Baker DW, Shapiro MF, Schur CL. Health Insurance and Access to Care for Symptomatic Conditions. Arch Intern Med. 2000;160(9):1269-1274. doi:10.1001/archinte.160.9.1269
Copyright 2000 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2000
The uninsured receive less medical care than the insured. However, it is not known whether the uninsured are less likely to seek medical care for highly "serious" or "morbid" symptoms.
Participants in the National Access to Care Survey were asked whether they had experienced any of 15 highly serious or morbid symptoms. Those who did were asked whether they received medical care and, if care was not received, whether care was thought to have been necessary.
A total of 574 respondents (16.4%) reported 794 new serious or morbid symptoms. Of these, 499 people (86.9%) had health insurance, and they reported 691 new symptoms; 75 (13.1%) lacked health insurance, and they reported 103 symptoms. The uninsured were less likely to have received medical care and more likely to say they did not receive care even though they thought it was needed (P=.001). Medical care was received for 45.4% of symptoms for the insured and 24.3% for the uninsured; care was not thought to have been needed for 41.0% of symptoms for the insured and 45.6% for the uninsured; and care was thought necessary but was not received for 13.6% of symptoms for the insured and 30.1% for the uninsured. In multivariate analysis, the uninsured were much less likely to have received care than the insured (adjusted odds ratio, 0.43; 95% confidence interval, 0.23-0.81).
Lack of health insurance is a major barrier to receiving medical care, even for highly serious and morbid symptoms.
DESPITE a large body of evidence suggesting that not having health insurance adversely affects health care use, costs, and outcomes,1- 8 approximately 44 million people in the United States remain uninsured.9,10 A variety of historical, political, and economic factors contribute to our society's unwillingness to amend this problem.10,11 One contributing factor is the perception that the uninsured get essential care anyway through public clinics and hospitals or uncompensated care from other providers.11 Studies suggest otherwise; the uninsured have fewer physician visits and are less likely to be hospitalized than the insured.1- 4 This difference in health care use between the insured and the uninsured has been called "the uninsured access gap."2
However, noting the differences in overall use of services between the insured and the uninsured does not address whether the uninsured receive needed medical care, because the optimal rate of health care use for a population is not known. It is as plausible that the insured overuse services as it is that the uninsured underuse them. Moreover, differences in rates of overall health care use or receipt of health care for specific conditions may reflect variations in individuals' perceived need for medical care rather than differences in the accessibility of health care for patients who desire it.
An alternative that may be easier to interpret is to look at symptom-specific use.12 In other words, if a patient had a specific condition such as chest pain, cough, or a sprained ankle, did she receive medical attention? Few studies have used this approach to examine whether the uninsured are less likely to receive needed medical care. Freeman et al1 reported that among adults with 1 or more of 5 "serious" symptoms (bleeding not due to nosebleeds or menstrual periods, loss of consciousness, shortness of breath with light exercise, chest pain when exercising, or a weight loss of more than 10 lb (4.5 kg) that was not the result of dieting), 33% of the uninsured said they received medical care compared with 59% of those who were insured. Similarly, uninsured children with pharyngitis, acute earache, recurrent ear infections, or asthma were less likely to receive medical care than children covered by some form of health insurance.13 However, the importance of these results is limited by the small number of conditions examined.
For the 1994 Robert Wood Johnson Foundation National Access to Care Survey, we developed a list of 15 symptoms that a national sample of physicians had rated as being highly serious or having a large negative effect on quality of life ("morbid" symptoms).12 Participants in the National Access to Care Survey, which was designed to provide nationally representative data, were asked whether they had experienced any of these 15 symptoms over the previous 3 months. Those reporting 1 or more symptoms were asked whether they received care and, if they did not receive care, whether they needed to see a physician. We analyzed the results of this survey to determine whether the uninsured were less likely than the insured to receive medical care for new highly serious or new morbid symptoms.
