Cumulative rates of being uninsured 1 or more times, being continuously uninsured, and having continuous private insurance coverage from 1992 to 2000. The number of individuals who continuously participated in Health and Retirement Study interviews in 1994, 1996, 1998, and 2000 was 5451, 4969, 4554, and 4188, respectively. Continuous private insurance coverage includes only individuals with comprehensive private insurance, those who said that they had basic insurance are not included.
Insurance coverage from 1994 to 2000 among the participants who were uninsured at baseline in 1992. Of the 991 individuals who were uninsured at baseline, the number of individuals who continuously participated in Health and Retirement Study interviews in 1994, 1996, 1998, and 2000 was 848, 740, 674, and 606, respectively. Individuals with basic insurance are included in private insurance.
Baker DW, Sudano JJ. Health Insurance Coverage During the Years Preceding Medicare Eligibility. Arch Intern Med. 2005;165(7):770-776. doi:10.1001/archinte.165.7.770
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
Adults in late middle age who lack health insurance are more likely to die or experience a decline in their overall health. Because most estimates of the uninsured are cross-sectional, the true number of individuals whose health is at risk from being uninsured is unclear.
We analyzed a nationally representative sample of 6065 US adults 51 to 57 years old who were interviewed in 1992, 1994, 1996, 1998, and 2000 as part of the Health and Retirement Study. Insurance coverage was determined at the time of each interview and classified as private, public, or uninsured. Longitudinal data were used to determine the proportion of individuals who were uninsured at any interview during the 8-year study period.
The proportion of participants who were uninsured at the time of the 1992, 1994, 1996, 1998, and 2000 interviews was 14.3%, 10.8%, 9.7%, 8.8%, and 8.2%, respectively. People frequently transitioned between having insurance and being uninsured. As a result, despite the declining prevalence of being uninsured, the percentage who were uninsured at least once during the 8-year period rose to 23.3% by 2000; few participants (2.6%) were continuously uninsured. Only 60.1% of participants were continuously enrolled in private insurance across all 5 interviews.
The proportion of US adults in late middle age at risk from being uninsured over a 10-year follow-up period was 2 to 3 times higher than cross-sectional estimates. At least one quarter of older adults will be uninsured at some point during the years preceding eligibility for Medicare.
Uninsured individuals are less likely to have a regular source of care, to use preventive services, to obtain timely care for acute medical problems, and to take medications for chronic illnesses.1- 3 As a consequence of these adverse effects on access to care and health care use, uninsured individuals have higher rates of morbidity and mortality.4 Adults in late middle age may be particularly vulnerable to adverse consequences from being uninsured because of their higher prevalence of chronic disease.5 Previous studies have shown that adults 51 to 61 years old who lack health insurance have higher risk-adjusted rates of decline in their overall health and physical functioning6,7 and higher risk-adjusted mortality rates.8
How many near elderly Americans are at risk from lack of health insurance? Few studies have observed individuals over long periods, and health insurance coverage changes frequently within a given year and over several years.9- 15 To examine this, we determined the number of times that participants in the Health and Retirement Study (HRS)16 said they had no health insurance at the 1992, 1994, 1996, 1998, and 2000 interviews. The HRS is a nationally representative, longitudinal study that targeted community-dwelling adults 51 to 61 years old in 1992 in the contiguous United States. To determine patterns of insurance coverage, we tracked a cohort of individuals 51 to 57 years old at baseline for the 8 years before they were old enough to be eligible for Medicare.
This study was approved by the Northwestern University Institutional Review Board. Analyses were conducted using publicly available HRS data files.17 In-home interviews were conducted by HRS for 7702 households (82% response rate), yielding 9824 age-eligible participants for wave 1 (1992). Follow-up interviews were conducted every 2 years. We restricted our study to 6148 participants 51 to 57 years old in 1992 who were younger than 65 years at the time of their interview in 2000; 83 individuals had missing insurance data at baseline and were excluded from analysis, leaving 6065 subjects. All HRS interviews obtained information on age, sex, race and ethnicity, marital status, education, household income, health behaviors, body mass index, self-reported overall health, and chronic conditions.18
Participants were asked about participation in employer-based insurance, individually purchased private insurance, and public health insurance programs (Medicare, Medicaid, CHAMPUS [Civilian Health and Medical Program Uniformed Service], and Department of Veterans Affairs). In addition, participants were asked: “Do you have any basic health insurance coverage purchased directly from an insurance company or through a membership organization?” The premiums for this basic health insurance varied widely (data not shown), and it is likely that the benefits (copayments and deductibles) also varied widely. Respondents were classified as uninsured if they said they did not have any form of public or private insurance. Those who had only catastrophic coverage were also classified as uninsured.
