaining care. This is likely to persist given that care for uninsured persons is driven largely by state and local policy, and health system changes are constraining clinicians' ability and willingness to serve uninsured persons in many parts of the country.

WITH THE FAILURE to enact legislation that would provide for universal health care coverage, the uninsured will continue to rely on the direct delivery of services through the health care safety net and other sources of care. The system of providing care to the uninsured varies considerably across the country and is driven largely by state and local policies that affect the organization and financing of the health care safety net, as well as voluntary efforts on the part of individual clinicians and community organizations.1,2 These differences may become even greater in the future as health care clinicians in many parts of the country are finding it more difficult to care for the uninsured because of reduced public subsidies for uncompensated care, greater difficulty in shifting the costs of care for the medically indigent onto other payers, increasing competition for paying patients, and increases in the number of uninsured persons. The result is that the ability of uninsured persons to obtain medical care is likely to vary substantially depending on where they live.

Previous research has focused primarily on documenting the access problems of the uninsured relative to the insured. These studies have convincingly shown that, relative to insured persons, uninsured persons are much more likely to have unmet health needs, are less likely to have a usual source of care, have lower health care use, and experience worse health outcomes, including increased mortality.37 However, very few studies have examined the extent to which access to care varies among the uninsured, and virtually nothing is known about how access to care viewed from the perspective of uninsured persons varies across communities.

The purpose of our study is to examine the extent of variation across communities in the ability of uninsured persons to obtain medical care and whether this variation is accounted for by differences across communities in the characteristics of uninsured persons, such as the need for health care and other factors related to the demand for care. Remaining variation after controlling for these factors would strongly suggest that other characteristics of the community, including health system factors, account for the variation in the ability of uninsured persons to obtain medical care.

Our study also examines whether the pattern of variation across communities for the uninsured is similar to that of the insured population, which would indicate whether the difficulties encountered by the uninsured reflect more generic problems of health care in the community that cannot be addressed simply by providing health insurance to the uninsured. We also simulate what the variation in access to care across communities would look like if all uninsured persons were given health insurance coverage.

Source of Data

The data for this study are from the Community Tracking Study (CTS) Household Survey. The CTS is a major initiative of the Robert Wood Johnson Foundation, Princeton, NJ, to track changes in the health care system over time and to gain a better understanding of how health care system changes are affecting people. To this end, the CTS includes site visits to selected communities to discuss health system changes with local representatives of employers, government, consumer groups, hospitals, physician groups, and health plans and nationally representative surveys of households, physicians, and employers. Data collection is to be repeated on a 2-year cycle. Although nationally representative, data collection is focused on 60 randomly selected communities or sites nationwide.8 Twelve of these sites were randomly selected to serve as case study sites, which involved the site visits described herein as well as considerably larger survey samples.

The selection of the 60 sites formed the first stage of sample selection for the household survey.9 Sites were defined as counties or groups of counties using conventionally accepted definitions of statistical and economic areas. Specifically, metropolitan statistical areas (MSAs) as defined by the US Office of Management and Budget, Washington, DC, were used to define metropolitan areas, whereas the nonmetropolitan portion of the Bureau of Economic Analysis economic areas was used to define nonmetropolitan sites. Nonmetropolitan sites include areas contiguous with MSAs and isolated sites clustered around economic centers too small to be designated as MSAs.

Based on this definition of sites, 60 sites were randomly selected with probability in proportion to population to ensure representation of the US population. Sites were stratified by region and according to medium and large metropolitan sites (≥200000 persons), small metropolitan sites (<200000 persons), and nonmetropolitan sites to ensure diversity in these areas. The 12 sites that were selected for case studies (high-intensity sites) were randomly selected from MSA sites with populations of 200000 or more.

Data collection for the household survey was performed by Mathematica Policy Research Inc, Princeton, NJ. Households were randomly selected within each of the 60 CTS study sites. The primary sample selection method was random-digit dialing (RDD). In addition, a small field sample was included to provide coverage of families and persons who did not have telephones or who had substantial interruptions (≥2 weeks) in telephone service during the survey year. Because of the high cost of conducting in-person interviews at all 60 sites, the field sample was limited to the 12 high-intensity sites. Within these sites, the general strategy was to identify and sample census block groups that included a relatively high proportion of households without telephone service, select housing units within these areas, screen the sample for eligible households (ie, no working telephones), and select eligible households for interviews.

