0
Articles |

Medicare Spending for Previously Uninsured Adults FREE

J. Michael McWilliams, MD, PhD; Ellen Meara, PhD; Alan M. Zaslavsky, PhD; and John Z. Ayanian, MD, MPP
[+] Article and Author Information

From Harvard Medical School, Brigham and Women's Hospital, and Harvard School of Public Health, Boston, and National Bureau of Economic Research, Cambridge, Massachusetts.


Acknowledgment: The authors thank Stuart B. Mushlin, MD, for helpful comments on an earlier draft of this manuscript.

Grant Support: By the Commonwealth Fund (grant 20060485).

Potential Conflicts of Interest:Consultancies: E. Meara (Employment Policies Institute), J.Z. Ayanian (RTI International, Verisk Health). Expert testimony: J.Z. Ayanian (U.S. House of Representatives Ways and Means Committee). Grants received: J.Z. Ayanian (National Institute on Aging).

Reproducible Research Statement:Study protocol: See Methods and the Appendix. Statistical code: Available from Dr. McWilliams (e-mail, mcwilliams@hcp.med.harvard.edu); SAS code for constructing inverse-probability-of-treatment weights are also available in appendix of reference (32). Data set: Health and Retirement Study survey data are available at http://hrsonline.isr.umich.edu/index.php?p=data. Linked Medicare claims data are restricted, and use requires approval by the Health and Retirement Study and the Centers for Medicare & Medicaid Services.

Requests for Single Reprints: J. Michael McWilliams, MD, PhD, Department of Health Care Policy, Harvard Medical School, 180 Longwood Avenue, Boston, MA 02115; e-mail, mcwilliams@hcp.med.harvard.edu.

Current Author Addresses: Drs. McWilliams, Meara, Zaslavsky, and Ayanian: Department of Health Care Policy, Harvard Medical School, 180 Longwood Avenue, Boston, MA 02115.

Author Contributions: Conception and design: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.

Analysis and interpretation of the data: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.

Drafting of the article: J.M. McWilliams, A.M. Zaslavsky.

Critical revision of the article for important intellectual content: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.

Final approval of the article: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.

Statistical expertise: J.M. McWilliams, E. Meara, A.M. Zaslavsky.

Obtaining of funding: J.M. McWilliams, J.Z. Ayanian.

Administrative, technical, or logistic support: J.M. McWilliams.

Collection and assembly of data: J.M. McWilliams.


Ann Intern Med. 2009;151(11):757-766. doi:10.7326/0000605-200912010-00149
Text Size: A A A

Background: Medicare spending after age 65 years may be higher for previously uninsured adults if suboptimal care before this age leads to irreversible complications, persistently elevated clinical risks, or delay of costly elective procedures.

Objective: To compare Medicare spending for previously uninsured and insured adults by using Medicare claims data.

Design: Longitudinal survey data and linked Medicare claims data were used to compare Medicare spending for beneficiaries age 65 to 74 years who were previously insured or previously uninsured before age 65 years. An inverse-probability-of-treatment weighting technique was used to adjust for fixed and time-varying sociodemographic and health characteristics before age 65 years. Condition-specific hospitalizations were compared, and their contribution to differences in Medicare spending was estimated.

Setting: Nationally representative Health and Retirement Study, 1992 to 2006.

Participants: 2951 continuously insured adults and 1616 adults who were continuously or intermittently uninsured before age 65 years.

Measurements: Mean adjusted annual Medicare spending (total and by type of service) and annual rates of condition-specific hospitalizations.

Results: Adjusted annual total Medicare spending was significantly higher for previously uninsured than previously insured adults ($5796 vs. $4773; difference, $1023 [95% CI, $29 to $2016]; P = 0.044). Among relevant clinical subgroups, previously uninsured adults had higher adjusted annual hospitalization rates than previously insured adults for complications related to cardiovascular disease or diabetes (9.1% vs. 6.4%; P = 0.002) and for joint replacements (2.5% vs. 1.3%; P = 0.006). Differences in these hospitalizations accounted for 65.7% of the $644 difference in annual Medicare inpatient spending between all previously uninsured and insured adults.

Limitation: Unobserved confounders could have explained spending differences.

Conclusion: Costs of expanded coverage before age 65 years may be partially offset by subsequent reductions in Medicare spending after age 65 years, particularly for uninsured adults with cardiovascular disease, diabetes, or severe arthritis.

Primary Funding Source: The Commonwealth Fund.

Editors' Notes
Context

  • Most proposals for health care reform in the United States provide insurance coverage for uninsured people.

Contribution

  • Medicare spends more for people who were uninsured before they become eligible for Medicare than for people who were insured before they become eligible for Medicare ($5796 vs. $4773 per year during the 10 years after joining Medicare).

Caution

  • Unobserved confounders might explain the difference.

Implication

  • Insurance coverage for uninsured adults might be a more valuable investment for the United States than previously thought.

—The Editors

Join the dialogue on health care reform. Comment on the perspectives published in Annals and offer ideas of your own. All thoughtful voices should be heard.

The burden of chronic disease has increased in the United States in the past decade (1), and many adults with chronic conditions lack health insurance (23). As rising health care costs and a slowing economy threaten to increase the number of uninsured adults (45) and thwart efforts to expand coverage (6), assessments of the value of health insurance for chronically ill adults may help to guide health care reform.

Growing evidence suggests that health insurance coverage benefits the health of adults with treatable conditions (78). Near-universal Medicare coverage after age 65 years increases use of health services (912), improves self-reported health trends (13), and reduces disparities in clinical measures of disease control (14), particularly for uninsured adults with cardiovascular disease or diabetes. Therefore, previously uninsured Medicare beneficiaries may have greater morbidity and require costlier care than if they had been continuously insured before age 65.

In a previous national study of adults aged 21 to 64 years, medical expenditures for uninsured adults who acquired insurance coverage were not subsequently higher than expenditures for continuously insured adults (15). However, uninsured adults in this study were predominantly young and healthy, only a small fraction of them gained coverage, and changes in insurance status may have been driven by individuals' choices and financial means. In contrast, nearly all Americans become eligible for Medicare at age 65 years. Moreover, because of a much higher prevalence of chronic disease, differences in Medicare spending between previously uninsured and insured adults may be much larger, but these differences have not been quantified by using Medicare claims. Consequently, these potential economic consequences to the Medicare program have not been considered in estimating the costs of expanding coverage for uninsured middle-aged adults (1617).