Analyses were conducted using survey modules (Stata, version 7; Stata Corp, College Station, Tex) to adjust analyses for the complex survey design and analytic weights. The actual number of study subjects is presented for all analyses, but the percentages shown are weighted percentages that represent estimates for the US population. The cohort of study participants 51 to 57 years at baseline was tracked from 1992 to 2000. Insurance coverage at each wave was determined for all study participants who completed the HRS interview, including the 516 people (8.5%) who missed 1 or more interviews but returned to the study.
Data from individual HRS interviews were used to determine the proportion of people who were uninsured at any HRS interview. We next determined the total number of interviews at which respondents reported being uninsured for all cohort participants who completed the 2000 HRS interview. Associations between baseline characteristics and the number of interviews at which a respondent was uninsured from 1992 to 2000 (range, 0-5) were determined using second-order corrected Pearson statistics for dichotomous variables19 and adjusted Wald statistics for continuous variables.20 Multivariate analyses were conducted using multinomial logistic regression to measure the association of baseline covariates with the likelihood of being uninsured 1, 2 to 3, or 4 to 5 times compared with being insured at all 5 HRS interviews. Odds ratios were converted to relative risks using published formulas.21
Finally, we analyzed cumulative patterns of insurance coverage by analyzing participants’ insurance coverage at the time of consecutive HRS interviews until the time at which a person did not complete an interview or died; if an individual missed an interview and then returned to the study, we used only the data until the time of initial dropout. We determined (1) the proportion of people who were continuously uninsured, (2) the proportion who had continuous private insurance, and (3) the proportion who were uninsured 1 or more times.
The baseline characteristics of the participants are shown in Table 1. The cumulative number of individuals who died by the time of each HRS interview was 85 (1.4%), 202 (3.3%), 316 (5.2%), and 460 (7.6%) for 1994, 1996, 1998, and 2000, respectively. Older age, male sex, worse self-reported overall health, having 1 or more chronic illness, and being uninsured at baseline were all associated with higher mortality from 1992 to 2002. The cumulative number of individuals who were lost to follow-up at the time of each HRS interview was 447 (7.4%), 688 (11.3%), 829 (13.7%), and 964 (15.9%) for 1994, 1996, 1998, and 2000, respectively. Spanish-speaking Hispanic individuals and those who did not attend college were more likely to be lost to follow-up from 1992 to 2002. Age, sex, income, baseline health insurance coverage, and health status were not independently associated with loss to follow-up.
The percentage of individuals who were uninsured declined markedly from 1992 (14.3%) to 1994 (10.8%) and then declined more slowly from 1994 to 2000 (8.2%; Table 2). The percentage of individuals with basic insurance coverage also declined steadily. The opposite trend occurred for Medicare, with a steady rise in the percentage of individuals covered by Medicare (presumably due to disability since all members of this cohort were younger than 65 years). The proportion of individuals covered by comprehensive private insurance, Medicaid, and the Department of Veterans Affairs or CHAMPUS remained relatively constant.
Despite the relatively stable overall distribution of insurance coverage at the 1994, 1996, 1998, and 2000 HRS interviews, insurance coverage for individual participants changed frequently. Of the 4641 participants who completed the 2000 HRS interview, 23.3% reported being uninsured at the time of 1 or more of the 5 HRS interviews; 11.6% at the time of 1 interview, 7.2% at the time of 2 or 3 interviews, and 4.5% at the time of 4 or 5 interviews. A total of 60.2% reported having private insurance coverage at the time of all interviews in which they participated. The rate of being uninsured was substantially higher among 1260 individuals who started the study in fair or poor health, 21.1% were uninsured at baseline, and 33.2% had been uninsured 1 or more times by 2000.
Women, African American and Hispanic individuals, and those with low educational attainment and/or low adjusted family income in 1992 were more likely to be uninsured 1 or more times. Baseline health status was also strongly related to future health insurance coverage (Table 3). Individuals in fair or poor health at baseline were far more likely to be uninsured 1 or more times than those in very good or excellent health. The number of chronic conditions was less closely related to insurance coverage. In multivariate analysis, African American and Hispanic participants and those of other ethnicities were more likely to be uninsured 1 or 2 or more times (Table 4). Baseline income showed a very strong relationship to being uninsured 2 or more times, and having less than a high school education was also independently associated with having multiple times without insurance. The number of chronic diseases was inversely related to having multiple times without health insurance; the adjusted relative risk of being uninsured 2 or more times was 0.39 (95% confidence interval [CI], 0.19-0.78) for participants with 4 or more chronic diseases compared with participants with none.