Although the combined RDD and field sample from the 60 sites is nationally representative (because the 60 sites were randomly selected), the clustering of the sample into 60 sites results in considerably smaller effective sample sizes (and therefore less precise estimates) than would be the case with simple random sampling or a sample based on a much larger number of sites. To increase the precision of national estimates, a supplemental national sample of households (using RDD sampling) was added to the sample for the 60 sites. The supplemental national sample was not restricted to any particular set of sites and closely approximated simple random sampling.

Within sampled households, families were defined to include all persons eligible for coverage under a typical private health insurance policy. These include the respondent, their spouse, children 17 years and younger, and children aged 18 to 23 years who were full-time students. Other relatives (including adult children who were not full-time students, grandparents, aunts, and uncles) or nonrelatives in the household were considered separate families and were interviewed separately. All interviews were conducted by telephone. Field sample subjects were provided cellular telephones by survey field staff to complete the interview. Interviews were conducted in Spanish for respondents who were not fluent in English or who preferred to conduct the interview in Spanish.

During the interview, information was obtained on all adults in the family as well as 1 randomly selected child in the family. Information about each individual's health insurance coverage, health care use, employment status, and demographic characteristics was provided by the family informant (the person most knowledgeable about the family's health care). Information on access to care, satisfaction with care, health status, and tobacco use was self-reported by all adults and proxy-reported for children (questions on tobacco use were not asked of children).

The final sample consisted of 32732 families (60446 individuals), including 635 families (950 persons) in the field sample (the non-RDD part of the sample) and 3276 families (6075 persons) who were part of the national supplement sample. Of the total sample, 7200 were uninsured on the day of the interview. The final sample reflects a response rate of 64% for families for the RDD sample and 82% for the field sample, which resulted in an overall combined response rate of 65%. For the RDD sample, this response rate takes into account families who were contacted and refused to participate and households for whom contact was not made (imputed to be eligible residences).

The potential for bias in estimates resulting from survey nonresponse cannot be assessed directly because there was no information collected on families who refused to participate in the survey. Person-level weights used for making population estimates (described herein) were poststratified to correct for any differences in nonresponse based on age, sex, race or ethnicity, and education. Even after the weighting adjustments, there were still some minor differences in sample characteristics between the CTS Household Survey and national data based on the Current Population Survey (CPS). Specifically, estimates of the percentage of persons who were uninsured were slightly lower in the CTS Household Survey than in the CPS (15.4% in CTS vs 17.7% in CPS). However, further analysis revealed that about half of this discrepancy is due to the fact that the CPS essentially measures uninsurance as a residual (ie, persons who do not specifically report any private or public health insurance coverage), whereas the CTS Household Survey included a verification for these individuals to confirm that they were uninsured. The CTS Household Survey sample is also slightly poorer (ie, having lower average family incomes) compared with the CPS sample. It is highly unlikely that these minor discrepancies have any meaningful effect on the results and major conclusions of this study given the fact that the discrepancies are quite small and the main purpose is to compare uninsured persons across communities based on their access to care rather than making nationally representative estimates of the uninsured population.

All estimates presented in this article were weighted to be representative of the civilian noninstitutionalized population of the continental United States as well as for each of the 60 sites. Weights were constructed to allow nationally representative estimates to be produced from the combined 60-site and national supplement samples and to restore proportionality to the sample made necessary by the complex sample design. Because the field sample was conducted only in the 12 high-intensity sites, an additional adjustment to the weights was made for the national supplement and low-intensity site samples. For the purposes of calculating weights, nontelephone households in these areas were represented by households who had experienced some interruption in their telephone service in the previous year (asked in the survey). Person-level weights also were poststratified to correspond with the distribution of the population by age, sex, education, and race or ethnicity as reported in the CPS.

The SEs used in tests of statistical significance were computed using SUDAAN computer software, Version 7.0 (Research Triangle Institute Inc, Research Triangle Park, NC) and take into account the complex survey design, including the clustering of the 60-site sample, the national supplement, the mixed sample frames, and selection of multiple families within a household.10

Measure of Access to Care

The measure of access to care used in this study is similar to that used in other recent studies based on survey data and reflects the ability of uninsured persons to obtain medical care.3,7 Individuals were asked 2 questions: (1) During the past 12 months, was there any time when you didn't get the medical care you needed? and (2) Was there any time during the past 12 months when you put off or postponed getting medical care you thought you needed? Follow-up questions ascertained the specific reasons why care was either not received or delayed. For this study, individuals are considered to have had difficulty obtaining medical care if the reasons cited for postponing or not receiving care include the cost of care (including lack of insurance), problems with health insurance or obtaining referrals, difficulty finding physicians or making appointments, proximity to clinicians, and transportation problems.