Medicare spending may be higher for previously uninsured adults with cardiovascular disease and diabetes if deficient care leads to irreversible complications before age 65 years (1822) or to elevated clinical risks that persist after age 65 years (2324). Uninsured adults may also delay costly elective procedures, such as joint replacements for severe arthritis, until they are eligible for Medicare. If differences in Medicare spending for previously uninsured and insured adults are driven by clinical conditions and services that reflect these hypothesized effects, then differences in spending after age 65 may be attributed to differences in insurance coverage before age 65 with greater confidence.

Using longitudinal national survey data linked with Medicare claims, we compared Medicare spending after age 65 for previously uninsured and insured adults. We expected hospitalizations for complications related to cardiovascular disease or diabetes and for joint replacements to be particularly influenced by previous insurance coverage. Therefore, we also compared rates of these condition-specific hospitalizations and estimated their contributions to differences in Medicare spending for previously uninsured and insured adults. On the basis of these analyses, we estimated reductions in Medicare spending from age 65 to 74 that could be achieved if all Americans currently 51 to 64 years of age were continuously insured before age 65.

Study Population

We analyzed data from the Health and Retirement Study, a nationally representative longitudinal survey (25). In 1992, this study enrolled noninstitutionalized adults and eligible spouses age 51 to 61 in the continental United States. Initial interviews conducted in 7705 households yielded 9749 participants (82% response rate). Biennial surveys were conducted through 2006, when participants were age 64 to 75. In each survey, participants who were eligible for Medicare were asked to provide their Medicare identification numbers for linkage to Medicare claims files available through 2005. Our study cohort included participants who were age 65 or older by 2005, were linked to their enrollment files, and were not continuously enrolled in Medicare managed care after age 65. We excluded adults who reported public insurance coverage in 1992 because insurance coverage before 1992 was not observed and preceding episodes of lack of insurance for some of these adults may have led to medical conditions, such as end-stage renal disease, that qualified them for public insurance and were not measured in the survey. Without this exclusion, higher subsequent health care spending for qualifying conditions present in 1992 would be inaccurately associated with a group uniformly classified as insured. In contrast, preceding insurance coverage was known for all subsequent transitions to public insurance, which were therefore included in the analysis.

Our study protocol was approved by the Harvard Medical School Committee on Human Studies and by the Privacy Board of the Centers for Medicare & Medicaid Services.

Study Variables

We used biennial survey data from 1992 to 2006 to assess longitudinal patterns of insurance coverage and sociodemographic and health characteristics before age 65. In each survey, participants answered detailed questions about sources of health insurance and, if insured, whether they lacked coverage at any time in the previous 2 years. As in earlier research (13), we classified participants as previously insured if they reported a continuous pattern of initial private coverage and subsequent private or public coverage in all surveys before age 65 and as previously uninsured if they reported continuous or intermittent uninsurance before age 65. After age 65, participants were classified as having supplemental insurance if they ever reported having employer-sponsored insurance, Medicaid, Medigap, or traditional Medicare with additional coverage. We determined vital status after age 65 and enrollment in Medicare Part B and Medicare managed care from Medicare denominator files.

For each participant, we used claims data from 1996 to 2005 to assess annual total Medicare spending after age 65 and spending in 7 distinct categories of service. We determined spending from reimbursements by Medicare to providers, inflated to constant 2006 U.S. dollars using the gross domestic product deflator (2627). For years in which adults were enrolled in Medicare managed care for less than 12 months, we inflated partial-year spending in traditional Medicare to annual estimates. We also determined annual utilization from claims data for 4 general services: hospital stays, outpatient institutional visits, physician office visits, and days in skilled-nursing facilities.

We assessed annual indicators of hospitalization for 2 prespecified categories of diagnosis-related groups (DRGs): complications related to cardiovascular disease or diabetes (DRGs 14 to 17, 36, 87, 103, 106, 107, 109 to 111, 113 to 118, 121 to 125, 127, 129 to 134, 138 to 140, 143 to 145, 294, 316, 317, 478, 479, 515 to 518, 524 to 527, and 533 to 536) and lower-extremity joint replacement (DRGs 209 and 471). These condition-specific hospitalizations were assessed for the entire study cohort and for 2 high-risk clinical subgroups: adults reporting cardiovascular disease (hypertension, heart disease, stroke) or diabetes by age 67 and adults reporting arthritis by age 67. We used this age threshold to allow adults with undiagnosed conditions to be diagnosed and report conditions in a biennial survey within 3 years of age eligibility for Medicare.

Statistical Analysis

Because intermittent uninsurance is common among near-elderly adults and predicts adverse health outcomes (13, 2830), we defined cohorts by longitudinal assessment of insurance coverage over multiple surveys before age 65 rather than by a single cross-sectional assessment. Our cohort definitions (continuously insured vs. continuously or intermittently uninsured) thus provided greater statistical power for comparisons of subsequent Medicare spending but also meant that we had to control for time-varying confounders. For example, uninsurance before age 65 may have caused health declines, in turn leading to higher Medicare spending, but declining health also may have caused uninsurance. If we did not control for health declines before the onset of uninsurance, we might mistakenly attribute higher Medicare spending to the effects of uninsurance before age 65 when higher spending was actually the result of earlier deteriorations in health.

A previous longitudinal analysis of these data demonstrated that coverage losses predicted subsequent health declines, but health declines did not predict coverage losses (31). Thus, effects of uninsurance on health are likely to predominate over the effects of health on uninsurance in this age group. Nevertheless, we used an inverse-probability-of-treatment weighting technique to adjust comparisons of previously uninsured and previously insured adults for both fixed and time-varying characteristics (Appendix) (3233). After the weighting adjustment, the likelihood of being uninsured at any given survey before age 65 was independent of the sociodemographic and health characteristics (listed in Table 1) observed up to that time. Because insured and uninsured adults had similar weighted distributions of health measures at the time of survey assessments before age 65, the weighting adjustment minimized differences in insurance status due to health declines before each survey while preserving differences in health that subsequently developed from differences in insurance status. Thus, we distinguished effects of uninsurance before age 65 on health from potentially confounding effects of health on uninsurance.