The National Access to Care Survey was designed by the Project HOPE Center for Health Affairs, Bethesda, Md, and implemented by the National Center for Health Statistics, Hyattsville, Md, and the Bureau of the Census, Washington, DC, with funding from the Robert Wood Johnson Foundation, Princeton, NJ. The survey is a national probability sample designed to provide data representing the entire US population. Information is collected on access to medical care, including health insurance coverage, usual source of medical care, use of health services, and barriers to obtaining care. The survey was fielded in the spring and summer of 1994 as a follow-up component to the 1993 National Health Interview Survey. Households were sampled, and 1 adult per household was randomly selected to be interviewed. Most people were interviewed by telephone, but persons who reported on the National Health Interview Survey that they did not have a telephone were interviewed in person. A total of 3480 people were interviewed (76% response rate).
Respondents were asked whether they had experienced any of 15 symptoms during the previous 3 months. A national sample of physicians previously identified these symptoms as being highly serious (eg, shortness of breath with light work or exercise) or greatly affecting patients' quality of life (morbid, eg, back or neck pain that impairs activities; Table 1).12 Four versions of the survey were used that presented the symptoms in different orders. For the first 3 symptoms reported by the respondent, it was determined whether medical care was received for the condition. If care was not received, the respondent was asked, "Did you think you needed to contact a physician or other medical person about this problem?" On the basis of these questions, the care received–perceived need for care for each symptom was classified as (1) medical care was received, (2) medical care was not received but the respondent did not think contact with a physician or other medical person was necessary, and (3) medical care was not received and the respondent thought that contact with a physician or other medical person was necessary. The survey also asked whether each symptom was "new" or "old" (ie, present prior to 3 months before the interview). Because this study was designed to look at obtaining medical care for acute, symptomatic conditions rather than chronic medical problems, only new symptoms were used for analysis (which constituted 37.2% of all symptoms). Most patients with old symptoms had either received care at some point in the past or thought that care was not necessary.
Respondents who stated that they did not receive medical care that they thought was necessary were asked why they had not received care, as an open-ended question. After the answer, the respondent was asked "Anything else?" until all reasons were elicited. Respondents were then asked (1) "Was your health affected in any way because you did not receive medical care?" and (2) "Did you have any personal, household, or work problems because you did not receive medical care for this problem?"
Symptoms were used as the unit of analysis. The association between insurance status and the care received–perceived need for care was analyzed for (1) all serious symptoms, (2) all morbid symptoms, and (3) all serious or morbid symptoms combined. The statistical significance of the associations between insurance status and care received–perceived need for care was determined using χ2 tests. The number of observations was too small to perform valid statistical tests for individual symptoms.
The independent association between insurance status, perceived need for care, and care received was analyzed using logistic regression to adjust for confounding variables. Age, sex, race, income, self-reported health, years of school completed, and whether the respondent had a regular source of care were defined a priori as covariates and included in all models regardless of statistical significance. Information on 1 or more of the covariates was missing for 14 reported symptoms, and these were dropped from the multivariate analysis leaving 785 symptoms for analysis. For the first logistic regression model, we analyzed whether the respondent said that care was not necessary. A dependent variable of "care not thought necessary" was set equal to 1 if the respondent said that medical care was not needed or equal to 0 for respondents who either received care or thought care was needed but did not receive care. Next, we analyzed whether care was actually received with the dependent variable set equal to 1 if care was received and 0 if either care was not thought necessary or care was thought necessary but not received. Finally, we repeated the second logistic regression after excluding those symptoms for which the respondent said that care was unnecessary, leaving 459 symptoms for analysis. This last analysis determined whether someone with a perceived need for medical care actually received it.
All analyses were conducted using Stata 6.0 (College Station, Tex). For multivariate analyses, SEs for beta coefficients were inflated using the "cluster" option to account for the fact that 32% of patients reported more than 1 symptom and were therefore represented more than once in the data set. Further adjustment of the confidence intervals for the complex survey sampling did not affect the results. A P value of .05 was used to determine final statistical significance.
A total of 574 people (16.4%) reported a new serious or morbid symptom. Of these, 499 (86.9%) had health insurance and 75 (13.1%) lacked health insurance. Respondent characteristics for the insured and the uninsured are shown in Table 2. The uninsured were younger and more likely to be nonwhite, poor, and lack a regular source of care. The uninsured were also less likely to have had education beyond high school. Self-reported overall health was similar for the 2 groups. The 574 insured respondents reported 691 new symptoms, and the 75 uninsured respondents reported 103 new symptoms. The types of symptoms reported were similar for the 2 groups, although the uninsured were somewhat less likely to report a morbid symptom (P=.04; Table 2).