Analytic Methods

Variation in access to care across communities for uninsured persons is assessed by a comparison of means across all 60 sites. The means for specific sites show the percentage of uninsured persons in that site who had difficulty obtaining medical care in the previous year. To determine whether variations in the percentage of uninsured persons having difficulty were accounted for by differences in the characteristics of the uninsured population, we computed estimates for each of the 60 sites that adjusted or controlled for differences in individual characteristics. Site-specific means and adjusted means are reported for only the 12 high-intensity sites, where sample sizes were large enough to permit separate reporting. The coefficient of variation is used to summarize the variation across all 60 sites and is computed by dividing the SD of the mean for the 60 site means (ie, using the site as the unit of analysis) by the overall mean for the 60 sites. A substantial decrease in the coefficient of variation for all 60 sites after controlling for individual characteristics would indicate that these characteristics account for much of the variation across communities in the ability of uninsured persons to obtain medical care.

The conceptual framework used to select individual characteristics for estimating adjusted means is based on the Anderson model of access to care, one of the most widely used and accepted models of access in the health services research literature.11,12 Access is conceptualized as a function of (1) the need for health care as indicated by the presence of symptoms or illness; (2) enabling factors, such as income and health insurance status, which allow individuals to satisfy a need for care; and (3) predisposing factors, which reflect preferences, styles of health care use, expectations, and other non–health-related factors that affect the demand for care.

In this analysis, 2 sets of adjustments to community estimates of access were performed. The first adjustment accounts for differences in access that are the result of differences across communities in the health care needs of the uninsured population. Variables used to reflect health care need included measures of perceived health status, age, and sex. Health status for adults was measured using the SF-12, a 12-item scale that was derived from the 36-item health status scale (SF-36) used in the Medical Outcomes Study.13 Previous studies have shown that the SF-12 scale performs well in tests of validity and reliability and is highly predictive of health care use.13,14 The SF-12 includes separate components for physical and mental health, and both are used in this study as separate continuous variables (high values on both scales represent better health). Because the SF-12 was designed to be used only for adults, a separate and less detailed measure of health status was used for children. This standard, single-item measure ascertained whether a parent perceived their child's health to be excellent, very good, good, fair, or poor. Binary variables were created for each of these categories (only for the children's health status variable), with the variable for poor health excluded from the model. Age and sex are also included as need factors, since many health problems are age and sex specific and previous research has shown these 2 variables to be strongly correlated with health care use.15

The second adjustment accounts for the additional variation in access to care that is due to differences across communities in enabling and predisposing characteristics of the uninsured. These include family income, educational attainment, family size, race or ethnicity, and whether the interview was conducted in Spanish. Given the number and breadth of individual characteristics controlled for in this final adjustment, any remaining variation in access across communities would strongly suggest that characteristics of the community—including the health system—account for at least some differences in access to care.

Adjusted estimates of site means were computed using ordinary least squares regression analyses with uninsured individuals as the unit of analysis. Specifically, variables for the individual characteristics listed herein as well as dummy variables representing the 60 communities were included as independent variables. Based on the regression results, the predicted percentage of uninsured persons with difficulty obtaining care was computed for each of the 60 communities. The extent to which variation in means across communities is reduced after adjusting for these sets of factors indicates the extent to which they account for the variation in access. Although nonlinear estimation techniques, such as logistic regression or probit, are normally used when the dependent variable is dichotomous (ie, whether an individual indicated difficulty obtaining care), computing site-specific means is complicated by the nonlinear aspects of the model. Since the effects of the independent variables were similar for both a logistic and weighted least squares regression, we used the ordinary least squares results to compute site-specific means because it was much more straightforward.

To test the effects that providing private health insurance to all uninsured persons would have on variations in access across communities, we simulated the probability of uninsured persons having difficulty obtaining care if all were provided with private insurance coverage. This was accomplished by estimating a multivariate model for privately insured persons predicting whether they had difficulty obtaining medical care (using the same independent variables mentioned herein as well as dummy variables representing the 60 sites). The coefficients from this regression equation (which represent the behavior of privately insured persons) and the population means of the independent variables for uninsured persons (which represent the characteristics of uninsured persons) were used to predict the percentage of uninsured persons in each community who would have difficulty obtaining care if they were provided with private health insurance. A similar simulation was conducted to test the effects of extending public coverage to all uninsured persons, which includes the beneficiaries of Medicaid as well as other state programs but excludes Medicare beneficiaries.