Table Jump PlaceholderTable 1.  Cohort Characteristics, by Insurance Coverage

In our main analyses, we sought to compare previously uninsured and insured adults who had equivalent access to care after age 65. Therefore, we also included an indicator of supplemental coverage when we constructed inverse-probability-of-treatment weights. Using a similar weighting method, we adjusted for survey and item nonresponse that led to missing Medicare identification numbers and claims data (34).

To estimate absolute differences between previously uninsured and insured adults, we fitted linear models predicting annual spending or utilization as a function of insurance coverage before age 65 and applied the weights we had constructed. For these analyses, we stratified the study cohort into adults with and without cardiovascular disease or diabetes by age 67 on the basis of self-reports. We also fitted logistic models for annual indicators of condition-specific hospitalization among the high-risk subgroups with cardiovascular disease or diabetes or arthritis. In our models of spending and utilization, we adjusted for geographic variation in spending by including fixed effects for Metropolitan Statistical Areas or non–Metropolitan Statistical Area counties because uninsured adults might have been disproportionately located in areas of higher or lower health care spending. We adjusted all analyses for the complex survey design (35) and repeated measures by using sampling weights and robust design-based variance estimators (36).

We assessed the contribution of condition-specific hospitalizations to differences in Medicare spending by comparing spending differences between previously uninsured and insured adults before and after adding annual counts of these hospitalizations as predictors to models. We also added the development of end-stage renal disease as an explanatory predictor.

For previously uninsured and insured adults with cardiovascular disease or diabetes, we expected differences in health care needs to persist after age 65 but to diminish with age as disease control, physical functioning, and risk for adverse cardiovascular outcomes improved for previously uninsured adults (1314). Therefore, in exploratory analyses we examined age profiles of Medicare spending and condition-specific hospitalizations to determine the ages at which differences narrowed, and we added to the models interactions between previous insurance coverage and these ages.

We conducted several sensitivity analyses. To determine whether differential selection into managed care may have biased our results, we further adjusted analyses using weights derived from a predictive model of continuous enrollment in managed care after age 65. Because health care provided by the Veterans Health Administration is not captured in Medicare claims, we excluded veterans. We also excluded observations for beneficiaries when they were not enrolled in Medicare Part B. For the rare instances in which Medicare was not the primary payer for Medicare-covered services (3.6% of beneficiary-years), we added reimbursements from other primary payers to spending measures and then repeated the main analyses. Finally, we compared annual rates of hospitalization for chronic obstructive pulmonary disease (COPD), a condition for which we hypothesized that uninsurance before age 65 would not lead to persistently greater health care needs after age 65 (Appendix).

All analyses were performed by using SAS statistical software, version 9.1 (SAS Institute, Cary, North Carolina). We report 2-sided P values without adjustment for multiple testing.

Role of the Funding Source

This study was funded by the Commonwealth Fund. The funding source had no role in the design, conduct, or reporting of the study or in the decision to submit the manuscript for publication.

Of the 9749 participants interviewed in 1992, we excluded 741 (7.6%) who died before age 65 and 971 (10.0%) who did not reach age 65 by 2005. Medicare identification numbers were not available for 1141 people (11.7%) who did not respond to surveys after age 65. Of the remaining 6896 participants, 5377 (78.0%) provided their identification numbers and were linked to their enrollment files. Rates of missing Medicare claims data due to survey and item nonresponse were similar for previously uninsured and insured adults (32.6% vs. 32.2%; P = 0.74). Of participants with linked claims files, we excluded 450 (8.4%) who had public insurance coverage in 1992 and 360 (6.7%) who were continuously enrolled in Medicare managed care after age 65. Of the remaining 4567 participants who were age 52 to 61 in 1992 and age 65 to 74 in 2005, 2951 (64.6%) reported continuous coverage, and 1616 (35.4%) reported being uninsured before age 65 (for 44% of the surveyed years on average).

Adults who were uninsured before age 65 differed significantly from continuously insured adults in many fixed and time-varying sociodemographic and health characteristics (Table 1). After age 65, previously uninsured adults were less likely to have supplemental insurance coverage than previously insured adults, but enrollment in Medicare managed care and Part B coverage did not differ between these groups (Table 1).

Mean adjusted annual Medicare spending was significantly higher for previously uninsured adults than for previously insured adults (Table 2); differences were most pronounced in the highest quartiles of their spending distributions (respective values for previously uninsured vs. previously insured adults: median, $707 vs. $706; 75th percentile, $3108 vs. $2608; 90th percentile, $12 463 vs. $9406; 95th percentile, $26 787 vs. $21 813). Differences in total Medicare spending were largely due to differences in inpatient and home health spending (Table 2) and were concentrated among the 66.7% of adults with cardiovascular disease or diabetes (Figure 1). These results did not substantially change in sensitivity analyses (data not shown). Descriptive plots and post hoc testing suggested that differences in total Medicare spending diminished after age 71 (P = 0.104 for comparison of age 65 through 71 with age 72 through 74).

Table Jump PlaceholderTable 2.  Adjusted Annual Per-Person Medicare Spending and Utilization After Age 65 Years, by Type of Service, Insurance Coverage Before Age 65, and History of Cardiovascular Disease or Diabetes
Grahic Jump Location
Figure 1.
Total Medicare spending, by age, insurance coverage before age 65 years, and history of CVD or diabetes.

Mean adjusted total Medicare spending is plotted by age and insurance coverage before age 65 for adults with (top) and adults without (bottom) a reported diagnosis of hypertension, heart disease, stroke, or diabetes by age 67. Annual Medicare spending after age 65 was significantly higher for previously uninsured adults than previously insured adults with CVD or diabetes, and this difference was particularly large for those age 65 to 71 (adjusted difference, $1400; P = 0.027) (top). Annual Medicare spending after age 65 did not differ for previously uninsured and insured adults without CVD or diabetes (bottom). Plots were truncated from age 74 to age 72 because of smaller sample sizes for estimates after age 72. Error bars are SEs. CVD = cardiovascular disease.

Grahic Jump Location

Previously uninsured adults had more hospital stays and outpatient institutional visits annually than previously insured adults, but not more physician office visits or days in skilled-nursing facilities (Table 2). Among adults with cardiovascular disease or diabetes, previously uninsured adults were more likely to be hospitalized for complications related to cardiovascular disease or diabetes, and specifically for myocardial infarction, heart failure, or stroke (Table 3). Descriptive plots suggested that differences in hospitalization risk narrowed with age (Figure 2), but this narrowing was not statistically significant. Among adults with arthritis, previously uninsured adults were more likely than previously insured adults to be hospitalized for joint replacement (Table 3).