The number of individuals who continuously participated in HRS interviews through 1994, 1996, 1998, and 2000 was 5451, 4969, 4554, and 4188, respectively. The proportion with continuous private insurance coverage steadily declined, by the 2000 HRS interview, only 60.1% of participants were continuously enrolled in private insurance at all 5 interviews (Figure 1). The proportion of participants who were uninsured at any point in time rose briskly from 1992 to 1994 and then more slowly through 2000. Of the 4188 people who continuously participated through 2000, 21.8% were uninsured at the time of at least 1 HRS interview. The proportion of participants who were continuously uninsured steadily declined; only 2.6% of the cohort was uninsured at all 5 interviews.
Transitions among the uninsured were particularly common. Of the 991 individuals who were uninsured at baseline, the proportion who were uninsured at the time of future interviews rapidly dropped to 55.6% in 1994 and then steadily declined to 37.0% in 2000 (Figure 2). The rapid rate of transition from uninsured in 1992 to insured in 1994 was due to transitions to both public insurance (14.5%) and private insurance (29.9%). After 1994, the proportion of this cohort with private insurance coverage remained relatively constant, while the proportion with public insurance coverage steadily increased, rising to 28.2% by 2000. Thus, the steady decline in the proportion of this cohort who were uninsured between 1994 and 2000 was almost completely explained by transitions to public insurance. Almost half (49.8%) of uninsured individuals who were in fair or poor health at baseline transitioned to public insurance coverage by the time of the 2000 interview.
Patterns of change in insurance coverage were different for individuals who became uninsured over the course of the study (ie, had private insurance at baseline and reported having lost insurance at a later interview). Between 1992 and 1994, 144 participants lost their private insurance, and between 1994 and 1996, 107 lost their private insurance. Of the 251 participants who lost private insurance, 128 (50.9%) regained private insurance in 2 years and 160 (63.7%) regained private insurance in 4 years. Four years after losing private insurance, 42 (16.7%) had become insured through a public program and only 49 (19.6%) still reported being uninsured.
This study provides the most extended analysis of health insurance coverage published to date. Over the 8-year study period, almost one fourth of the study population reported being uninsured at the time of at least 1 of the 5 HRS interviews. Thus, while cross-sectional rates of being uninsured ranged from 8.2% to 14.3%, the cumulative rate of reporting being uninsured over the 8-year study period was 2 to 3 times higher. Using this same database, we have previously reported that the continuously uninsured (ie, individuals who were uninsured at the time of their 1992 and 1994 HRS interview) and the intermittently uninsured (ie, individuals with private insurance at the time of their 1992 interview who said they were uninsured in 1994) had worse risk-adjusted health outcomes than individuals who were covered continuously by private insurance.6,7 Thus, the figures from the current study suggest that one fourth of adults in late middle age are at increased risk of a health decline due to being uninsured in the years before retirement and Medicare eligibility.
Our findings of frequent transitions between being insured and uninsured and the large difference between cross-sectional and longitudinal rates of being uninsured are consistent with previous studies conducted over shorter periods.9- 15 For example, analyses of the Survey of Income and Program Participation found that 6.7% of Americans aged 55 to 64 years were uninsured for all of 1998, but 14.0% were uninsured at any time during the year.15 When the Survey of Income and Program Participation tracked individuals for 4 full years from 1996 through 1999, 21.0% of people aged 55 to 64 years were uninsured 1 or more times, although one third of those who were uninsured said it was for only a brief period (4 months or less).15
Our results show the tremendous importance of public insurance programs as a safety net for limiting the number of uninsured. Among individuals who were uninsured at baseline, over one fourth were covered by public insurance by the end of the study (Figure 2). Among those who became uninsured in 1994 or 1996, almost 1 in 5 transitioned to public insurance during the 4 years after first losing insurance coverage. This safety net is essential for older adults with severe health problems whose ability to get private health insurance coverage is limited because of restricted employment opportunities and the high cost of individual health insurance policies for people with preexisting conditions. However, many people fall through the safety net. One in 3 individuals in fair or poor health at baseline were uninsured at the time of at least 1 of the interviews during the 8-year study period.