The effects that expanding insurance coverage would have on variations in access across all 60 communities were examined by comparing the predicted percentage of uninsured persons having difficulty obtaining care in the group of sites where uninsured persons had the most difficulty obtaining care with the group of sites where uninsured persons had the least difficulty obtaining care. The 20% of sites (12 sites) that had the highest percentage of uninsured persons with difficulty obtaining care (controlling for individual characteristics) was used to define sites with the worst access, whereas the 20% of sites that had the lowest percentage of uninsured persons reporting difficulty obtaining health care was used to define sites with the best access. All estimates (actual and predicted means for the uninsured) are based on the weighted average for those groups of sites.

Difficulty Obtaining Care for the Uninsured

Nationally, about 14% of uninsured persons reported that they did not get needed medical care in the previous year, and 29% had to postpone getting needed medical care (Table 1). Overall, the rate of access problems for uninsured persons is about twice as high as for persons with private insurance and Medicaid and 3 times as high as for the Medicare population. Although differences in reported difficulty by health insurance coverage are similar to other recent national estimates,16 it is noteworthy that national estimates obtained in the late 1980s showed that Medicaid beneficiaries had levels of unmet need similar to those of uninsured persons.17 However, this cannot be interpreted as an improvement in access to care for Medicaid beneficiaries (eg, as a result of Medicaid managed care) because of differences across surveys in the way that questions on access to care were asked.18

Table Graphic Jump LocationTable 1.—Percentage of Subjects Who Reported Difficulty Obtaining Care*

Variations across communities on the 2 component measures (ie, did not get care, postponed getting care) were not substantially different from the combined measure, so the results for variations across communities are reported for only the combined measure (Table 2). These results show that the rate of access problems among the uninsured population varies more than 2-fold across the 12 high-intensity sites, from more than 40% in communities that had the highest rate of access problems among the uninsured (Cleveland, Ohio; Lansing, Mich; Seattle, Wash) to less than 20% in the community with the lowest rate of access problems among the uninsured (Orange County, California). There were numerous differences among the 12 high-intensity sites; of a total of 66 pairwise comparisons that tested differences between individual communities, 29 indicated statistically significant differences.

Table Graphic Jump LocationTable 2.—Variations Across Communities in the Percentage of Subjects Who Had Difficulty Obtaining Care*

Adjusting for need reduces the coefficient of variation by 1.8, whereas controlling for other characteristics of the uninsured population reduces the coefficient of variation by another 1.3. These reductions indicate that about 9% of the total variation across all 60 sites is accounted for by differences in the health care needs of the population, while another 6% of the variation is accounted for by other characteristics of the uninsured population. Further analyses (data not shown) showed that Hispanic ethnicity and the Spanish interview variable were the most important population characteristics accounting for the variation. This is due to both the high concentration of Hispanics in certain communities and the strong independent effects of these variables on difficulty in obtaining care (Table 3). In the 12 high-intensity sites, most of the biggest changes in the site-specific estimates were for sites that either had large Hispanic populations (eg, Miami, Fla; Orange County) or very small Hispanic populations (eg, Lansing; Cleveland; Syracuse, NY). Nevertheless, the substantial amount of unexplained variation across all 60 communities strongly suggests that factors other than characteristics of the population account for variations in access to care of the uninsured.

Table Graphic Jump LocationTable 3.—Results of Ordinary Least Squares Regression Analysis Used for Computing Adjusted Site Means for Uninsured Persons*
Comparison With the Privately Insured

Variations across all 60 sites in the ability of privately insured persons to obtain health care appear to be somewhat less than for the uninsured, and access problems among the privately insured are consistently lower than that of the uninsured across the 12 high-intensity sites. The pattern of variation in access for the privately insured differs from that of the uninsured. There was virtually no correlation in the percentage reporting difficulty between the uninsured and privately insured across 60 communities, as indicated by a Pearson r of 0.04. A visual inspection of the 12 high-intensity sites also indicates that communities with relatively high levels of access problems among the uninsured do not necessarily have the highest levels of access problems for the privately insured. This suggests that the access problems encountered by uninsured persons are not generic problems of the entire community but are specific to the uninsured sector of the health care market.

The lack of correlation may also reflect the fact that uninsured and privately insured persons report different reasons for encountering difficulties in obtaining care. Uninsured persons overwhelmingly cite cost concerns as the reason for having difficulty (90%, data not shown), whereas, in addition to cost concerns (48%), privately insured persons also cite problems with health insurance and obtaining referrals (28%) and the ease and convenience of using the system (33%).