Table Jump PlaceholderTable 3.  Condition-Specific Hospitalizations After Age 65 Years for Previously Uninsured and Insured Adults
Grahic Jump Location
Figure 2.
Annual rates of hospitalization for complications related to cardiovascular disease or diabetes, by age and insurance coverage before age 65 years.

Among adults with cardiovascular disease or diabetes, annual rates of hospitalization after age 65 for complications related to these conditions were significantly higher for previously uninsured adults than previously insured adults. This increased risk in previously uninsured adults diminished with increasing age, but not significantly (P = 0.38 for interaction between previous insurance coverage and age). Error bars are SEs.

Grahic Jump Location

Of 5485 hospital admissions after age 65, 1764 (32.2%) were for complications related to cardiovascular disease or diabetes and 274 (5.0%) were for joint replacements. On the basis of annual counts of hospitalizations for complications related to cardiovascular disease or diabetes among previously uninsured and insured adults (0.16 per person vs. 0.10 per person; P = 0.04), 1 more of these hospitalizations occurred annually after age 65 for every 17 (1/0.06) previously uninsured adults with these conditions. Each of these hospitalizations was associated with an additional $11 000 in annual inpatient Medicare spending on average, as estimated when annual hospitalization counts were added as predictors to spending models. After adjustment for hospitalizations for complications related to cardiovascular disease or diabetes and for joint replacements, the adjusted difference in annual inpatient spending for previously uninsured and insured adults was reduced by 65.7% and was no longer statistically significant (from $644 per person to $221 per person; P = 0.29). Similarly, adjustment for condition-specific hospitalizations and development of end-stage renal disease reduced the adjusted difference in annual total Medicare spending by 62.9% (from $1023 per person to $380 per person; P = 0.25).

In this nationally representative cohort, adjusted Medicare spending was significantly higher for previously uninsured adults than previously insured adults. Consistent with clinical evidence (1824), differences in inpatient spending were largest for adults with cardiovascular disease or diabetes because previously uninsured adults were substantially more likely than previously insured adults to be hospitalized after age 65 for costly complications related to these treatable conditions. Adjusted rates of hospitalization for joint replacement were also significantly higher for previously uninsured adults with arthritis, suggesting that uninsured adults may delay costly procedures that enhance physical functioning and quality of life (3742). Despite accounting for just 37% of admissions, hospitalizations for complications related to cardiovascular disease or diabetes and joint replacements accounted for nearly two thirds of the adjusted difference in inpatient spending for all previously uninsured and insured adults.

Higher Medicare spending after age 65 for previously uninsured adults suggests that the costs of expanding coverage for these adults may be lower than expected. For example, a recent analysis estimated that universal coverage would increase total annual medical spending for adults age 19 to 64 by $113 billion (16). However, this estimate did not account for potential reductions in subsequent spending after age 65 due to improved health from prevention of complications or to earlier use of elective procedures, such as joint replacements.

For adults who are currently age 51 to 64, providing coverage to the uninsured for an average of 4 years (based on our study cohort) to achieve universal continuous coverage through age 65 would increase health care spending before age 65 by an estimated $197 billion (Appendix). Projecting from our findings, if the current population of adults age 51 to 64 had continuous insurance coverage from now through age 65, the present value of subsequent reductions in Medicare spending between age 65 and 74 for these adults would be approximately $98 billion (Appendix). Thus, from a societal perspective, subsequent reductions in Medicare spending after age 65 may offset almost half of this increased spending before age 65. Because Medicare pays for only 49% of all health care spending for elderly adults (43), the reduction in total medical spending after age 65 may be even greater.

Our calculations assumed that expanding coverage would not affect mortality for uninsured near-elderly adults, but gains in life expectancy may be substantial for these persons (4446). We compared increases in health care spending associated with provision of insurance coverage before age 65 to subsequent decreases in spending after age 65 associated with better health. However, decreased annual spending due to better health may also be matched by increased lifetime spending due to greater longevity (47). Thus, survival gains for uninsured adults would add years of spending before and after age 65 to our calculated costs of expanded coverage before age 65 and thereby reduce the relative size of partially offsetting decreases in Medicare spending after age 65, particularly if these survival gains disproportionately increased the number of severely ill adults living beyond age 65. However, if providing insurance coverage to uninsured adults added valuable years to the lives of many near-elderly and elderly Americans, this would be an important benefit for policymakers to consider.

Strengths of our study included detailed longitudinal survey data linked to Medicare claims, rigorous methods to adjust differences in Medicare spending for both fixed and time-varying confounders before age 65, and assessments of clinically relevant hospitalizations as contributors to these spending differences. Our findings are consistent with a previous analysis of self-reported hospitalizations after age 65 (10) and quasi-experimental studies suggesting distinctive health benefits of insurance coverage for near-elderly adults, particularly those with cardiovascular disease or diabetes (1314, 46, 4849). For adults with these chronic conditions, improvements in blood pressure, blood glucose, and cholesterol control associated with gaining coverage (14) may substantially reduce subsequent annual health care costs.

Our study had several limitations. First, exploratory analyses suggested that differences in Medicare spending and complications related to cardiovascular disease or diabetes eventually narrowed with age, but we had limited statistical power to discriminate between short-term and long-term effects of uninsurance on subsequent utilization and spending. The age-related narrowing that was evident in descriptive plots of hospitalizations for complications of cardiovascular disease or diabetes was consistent with relative health improvements for previously uninsured adults after acquisition of Medicare coverage (13). This narrowing was also consistent with previously demonstrated legacy effects of poorly controlled blood pressure and blood glucose among adults with diabetes, some of which persist indefinitely after better disease control is achieved (for example, higher risk for myocardial infarction due to previous lack of metformin therapy) and some of which diminish gradually over several years (for example, higher risk for stroke or microvascular complications due to poor previous blood pressure control) (2324). However, greater mortality rates among previously uninsured adults may have also contributed to narrowing differences (44). Because of limited sample sizes, we also could not precisely assess the contributions of other key conditions, such as cancer, to differences in Medicare spending between previously uninsured and insured adults.