The results of our multivariate analyses emphasize that the classic factors associated with being uninsured (ie, race or ethnicity, education, and income) are even more strongly associated with being uninsured 2 or more times. For example, compared with individuals whose adjusted family income was 500% or more of the federal poverty level, for participants whose adjusted family income was less than 100% of the federal poverty level, the adjusted relative risk for being uninsured once was 3.84 (95% CI, 2.72-5.42) but the adjusted relative risk increased to 10.90 (95% CI, 6.84-17.5) for being uninsured 2 or more times. Intermittent lack of insurance may contribute to racial, ethnic, and socioeconomic differences in health outcomes, and studies that include only health insurance coverage at baseline may underestimate the cumulative negative effect of lack of insurance coverage on population health.
There are several important limitations to this study. The dropout rate during the study was 23.5%, which could bias our findings. Individuals who lost health insurance may have been more likely to drop out of the study, which could lead to an underestimate of the true proportion of the original cohort who were uninsured during the study period. We also had no information on covered services for participants with basic insurance coverage; some plans may have covered health care costs in only the most extreme circumstances. Thus, our figures may underestimate the proportion of people with severe access barriers due to lack of adequate insurance coverage. We also did not include episodes of being uninsured between HRS interviews. The Current Population Survey reported that 12.6% of US adults aged 45 to 64 years were uninsured for all of 2000, compared with our finding that 8.2% of adults aged 51 to 57 years were uninsured at the time of the 2000 interview.22 The clinical importance of brief periods without insurance is not known. Some uninsured individuals may have good access to care through local free clinics and public hospitals with sliding-scale payment plans. Individuals with this access may face few or no economic barriers to care, although they may face serious nonfinancial barriers, such as long waiting times.23,24
The proportion of individuals who reported being uninsured dropped sharply from 1992 to 1994. Part of this may have been due to an improvement in the US economy around this time. It is also possible that some participants gave incomplete information about their insurance coverage during their baseline interview, resulting in an incorrect uninsured classification. In addition, individuals who were uninsured at baseline had higher dropout and mortality rates from 1992 to 1994. The higher dropout rates could partly account for the higher than expected drop in the number of uninsured individuals during this period.
There has been increasing concern over the number of near elderly Americans who lack health insurance.25,26 Our findings suggest that the problem is even larger than previously believed. National estimates of the proportion of people who are uninsured each year are deceiving, giving the impression that this is an accurate estimate of the population at risk from being uninsured. In fact, there is a large number of individuals who are transitioning in and out of insurance coverage, with only a minority being classified as uninsured at a given point. Those engaged in policy debates about uninsured individuals should recognize that the number of Americans whose health and economic well-being are at risk from lack of insurance coverage far exceeds the figures released from annual cross-sectional surveys.
To date, policies designed to increase rates of private insurance coverage for adults and to prevent individuals from losing insurance coverage have been unsuccessful.12 One proposed option is to allow uninsured individuals to buy into Medicaid or Medicare with the help of government subsidies for the poor and near poor.27 This option could be available to all individuals over a certain age (eg, 55 years), or the extension of eligibility could be limited to individuals with chronic diseases. A model of this approach is the Breast and Cervical Cancer Prevention and Treatment Act of 2000 (Public Law 106-354), which gives states the option to extend Medicaid eligibility to women who were screened through the Centers for Disease Control and Prevention’s National Breast and Cervical Cancer Early Detection Program and found to have breast or cervical cancer. Cardiovascular disease kills far more people than breast and cervical cancer.28 Therefore, from the perspectives of both equity and public health, it seems justified to create a similar program for people with medical conditions such as hypertension, diabetes, heart disease, hypercholesterolemia, and chronic lung disease. McWilliams and colleagues8 found that excess mortality rates in uninsured older adults were concentrated in those with diabetes, hypertension, and heart disease. Thus, expanding the availability of insurance coverage to this target population may achieve the greatest health benefits. There is now substantial evidence that uninsured adults in their preretirement years are at greatly increased risk for adverse health outcomes, and urgent policy measures are needed to expand coverage for this group.26
Correspondence: David W. Baker, MD, MPH, 676 N St Clair St, Suite 200, Chicago, IL 60611 (email@example.com).
Accepted for Publication: December 17, 2004.
Financial Disclosure: None.
Funding/Support: Supported by grant 2 R01 HS10283-03 from the Agency for Healthcare Research and Quality, Rockville, Md.