Implications for Expanding Health Insurance Coverage

The simulation results show that expanding health insurance coverage would substantially reduce the amount of difficulty that uninsured persons experience in obtaining medical care. Overall, the percentage of uninsured persons who experienced difficulty obtaining care would decrease from about 31% to 17% if they were given private insurance coverage and drop to about 21% if they were given public coverage (mostly Medicaid) (Table 4). Moreover, the effects of expanding insurance, whether through private or public coverage, would result in much greater uniformity across all 60 sites in the level of difficulty that uninsured persons currently experience. Thus, providing all uninsured persons with insurance would not only improve their ability to obtain medical care, it would also substantially reduce the effects of place of residence as a major factor in the ability of uninsured persons to obtain medical care.

Table Graphic Jump LocationTable 4.—Simulation Results for Effects of Expanding Insurance Coverage to Uninsured Persons*

This is not to say that the actual effect of providing health insurance to the uninsured would be uniform across communities. Indeed, the simulations in Table 3 strongly suggest that the effect of providing insurance would be much greater in communities that currently have the worst access to care for uninsured persons and that the effect would be more modest in those communities that already have relatively good access to care for the uninsured. However, these differential effects would result in greater uniformity across communities in access to care for persons who were previously uninsured.

These simulations also suggest that expanding health insurance coverage would be a more effective method of improving access to care for the uninsured than through improvements in the current system of direct delivery of services by safety net providers. That is, even if all sites were to achieve a level of access to care for the uninsured that was as good as the best sites (eg, through improvements in the health care safety net), the improvement in access to care for all uninsured persons would not be as great as providing all uninsured persons with health insurance coverage.

The results of this study strongly suggest that there are differences across communities in the ability of uninsured persons to obtain medical care. For those who are uninsured, therefore, residence is an important factor in determining the difficulty in obtaining care. Furthermore, the variations across communities in access to care for uninsured persons were largely not the result of differences in the need for health care or other characteristics of the uninsured population.

However, identifying the particular characteristics of communities that affect access for uninsured persons is far more difficult. The site visits that were conducted in each of the 12 high-intensity sites included interviews with major safety net providers, such as public hospitals and community health centers. Although much was learned from these interviews regarding how safety net providers are coping with and responding to changes in the health care system, it is less clear from these findings why the uninsured in Newark, NJ, and Orange County would have less difficulty obtaining health care than in other communities.19 Both communities have safety net providers that serve the uninsured as well as some public subsidy of charity care, but the safety net in these communities was not viewed by site visit respondents as particularly extensive, well funded, or highly organized. Like most of the other communities in the study, safety net providers in Newark and Orange County were struggling financially and were reportedly threatened with reduced funding and capacity to serve the uninsured population.19

Other characteristics of these 2 communities may explain the relatively low rate of difficulties among the uninsured. Both the Newark metropolitan area and Orange County are relatively wealthy areas (although with pockets of high poverty) with a higher-than-average supply of physicians. Clinicians in wealthy areas with many paying patients may not feel as threatened financially by the uninsured and may be more willing to treat them for free or at reduced cost. Unpublished results from the CTS physician survey show that physicians in Newark provided more charity care (approximately 10 hours per month on average) than physicians in any of the other 12 high-intensity sites.

Also, both communities are part of the 2 largest metropolitan areas in the United States (Los Angeles, Calif, and New York, NY). The larger areas expand the options of uninsured persons by allowing them to cross over into contiguous areas that also have extensive health care facilities, including safety net providers. Indeed, site visit respondents in Orange County reported that some uninsured patients in that community used the more extensive public facilities of nearby Los Angeles County.19

Of the small amount of variation that was accounted for by population characteristics, the racial and ethnic composition of the population, especially Hispanic persons, stood out as one of the most salient. Since uninsured Hispanic persons and those who conducted the interview in Spanish reported fewer problems in obtaining medical care, the level of access problems for uninsured persons is understated in communities with a large uninsured Hispanic population, unless this factor is explicitly controlled for. These findings are somewhat surprising since access to care among Hispanic persons is often presumed to be worse because of language and other cultural barriers, although the available research on access to care for uninsured Hispanic persons is inconclusive.7,15,20 The findings may reflect either lower demand because of different expectations, preferences, or patterns of care seeking or more extensive support systems among Hispanic groups in some communities that include referrals to culturally sensitive clinicians, as has been found for Cuban immigrants in Miami.21 Unfortunately, sample sizes for uninsured Hispanic persons are too small in most communities to permit a more detailed analysis of differences in the access to care of uninsured Hispanic persons. Nevertheless, the findings do suggest that the ethnic and cultural composition of the uninsured population is an important consideration for communities in addressing the problem of uninsured persons.