Second, we could not observe health care utilization or spending for Medicare beneficiaries enrolled in managed care plans. However, managed care enrollment did not differ between previously uninsured and insured adults, and results were similar in analyses weighted to reflect the likelihood of continuous enrollment in Medicare managed care based on characteristics before age 65. Findings were similarly robust to sensitivity analyses addressing enrollment in Part B, veterans' benefits, and Medicare-covered services reimbursed by other primary payers.

Third, preexisting differences in unobserved characteristics between insured and uninsured near-elderly adults, such as health care preferences or disease severity, may have explained subsequent differences in adjusted Medicare spending. However, uninsured adults are likely to have weaker preferences for health care (5051), and their lower demand for care might persist after age 65. Therefore, differences in Medicare spending for previously uninsured and insured adults with equal preferences for care may be larger than our results suggest. Furthermore, if higher Medicare spending for previously uninsured adults with cardiovascular disease or diabetes was driven primarily by unmeasured differences in disease severity that were not caused by uninsurance before age 65, their elevated health care needs should have persisted indefinitely after acquisition of Medicare coverage. Yet, their risk for hospitalization for complications related to these conditions appeared to diminish relative to previously insured adults after several years of eligibility, coinciding with improved trends in health status demonstrated in a previous study (13).

Finally, our inverse-probability-of-treatment weighting may have underestimated effects of uninsurance on subsequent Medicare spending if measured differences in health at the time of surveys were due to preceding differences in coverage.

Our findings have important policy implications. Medicare spending between age 65 and 74 was significantly higher for previously uninsured adults than for previously insured adults who were otherwise similar in numerous measured characteristics before age 65. Greater spending for previously uninsured adults was largely explained by hospitalizations for complications related to cardiovascular disease or diabetes and for joint replacements. Thus, extending insurance to uninsured adults may result in substantial health gains for many older working-age adults, and subsequent reductions in Medicare spending after age 65 may partially offset increased spending from expanded coverage before age 65. These benefits suggest that health insurance coverage for uninsured adults older than 50 years would be a more valuable investment for the United States than previously thought.

Paez KA, Zhao L, Hwang W.  Rising out-of-pocket spending for chronic conditions: a ten-year trend. Health Aff (Millwood). 2009; 28:15-25. PubMed
CrossRef
 
Ayanian JZ, Weissman JS, Schneider EC, Ginsburg JA, Zaslavsky AM.  Unmet health needs of uninsured adults in the United States. JAMA. 2000; 284:2061-9. PubMed
 
Wilper AP, Woolhandler S, Lasser KE, McCormick D, Bor DH, Himmelstein DU.  A national study of chronic disease prevalence and access to care in uninsured U.S. adults. Ann Intern Med. 2008; 149:170-6. PubMed
 
Chernew M, Cutler DM, Keenan PS.  Increasing health insurance costs and the decline in insurance coverage. Health Serv Res. 2005; 40:1021-39. PubMed
 
Emanuel EJ.  The cost-coverage trade-off: “it's health care costs, stupid”. JAMA. 2008; 299:947-9. PubMed
 
Gruber J.  Universal health insurance coverage or economic relief—a false choice. N Engl J Med. 2009; 360:437-9. PubMed
 
Institute of Medicine Committee on the Consequences of Health Insurance Status.  America's Uninsured Crisis: Consequences for Health and Health Care. Washington, DC: National Academies Pr; 2009.
 
McWilliams JM.  Health consequences of uninsurance among adults in the United States: recent evidence and implications. Milbank Q. 2009; 87:443-94. PubMed
 
Card D, Dobkin C, Maestas N.  The impact of nearly universal insurance coverage on health care utilization: evidence from Medicare. American Economic Review. 2008; 98:2242-2258.
 
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ.  Use of health services by previously uninsured Medicare beneficiaries. N Engl J Med. 2007; 357:143-53. PubMed
 
Decker SL.  Medicare and the health of women with breast cancer. Journal of Human Resources. 2005; 40:948-68.
 
McWilliams JM, Zaslavsky AM, Meara E, Ayanian JZ.  Impact of Medicare coverage on basic clinical services for previously uninsured adults. JAMA. 2003; 290:757-64. PubMed
 
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ.  Health of previously uninsured adults after acquiring Medicare coverage. JAMA. 2007; 298:2886-94. PubMed
 
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ.  Differences in control of cardiovascular disease and diabetes by race, ethnicity, and education: U.S. trends from 1999 to 2006 and effects of medicare coverage. Ann Intern Med. 2009; 150:505-15. PubMed
 
Ward L, Franks P.  Changes in health care expenditure associated with gaining or losing health insurance. Ann Intern Med. 2007; 146:768-74. PubMed
 
Hadley J, Holahan J, Coughlin T, Miller D.  Covering the uninsured in 2008: current costs, sources of payment, and incremental costs. Health Aff (Millwood). 2008; 27:w399-415. PubMed
 
Sheils J, Chen YJ.  Medicare Buy-In Options: Estimating Coverage and Costs. New York: The Commonwealth Fund; 2001. Report no. 441.
 
Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults.  Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA. 2001; 285:2486-97. PubMed
 
Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr., et al. National Heart, Lung, and Blood Institute Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure.  The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA. 2003; 289:2560-72. PubMed
 
UK Prospective Diabetes Study (UKPDS) Group.  Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet. 1998; 352:854-65. PubMed
 
UK Prospective Diabetes Study Group.  Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes: UKPDS 38. BMJ. 1998; 317:703-13. PubMed
 
Gaede P, Lund-Andersen H, Parving HH, Pedersen O.  Effect of a multifactorial intervention on mortality in type 2 diabetes. N Engl J Med. 2008; 358:580-91. PubMed
 
Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA.  10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med. 2008; 359:1577-89. PubMed
 
Holman RR, Paul SK, Bethel MA, Neil HA, Matthews DR.  Long-term follow-up after tight control of blood pressure in type 2 diabetes. N Engl J Med. 2008; 359:1565-76. PubMed
 
Health and Retirement Study.  1992-2006 final release public use datasets. Ann Arbor: University of Michigan; 2009.
 
Newhouse JP.  An iconoclastic view of health cost containment. Health Aff (Millwood). 1993; 12:Suppl152-71. PubMed
 
.  Economic Report of the President, Transmitted to the Congress February 2008. Washington, DC: U.S. Government Printing Office; 2008.
 