Among respondents who reported any serious or morbid symptom, those who lacked health insurance were less likely to have received medical care and more likely to say they did not receive care even though they thought it was needed (P=.001; Table 3). Medical care was received for 45.4% of symptoms for the insured and 24.3% for the uninsured; care was not thought to have been needed for 41.0% of symptoms for the insured and 45.6% for the uninsured; and care was thought necessary but was not received for 13.6% of symptoms for the insured and 30.1% for the uninsured. The results were similar when serious and morbid symptoms were analyzed separately (Table 3). Although differences could not be analyzed for individual symptoms because of small numbers, the uninsured were less likely to have received medical care for all 9 of the most frequently reported symptoms (P=.002 by the sign rank test; Table 3).
To analyze the independent relationship between insurance status, perceived need for care, and actual medical care received after adjusting for other patient characteristics, we used logistic regression. First, for all serious and morbid symptoms (N=785), we analyzed whether the respondent said care was not needed. There was no difference in the perceived need for care for the uninsured and the insured; the adjusted odds ratio for saying care was not necessary was 1.09 (95% confidence interval [CI], 0.66-1.81) for the uninsured compared with the insured. Respondents were less likely to say medical care was unnecessary if they reported their overall health as "fair" or "poor" and more likely to say medical care was unnecessary if their income was below 150% of poverty level (Table 4), although both of these did not reach statistical significance.
When all serious and morbid symptoms were analyzed, the uninsured were much less likely to have actually received care compared with the uninsured. The adjusted odds ratio for receiving medical care for the uninsured compared with the insured was 0.43 (95% CI, 0.23-0.81). Older individuals were also more likely to have actually received care. Finally, we repeated this logistic regression using only those symptoms for which the respondent thought care was needed (n=459). The insured received care for 76.9% of these symptoms, and the uninsured received care for 44.8% of symptoms. After adjusting for other variables, the uninsured were much less likely to have received care (adjusted odds ratio was 0.28 compared with the insured [P=.002; 95% CI, 0.13-0.62]; Table 5). Age was the only other variable significantly associated with receipt of medical care, although the sample size had limited power to detect important differences for other variables.
To analyze other possible confounding factors that could affect the relationship between insurance status and use of health care services, we performed several additional analyses. Separate logistic regression models for serious and morbid symptoms gave similar results. The relationship between insurance status and use of health care services was unchanged when the number of serious and morbid symptoms reported by respondents was included in the logistic models, and addition of dummy variables for specific symptoms also did not change the results. Finally, we ran models in which we excluded from analysis the 6 symptoms reported by 2 or fewer uninsured respondents, but the results were unchanged.
Of the insured respondents who did not receive medical care that was thought necessary for 1 or more new symptoms, the most common reason cited for not receiving care was that the physician did not accept their insurance (33.3%), followed by inability to pay (22.8%) and lack of convenient hours (15.8%). Among the uninsured respondents who did not receive medical care that was thought necessary, 95.2% said they could not pay for care (P<.001 compared with the insured). A total of 63.2% of the uninsured said that not receiving care affected their health, and 57.1% said they had personal, household, or work problems because they did not receive medical care. Among the insured, these figures were 12.7% and 19.6%, respectively (P<.001 and P=.001 compared with the uninsured).
These results show that uninsured adults are far less likely to receive medical care when they develop new symptoms that may represent serious medical conditions or that have major adverse effects on quality of life. Moreover, the lower rate of health care use by the uninsured was not explained by differences in perceived need for care; the proportion of people who thought medical care was unnecessary was similar regardless of insurance status. Our findings suggest that lack of health insurance is a major barrier to obtaining health care for important symptomatic conditions. The widely held perception that "the uninsured get care anyway"11 seems false.