Some limitations of this study should be noted. Although the measure of access used in this study is one of the most specific and direct measures of access available on surveys of households and consumers, it is based on patients' perceptions of their need for services, which may be influenced by their expectations and preferences, and not on a clinical evaluation. The measure does not capture more detailed aspects of health care delivery or distinctions among preventive care, specialty care, inpatient care, diagnostic procedures, and surgical procedures. The health consequences of delaying or not obtaining care were not ascertained in the study, although other surveys found that most uninsured persons who reported difficulty in obtaining care perceived their symptoms to be serious.3

Also, the small samples of uninsured persons in some communities resulted in some fairly large confidence intervals around estimates of access problems, particularly in communities that were not part of the 12 high-intensity sites. Although the SEs of estimates were taken into account when discussing differences among individual communities, some of the variation across all 60 sites may be due to large sample variances for some sites.

Finally, we cannot with absolute certainty attribute the variation across communities to community-level or health care system factors, despite the extensive set of individual characteristics that were controlled for in the analysis. It is possible that some unmeasured individual characteristics account for all or part of the remaining variation, although the individual characteristics controlled for in this analysis are the major determinants of health care use and access that have been identified in the literature.

Future research should examine more systematically the effects of specific market characteristics on health care use and access by the uninsured. Because of the lack of any serious political consideration being given to universal health insurance coverage, such an understanding is important because any improvements in access to care for the uninsured in the foreseeable future will have to come through the direct delivery of services via the health care safety net and other clinicians. Such improvements are likely to be difficult. The health care safety net is less a highly organized, integrated, and well-financed medical care delivery system than a loosely organized patchwork of publicly subsidized hospitals, clinics, local health departments, and other individual clinicians who care for the uninsured on a more ad hoc basis. Also, there is very little understanding as to how the uninsured actually obtain health care and even less understanding as to how access to care of the uninsured is affected by specific aspects of the medical care delivery system.

Although increasing public and private subsidies to major providers of care to the uninsured (as is currently done through Medicaid and Medicare Disproportionate Share Hospital payments, various types of state and local uncompensated care pools, and local tax assessments for public clinics and hospitals) would seem to be the most obvious way to improve access, it is important to note that there is very little understanding as to whether and how these subsidies actually result in improved access to care for uninsured persons. These funds are directed at clinicians rather than uninsured persons, and much of the existing funding is used to help support hospital uncompensated care costs. But while hospitals have long been the providers of last resort in most communities, poor access to primary care that emphasizes prevention and the detection and early treatment of health problems may unnecessarily increase uncompensated care costs for inpatient care. Research has shown that uninsured persons have higher rates of avoidable hospitalizations than insured persons.22

Although federally funded community health centers play an important role in providing primary care to the medically indigent in areas where they exist, they serve relatively few uninsured persons nationally, perhaps in part because many uninsured persons do not live in medically underserved areas where community health centers tend to be located.23 Uninsured persons see private physicians more than most people realize (more than one third of uninsured persons identify a physician's office as their usual source of care, according to unpublished findings from the CTS Household Survey), but the vast majority of private physicians receive no subsidies for providing care for free or at reduced fees.

Finally, poor access to care in some communities and good access in other communities may in part be based on differences in cultural attitudes regarding care to the uninsured and indigenous voluntary efforts that are highly idiosyncratic to particular communities, factors that cannot be manipulated easily through public policy interventions. Because the limited federal role in providing care to the medically indigent is likely to shrink as a result of reductions in Medicare and Medicaid Disproportionate Share Hospital payments due to the Balanced Budget Act of 1997, policies that are directed at providing care to the uninsured will continue to be driven largely at the state and local level. As a result, variation across communities in access to care for the uninsured is likely to persist or grow even larger.