Baker DW, Sudano JJ.  Health insurance coverage during the years preceding medicare eligibility. Arch Intern Med. 2005; 165:770-6. PubMed
 
Baker DW, Sudano JJ, Albert JM, Borawski EA, Dor A.  Lack of health insurance and decline in overall health in late middle age. N Engl J Med. 2001; 345:1106-12. PubMed
 
Sudano JJ Jr, Baker DW.  Intermittent lack of health insurance coverage and use of preventive services. Am J Public Health. 2003; 93:130-7. PubMed
 
Baker DW, Sudano JJ, Albert JM, Borawski EA, Dor A.  Loss of health insurance and the risk for a decline in self-reported health and physical functioning. Med Care. 2002; 40:1126-31. PubMed
 
Hernán MA, Brumback B, Robins JM.  Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology. 2000; 11:561-70. PubMed
 
Robins JM, Hernán MA, Brumback B.  Marginal structural models and causal inference in epidemiology. Epidemiology. 2000; 11:550-60. PubMed
 
Rubin DB, Little RJ.  Statistical Analysis With Missing Data. 2nd ed. Hoboken, NJ: J Wiley & Sons; 2002.
 
Heeringa SG, Connor JH.  Technical description of the Health and Retirement Survey sample design. Vol. 2009. Ann Arbor, MI: Univ of Michigan; 1995. Accessed athttp://hrsonline.isr.umich.edu/sitedocs/userg/HRSSAMP.pdfon 23 July 2009.
 
Lohr SL.  Sampling: Design and Analysis. Pacific Grove, CA: Duxbury Pr; 1999.
 
Laupacis A, Bourne R, Rorabeck C, Feeny D, Wong C, Tugwell P. et al.  The effect of elective total hip replacement on health-related quality of life. J Bone Joint Surg Am. 1993; 75:1619-26. PubMed
 
NIH Consensus Development Panel on Total Hip Replacement.  NIH consensus conference: Total hip replacement. JAMA. 1995; 273:1950-6. PubMed
 
Hawker G, Wright J, Coyte P, Paul J, Dittus R, Croxford R. et al.  Health-related quality of life after knee replacement. J Bone Joint Surg Am. 1998; 80:163-73. PubMed
 
Heck DA, Robinson RL, Partridge CM, Lubitz RM, Freund DA.  Patient outcomes after knee replacement. Clin Orthop Relat Res. 1998; 93-110. PubMed
 
van Essen GJ, Chipchase LS, O'Connor D, Krishnan J.  Primary total knee replacement: short-term outcomes in an Australian population. J Qual Clin Pract. 1998; 18:135-42. PubMed
 
Hamel MB, Toth M, Legedza A, Rosen MP.  Joint replacement surgery in elderly patients with severe osteoarthritis of the hip or knee: decision making, postoperative recovery, and clinical outcomes. Arch Intern Med. 2008; 168:1430-40. PubMed
 
Centers for Medicare & Medicaid Services, Office of the Actuary, National Health Statistics Group.  Personal health care spending by age group and source of payment, calendar year 2004. Accessed atwww.cms.hhs.gov/NationalHealthExpendData/downloads/2004-age-tables.pdfon 23 July 2009.
 
McWilliams JM, Zaslavsky AM, Meara E, Ayanian JZ.  Health insurance coverage and mortality among the near-elderly. Health Aff (Millwood). 2004; 23:223-33. PubMed
 
Baker DW, Sudano JJ, Durazo-Arvizu R, Feinglass J, Witt WP, Thompson J.  Health insurance coverage and the risk of decline in overall health and death among the near elderly, 1992-2002. Med Care. 2006; 44:277-82. PubMed
 
Card D, Dobkin C, Maestas N.  Does Medicare save lives? Quarterly Journal of Economics. 2009; 124:531-96.
 
Lubitz J, Cai L, Kramarow E, Lentzner H.  Health, life expectancy, and health care spending among the elderly. N Engl J Med. 2003; 349:1048-55. PubMed
 
Dor A, Sudano J, Baker DW.  The effect of private insurance on the health of older, working age adults: evidence from the health and retirement study. Health Serv Res. 2006; 41:759-87. PubMed
 
Hadley J, Waidmann T.  Health insurance and health at age 65: implications for medical care spending on new Medicare beneficiaries. Health Serv Res. 2006; 41:429-51. PubMed
 
Monheit AC, Vistnes JP.  Health insurance enrollment decisions: preferences for coverage, worker sorting, and insurance take-up. Inquiry. 2008; 45:153-67. PubMed
 
Levy H, DeLeire T.  What do people buy when they don't buy health insurance and what does that say about why they are uninsured? Inquiry. 2008; 45:365-79. PubMed
 
Cole SR, Hernán MA.  Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008; 168:656-64. PubMed
 
Bang H, Robins JM.  Doubly robust estimation in missing data and causal inference models. Biometrics. 2005; 61:962-73. PubMed
 
Cassel CM, Sarndal CE, Wretman JH.  Some uses of statistical models in connection with the nonresponse problem. Madow WG, Olkin I Incomplete Data in Sample Surveys III. Symposium on Incomplete Data, Proceedings. New York: Academic Pr; 1983.
 
Czajka JL, Hirabayashi SM, Little RJA, Rubin DB.  Projecting from advance data using propensity modeling: an application to income and tax statistics. Journal of Business and Economics Statistics. 1992; 10:117-131.
 
Kalton G, Piesse A.  Survey research methods in evaluation and case-control studies. Stat Med. 2007; 26:1675-87. PubMed
 
Lung Health Study Research Group.  Effect of inhaled triamcinolone on the decline in pulmonary function in chronic obstructive pulmonary disease. N Engl J Med. 2000; 343:1902-9. PubMed
 
Anthonisen NR, Connett JE, Kiley JP, Altose MD, Bailey WC, Buist AS. et al.  Effects of smoking intervention and the use of an inhaled anticholinergic bronchodilator on the rate of decline of FEV1. The Lung Health Study. JAMA. 1994; 272:1497-505. PubMed
 
Burge PS, Calverley PM, Jones PW, Spencer S, Anderson JA, Maslen TK.  Randomised, double blind, placebo controlled study of fluticasone propionate in patients with moderate to severe chronic obstructive pulmonary disease: the ISOLDE trial. BMJ. 2000; 320:1297-303. PubMed
 