Our estimate of the effect of being uninsured on use of health care for symptomatic conditions is similar to that from previous surveys. Freeman et al1 reported that the uninsured were 44% less likely to have seen a physician for any of 5 serious symptoms compared with the insured. Thus, our study expands the generalizability of this previous work to include a larger number and broader type of symptoms. The reductions in health care use for new, symptomatic conditions seen in our study and in that of Freeman et al are also consistent with studies of differences in the overall use of ambulatory health care by the uninsured. Long and Marquis2 found that uninsured adults averaged 2.7 visits per year compared with 4.4 visits per year for the insured, a 39% relative reduction.2
The most common reason cited by the uninsured for not receiving necessary care was inability to pay. Because the uninsured often must make substantial out-of-pocket payments to receive medical care, it is useful to compare the results of this study with previous studies of the effect of copayments on health care use for symptomatic conditions. Of patients in the Health Insurance Experiment who reported 1 of 5 serious symptoms, 17.9% of those in the cost-sharing group received medical care compared with 22.3% of those in the "free care" group (a 20% relative reduction).14 At a large managed care plan in California, a $25 to $35 copayment for emergency department care reduced the number of visits by 13% to 15% for conditions classified as "often an emergency."15 In contrast, our study found that medical care was received for 24.3% of symptoms reported by the uninsured compared with 45.4% among the insured, a 46% relative reduction. Thus, the effect of being uninsured seems to have greater negative consequences than even relatively large copayments for the insured population.
It is somewhat surprising that medical care was received for less than half of all symptoms regardless of insurance status, although this is consistent with previous studies.2,14 No patient characteristics were associated with reporting that care was unnecessary, suggesting that perceived need for medical care for these conditions may be highly individualized or related to characteristics of the symptoms that were not captured in the survey. For example, we did not measure symptom severity, so people who thought that care was not needed may have had less severe or less prolonged symptoms. It is therefore possible that the uninsured had more mild symptoms than the insured, and this could account for some of the differences in use of medical care. However, the majority of the uninsured who did not receive medical care that was thought necessary said that not receiving care adversely affected their health, work, home situation, or personal life.
There are several limitations to this study. First, the cross-sectional nature of the survey prevents us from determining both the future use of health care services among those who had not received care for symptomatic conditions and the long-term health consequences of not receiving or delaying medical care. Most participants who had persistent symptoms for more than 3 months (old symptoms) reported that they had either received medical care or did not think that medical care was needed. This may indicate that if serious or highly morbid symptoms persist long enough, the uninsured will eventually seek and obtain medical care. Secondly, we do not have information on the adequacy of coverage for outpatient services for those who reported having insurance. A substantial number of the insured seem to have had inadequate coverage for outpatient care based on the fact that 23% of the insured who did not receive needed care said this was due to their inability to pay. By including this group of individuals with inadequate insurance coverage within the category of insured, our analysis may underestimate the adverse effects of being uninsured on health care use.
In addition, there were a relatively small number of survey respondents who were uninsured and reported new serious or morbid symptoms. Larger studies are necessary to more accurately estimate the effect of lack of health insurance on health care use for highly serious and morbid symptoms. Nevertheless, because this study was part of a national probability sample, the respondents are representative of the uninsured population in the United States. Finally, although the symptoms used in the survey were selected because they were likely to have important consequences if left untreated, we have no objective measure of the health consequences of the lower use of services by the uninsured.
The results of this study are particularly disturbing in light of projected increases in the number of uninsured and anticipated decreases in the ability of providers to deliver uncompensated care.16 The US General Accounting Office estimates that the number of uninsured Americans will increase to 45 million over the next 5 years.17 Meanwhile, pressures from managed care will probably reduce delivery of uncompensated care,10,18,19 and "safety net providers" that serve as a critical source of care for the uninsured are likely to face increasing financial pressure as well.20,21 It is ironic that at a time when our country is establishing a "Patients' Bill of Rights" we still have not established the right to be a patient. The ability of the uninsured to obtain medical care for even the most serious and disabling conditions is likely to worsen in the coming years.
Accepted for publication September 15, 1999.
This study was supported in part by a grant from the Robert Wood Johnson Foundation, Princeton, NJ.
Preliminary results from this project were presented at the national meeting of the Association for Health Services Research, Chicago, Ill, June 22, 1998.
We dedicate this article to the memory of Howard Freeman, PhD. We would also like to thank Marc Berk, PhD, at Project HOPE, Millwood, Va, for his help in the design of the symptom-response questions and his feedback throughout the project.
Reprints: David W. Baker, MD, MPH, MetroHealth Medical Center, 2500 MetroHealth Dr, Room R221, Cleveland, OH 44109-1998 (e-mail: email@example.com).