Feldman R, Baxter RJ, Omata T. Staying in the game: health system change and challenges in caring for the poor. Presented at: meeting of the Center for Studying Health System Change, Analysis of Health Systems in 12 Communities; September 22, 1997; Washington, DC.
Lipson DJ, Naierman N. Safety-net providers.  Health Aff (Millwood).1996;15:33-48.
Donelan K, Blendon RJ, Hill CA.  et al.  Whatever happened to the health insurance crisis in the United States?  JAMA.1996;276:1346-1350.
Rask KJ, Williams MV, Parker RM, McNagny SE. Obstacles predicting lack of regular provider and delays in seeking care for patients at an urban public hospital.  JAMA.1994;271:1931-1933.
Franks P, Clancy CM, Gold MR. Health insurance and mortality.  JAMA.1993;270:737-741.
Hadley J, Steinberg EP, Feder J. Comparison of uninsured and privately insured hospital patients.  JAMA.1991;265:374-379.
Berk ML, Schur CL, Cantor JC. Ability to obtain health care: recent estimates from The Robert Wood Johnson Foundation National Access to Care Survey.  Health Aff (Millwood).1995;14:139-146.
Kemper P, Blumenthal D, Corrigan JM.  et al.  The design of the Community Tracking Study.  Inquiry.1996;33:195-206.
Metcalf CE, Kemper P, Kohn LT, Pickreign JD. Site Definition and Sample Design for the Community Tracking Study . Washington, DC: Center for Studying Health System Change; 1996. Technical publication 1.
Shah BV, Barnwell BG, Bieler FS. SUDAAN User's Manual, Release 7.0 . Research Triangle Park, NC: Research Triangle Institute Inc; 1996.
Anderson R, Newman JF. Societal and individual determinants of medical care utilization in the United States.  Milbank Memorial Fund Q Health Soc.1973;51:95-124.
Anderson RM. Revisiting the behavioral model and access to medical care: does it matter?  J Health Soc Behav.1995;3:1-10.
Ware JE, Kosinksi M, Keller SD. A 12-item short-form health survey.  Med Care.1996;34:220-233.
Kravitz RL, Greenfield S, Rogers W.  et al.  Differences in the mix of patients among medical specialties and systems of care.  JAMA.1992;267:1617-1623.
Cunningham P, Cornelius L. Use of Health Care: Findings From the SAIAN and the Household Survey . Rockville, Md: Agency for Health Care Policy and Research, Public Health Service; 1993. AHCPR publication 93-0041. National Medical Expenditure Survey Research Findings 16.
Weinick RM, Zuvekas SH, Drilea SK. Access to Health Care: Sources and Barriers, 1996 . Rockville, Md: Agency for Health Care Policy and Research, Public Health Service; 1997. AHCPR publication 98-0001. National Medical Expenditure Survey Research Findings 3.
Himmelstein DU, Woolhandler S. Care denied: US residents who are unable to obtain needed medical services.  Am J Public Health.1995;85:341-344.
Berk ML, Schur CL. Measuring access to care: improving information for policymakers.  Health Aff (Millwood).1998;17:180-186.
Kohn LK, Kemper P, Baxter RJ, Feldman RL, Ginsburg P. Health System Change in Twelve Communities . Washington, DC: Center for Studying Health System Change; 1997.
Bashshur RL, Homan RK, Smith DG. Beyond the uninsured: problems in access to care.  Med Care.1994;32:409-419.
Portes A, Kyle D, Eaton WW. Mental illness and help-seeking behavior among Mariel Cuban and Haitian refugees in south Florida.  J Health Soc Behav.1992;33:283-298.
Weissman JS, Gatsonis C, Epstein AM. Rates of avoidable hospitalization by insurance status in Massachusetts and Maryland.  JAMA.1992;268:2388-2394.
Cunningham PJ, Tu HT. A changing picture of uncompensated care.  Health Aff (Millwood).1997;16:167-175.

Figures

Tables

Table Graphic Jump LocationTable 1.—Percentage of Subjects Who Reported Difficulty Obtaining Care*
Table Graphic Jump LocationTable 2.—Variations Across Communities in the Percentage of Subjects Who Had Difficulty Obtaining Care*
Table Graphic Jump LocationTable 3.—Results of Ordinary Least Squares Regression Analysis Used for Computing Adjusted Site Means for Uninsured Persons*
Table Graphic Jump LocationTable 4.—Simulation Results for Effects of Expanding Insurance Coverage to Uninsured Persons*