Calverley PM, Anderson JA, Celli B, Ferguson GT, Jenkins C, Jones PW, et al. TORCH Investigators.  Salmeterol and fluticasone propionate and survival in chronic obstructive pulmonary disease. N Engl J Med. 2007; 356:775-89. PubMed
 
Pauwels RA, Löfdahl CG, Laitinen LA, Schouten JP, Postma DS, Pride NB. et al.  Long-term treatment with inhaled budesonide in persons with mild chronic obstructive pulmonary disease who continue smoking. European Respiratory Society Study on Chronic Obstructive Pulmonary Disease. N Engl J Med. 1999; 340:1948-53. PubMed
 
Reilly JJ.  COPD and declining FEV1—time to divide and conquer? [Editorial]. N Engl J Med. 2008; 359:1616-8. PubMed
 
Tashkin DP, Celli B, Senn S, Burkhart D, Kesten S, Menjoge S, et al. UPLIFT Study Investigators.  A 4-year trial of tiotropium in chronic obstructive pulmonary disease. N Engl J Med. 2008; 359:1543-54. PubMed
 
DeNavas-Walt C, Proctor BD, Smith J.  Income, poverty, and health insurance coverage in the United States: 2007. Washington, DC: U.S. Census Bureau; 2008. Report no. P60-235.
 
Meara E, White C, Cutler DM.  Trends in medical spending by age, 1963-2000. Health Aff (Millwood). 2004; 23:176-83. PubMed
 
Appendix
Inverse-Probability-of-Treatment Weighting to Control for Time-Varying Confounders

We used logistic regression to model insurance status at each biennial survey before age 65 as a function of numerous characteristics (listed in Table 1) assessed in the same survey and insurance status from the preceding survey. The cumulative products of predicted probabilities from this model yielded participants' probabilities of having their observed longitudinal coverage pattern given these fixed and time-varying characteristics. Individual weights were constructed from the inverse of these cumulative probabilities, and their variance was reduced by estimating stabilized weights as described elsewhere (3233). In recommended diagnostic checks (52), we examined the distribution of these stabilized weights to confirm that there were no extreme values and that the mean was close to 1.0 (mean, 1.0098). In a sensitivity analysis, we also truncated the distribution of these weights to the 5th and 95th percentiles. Estimated differences in Medicare spending remained statistically significant after this truncation. Finally, to check the robustness of weighted estimates to alternative modeling approaches (53), we added to spending models all baseline sociodemographic and health characteristics from 1992 that were already included in the predictive model used to derive inverse-probability-of-treatment weights.

Assuming no unmeasured confounders, these weights approximated sequential randomization of insurance status at each survey before age 65, conditional on concurrent characteristics. Thus, whereas standard regression or propensity-score methods would allow comparisons of Medicare spending after age 65 by insurance status at a specific age or time before age 65 (for example, insured vs. uninsured at baseline in 1992) with adjustment for observed characteristics at that time, this weighting approach facilitated adjusted comparisons of Medicare spending by entire histories of insurance coverage from age 52 through age 64. We could therefore use simple weighted models (marginal structural models) to compare adults who were continuously insured with adults who were continuously or intermittently uninsured before age 65 without bias from observed time-varying confounders.

We confirmed that fixed and time-varying characteristics in the weighted population before age 65 were similar by insurance status at the time of surveys. To do so, we weighted participants' successive self-reports before age 65 (for example, of general health status or household income) by their inverse-probability-of-treatment weights cumulatively determined through each corresponding survey (the last set of which was used in main comparisons of Medicare spending by previous coverage, as described above) (33). We then conducted pooled comparisons of characteristics from all surveys before age 65 by insurance status assessed in corresponding surveys, and compared differences before and after applying these weights. Relative to unweighted differences, weighted differences between insured and uninsured adults were substantially reduced for all fixed and time-varying sociodemographic and health characteristics. Differences between insured and uninsured adults in general health status, change in general health, mobility, agility, difficulties with activities of daily living, and smoking status were reduced by 66% to 94%. Large differences in household income were almost entirely eliminated, and racial differences were reversed so that uninsured adults were slightly more likely to be white. Each of 14 other categorical sociodemographic and health characteristics that were considered differed by no more than 1 to 2 percentage points in weighted comparisons. Thus, inverse-probability-of-treatment weights evenly balanced fixed and time-varying characteristics between insured and uninsured adults at the time of surveys before age 65 so that differences in health insurance status were not related to preexisting differences in these observed characteristics.

In the weighted population, uninsurance at any survey wave before age 65 was still permitted to predict subsequent health declines that might accumulate and lead to greater Medicare spending after age 65. To characterize some of these potentially mediating effects, we compared health characteristics at the last survey wave before age 65 by longitudinal assessments of coverage (continuously insured adults vs. continuously or intermittently uninsured adults) and weighted these comparisons by participants' last set of inverse-probability-of-treatment weights before age 65 (those used in our main analyses of Medicare spending). Relative to the diminished health differences described above, these weighted differences were 67% to 195% greater, indicating that adults who were uninsured at any survey wave after age 52 subsequently developed worse general health status, more functional limitations, and more difficulties with activities of daily living by their final survey before age 65 than adults who were continuously insured. Thus, these analyses confirmed that our inverse-probability-of-treatment weighting effectively minimized confounding effects of observed time-varying health characteristics on insurance coverage while allowing mediating effects of insurance coverage on health.

We combined sampling, nonresponse, and treatment weights in the following manner (5456). First, nonresponse weights were estimated by using a predictive model weighted by survey sampling weights. Second, inverse-probability-of-treatment weights were estimated by using a predictive model weighted by the product of nonresponse and sampling weights. Third, sampling, nonresponse, and treatment weights were multiplied to obtain combined weights used for marginal structural models of the effects of insurance coverage before age 65 on Medicare spending after age 65.