References

Feldman R, Baxter RJ, Omata T. Staying in the game: health system change and challenges in caring for the poor. Presented at: meeting of the Center for Studying Health System Change, Analysis of Health Systems in 12 Communities; September 22, 1997; Washington, DC.
Lipson DJ, Naierman N. Safety-net providers.  Health Aff (Millwood).1996;15:33-48.
Donelan K, Blendon RJ, Hill CA.  et al.  Whatever happened to the health insurance crisis in the United States?  JAMA.1996;276:1346-1350.
Rask KJ, Williams MV, Parker RM, McNagny SE. Obstacles predicting lack of regular provider and delays in seeking care for patients at an urban public hospital.  JAMA.1994;271:1931-1933.
Franks P, Clancy CM, Gold MR. Health insurance and mortality.  JAMA.1993;270:737-741.
Hadley J, Steinberg EP, Feder J. Comparison of uninsured and privately insured hospital patients.  JAMA.1991;265:374-379.
Berk ML, Schur CL, Cantor JC. Ability to obtain health care: recent estimates from The Robert Wood Johnson Foundation National Access to Care Survey.  Health Aff (Millwood).1995;14:139-146.
Kemper P, Blumenthal D, Corrigan JM.  et al.  The design of the Community Tracking Study.  Inquiry.1996;33:195-206.
Metcalf CE, Kemper P, Kohn LT, Pickreign JD. Site Definition and Sample Design for the Community Tracking Study . Washington, DC: Center for Studying Health System Change; 1996. Technical publication 1.
Shah BV, Barnwell BG, Bieler FS. SUDAAN User's Manual, Release 7.0 . Research Triangle Park, NC: Research Triangle Institute Inc; 1996.
Anderson R, Newman JF. Societal and individual determinants of medical care utilization in the United States.  Milbank Memorial Fund Q Health Soc.1973;51:95-124.
Anderson RM. Revisiting the behavioral model and access to medical care: does it matter?  J Health Soc Behav.1995;3:1-10.
Ware JE, Kosinksi M, Keller SD. A 12-item short-form health survey.  Med Care.1996;34:220-233.
Kravitz RL, Greenfield S, Rogers W.  et al.  Differences in the mix of patients among medical specialties and systems of care.  JAMA.1992;267:1617-1623.
Cunningham P, Cornelius L. Use of Health Care: Findings From the SAIAN and the Household Survey . Rockville, Md: Agency for Health Care Policy and Research, Public Health Service; 1993. AHCPR publication 93-0041. National Medical Expenditure Survey Research Findings 16.
Weinick RM, Zuvekas SH, Drilea SK. Access to Health Care: Sources and Barriers, 1996 . Rockville, Md: Agency for Health Care Policy and Research, Public Health Service; 1997. AHCPR publication 98-0001. National Medical Expenditure Survey Research Findings 3.
Himmelstein DU, Woolhandler S. Care denied: US residents who are unable to obtain needed medical services.  Am J Public Health.1995;85:341-344.
Berk ML, Schur CL. Measuring access to care: improving information for policymakers.  Health Aff (Millwood).1998;17:180-186.
Kohn LK, Kemper P, Baxter RJ, Feldman RL, Ginsburg P. Health System Change in Twelve Communities . Washington, DC: Center for Studying Health System Change; 1997.
Bashshur RL, Homan RK, Smith DG. Beyond the uninsured: problems in access to care.  Med Care.1994;32:409-419.
Portes A, Kyle D, Eaton WW. Mental illness and help-seeking behavior among Mariel Cuban and Haitian refugees in south Florida.  J Health Soc Behav.1992;33:283-298.
Weissman JS, Gatsonis C, Epstein AM. Rates of avoidable hospitalization by insurance status in Massachusetts and Maryland.  JAMA.1992;268:2388-2394.
Cunningham PJ, Tu HT. A changing picture of uncompensated care.  Health Aff (Millwood).1997;16:167-175.

Letters

CME
Meets CME requirements for:
Browse CME for all U.S. States
Accreditation Information
The American Medical Association is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The AMA designates this journal-based CME activity for a maximum of 1 AMA PRA Category 1 CreditTM per course. Physicians should claim only the credit commensurate with the extent of their participation in the activity. Physicians who complete the CME course and score at least 80% correct on the quiz are eligible for AMA PRA Category 1 CreditTM.
Note: You must get at least of the answers correct to pass this quiz.
You have not filled in all the answers to complete this quiz
The following questions were not answered:
Sorry, you have unsuccessfully completed this CME quiz with a score of
The following questions were not answered correctly:
Commitment to Change (optional):
Indicate what change(s) you will implement in your practice, if any, based on this CME course.
Your quiz results:
The filled radio buttons indicate your responses. The preferred responses are highlighted
For CME Course: A Proposed Model for Initial Assessment and Management of Acute Heart Failure Syndromes
Indicate what changes(s) you will implement in your practice, if any, based on this CME course.

Multimedia

Some tools below are only available to our subscribers or users with an online account.

Web of Science® Times Cited: 77

Related Content

Customize your page view by dragging & repositioning the boxes below.

Articles Related By Topic
Related Collections
PubMed Articles