Hospitalizations for COPD Exacerbations

In contrast to those with cardiovascular disease or diabetes, previously uninsured and insured adults with COPD may have similar health care needs after age 65 as a result of this condition because treatments for COPD improve symptoms and prevent reversible complications but generally do not alter the rate of disease progression. Specifically, clinical trials suggest that underuse of inhaled therapies by uninsured adults with COPD would increase their risk for exacerbations while they are uninsured but would not accelerate declining lung function or lead to irreversible complications (5763). Thus, poor control of this condition before age 65 should not lead to persistently greater health care needs after age 65 if Medicare effectively provides access to appropriate care because regular use of these treatments quickly reduces the risk for exacerbations. We compared adjusted annual rates of hospitalization for COPD exacerbation (DRG 88) among previously uninsured and insured adults who reported a diagnosis of COPD by age 67 or were active smokers at baseline. As hypothesized, previously uninsured adults were not more likely than previously insured adults in this high-risk group to be hospitalized for COPD exacerbations (adjusted odds ratio, 0.52 [CI, 0.27 to 1.01]); P = 0.053). The fact that differences in Medicare spending varied across conditions in a clinically predicted manner strengthens the attribution of higher spending after age 65 to uninsurance before age 65. Furthermore, the non–statistically significant lower rate of COPD exacerbations for previously uninsured adults suggests that all else being equal, they may have had weaker preferences for inpatient services, which would tend to bias effects of uninsurance on subsequent Medicare spending toward the null.

Estimation of Medicare Cost Offset

On the basis of the current number of adults age 51 to 64 in the United States (approximately 50.8 million) (64) and the proportions of adults in our study cohort who were continuously or intermittently uninsured (32.0%) and who died (7.6%) before age 65, approximately 15 million adults age 51 to 64 are expected to be uninsured at some point before age 65 and to survive to age 65. For each age from 51 to 64, we calculated the present value of the cumulative predicted difference in Medicare spending from age 65 to 74 between a previously uninsured and insured adult, using a discount rate of 3% and the adjusted annual spending difference estimated in our study ($1023 per person in 2006 dollars). These cumulative differences were adjusted to reflect mortality after age 65 among previously uninsured adults (2.4% annually on average). Differences ranged from $5371 to $7888 because future Medicare spending was discounted more heavily for younger adults than older adults and averaged $6556 per previously uninsured adult. Assuming that the 15 million adults expected to be uninsured before age 65 are evenly distributed across current ages of 51 to 64, the present value of the estimated reduction in subsequent Medicare spending between age 65 and 74 for these adults would total approximately $98 billion if they were continuously insured before age 65.

On the basis of previously published estimates of differences in adjusted annual per capita medical spending for insured and uninsured adults age 19 to 64 ($1948 per person in 2008 dollars) (16), the proportion of uninsured adults who are age 50 to 64 (17.9%) (16), and per capita spending for near-elderly adults relative to younger adults (approximately 2 times higher) (65), we estimated that providing coverage for uninsured adults over age 50 would increase annual health care spending for these adults by $3304 per person (in 2008 dollars). Assuming that 16.3 million adults age 51 to 64 are or will be uninsured before age 65 (50.8 million × 0.32) for an average of 4 years (based on our study cohort), providing continuous coverage for these adults would increase health care spending before age 65 by an estimated $53.7 billion per year for 4 years. Deflating this annual amount to 2006 dollars and discounting 4 consecutive years of spending to present value yielded an estimated $197 billion in increased spending before age 65 if the current population of adults age 51 to 64 had continuous insurance coverage through age 65.

Figures

Grahic Jump Location
Figure 1.
Total Medicare spending, by age, insurance coverage before age 65 years, and history of CVD or diabetes.

Mean adjusted total Medicare spending is plotted by age and insurance coverage before age 65 for adults with (top) and adults without (bottom) a reported diagnosis of hypertension, heart disease, stroke, or diabetes by age 67. Annual Medicare spending after age 65 was significantly higher for previously uninsured adults than previously insured adults with CVD or diabetes, and this difference was particularly large for those age 65 to 71 (adjusted difference, $1400; P = 0.027) (top). Annual Medicare spending after age 65 did not differ for previously uninsured and insured adults without CVD or diabetes (bottom). Plots were truncated from age 74 to age 72 because of smaller sample sizes for estimates after age 72. Error bars are SEs. CVD = cardiovascular disease.

Grahic Jump Location
Grahic Jump Location
Figure 2.
Annual rates of hospitalization for complications related to cardiovascular disease or diabetes, by age and insurance coverage before age 65 years.

Among adults with cardiovascular disease or diabetes, annual rates of hospitalization after age 65 for complications related to these conditions were significantly higher for previously uninsured adults than previously insured adults. This increased risk in previously uninsured adults diminished with increasing age, but not significantly (P = 0.38 for interaction between previous insurance coverage and age). Error bars are SEs.

Grahic Jump Location

Tables

Table Jump PlaceholderTable 1.  Cohort Characteristics, by Insurance Coverage
Table Jump PlaceholderTable 2.  Adjusted Annual Per-Person Medicare Spending and Utilization After Age 65 Years, by Type of Service, Insurance Coverage Before Age 65, and History of Cardiovascular Disease or Diabetes
Table Jump PlaceholderTable 3.  Condition-Specific Hospitalizations After Age 65 Years for Previously Uninsured and Insured Adults

References

Paez KA, Zhao L, Hwang W.  Rising out-of-pocket spending for chronic conditions: a ten-year trend. Health Aff (Millwood). 2009; 28:15-25. PubMed
CrossRef
 
Ayanian JZ, Weissman JS, Schneider EC, Ginsburg JA, Zaslavsky AM.  Unmet health needs of uninsured adults in the United States. JAMA. 2000; 284:2061-9. PubMed
 
Wilper AP, Woolhandler S, Lasser KE, McCormick D, Bor DH, Himmelstein DU.  A national study of chronic disease prevalence and access to care in uninsured U.S. adults. Ann Intern Med. 2008; 149:170-6. PubMed
 
Chernew M, Cutler DM, Keenan PS.  Increasing health insurance costs and the decline in insurance coverage. Health Serv Res. 2005; 40:1021-39. PubMed
 
Emanuel EJ.  The cost-coverage trade-off: “it's health care costs, stupid”. JAMA. 2008; 299:947-9. PubMed
 
Gruber J.  Universal health insurance coverage or economic relief—a false choice. N Engl J Med. 2009; 360:437-9. PubMed
 
Institute of Medicine Committee on the Consequences of Health Insurance Status.  America's Uninsured Crisis: Consequences for Health and Health Care. Washington, DC: National Academies Pr; 2009.
 
McWilliams JM.  Health consequences of uninsurance among adults in the United States: recent evidence and implications. Milbank Q. 2009; 87:443-94. PubMed
 
Card D, Dobkin C, Maestas N.  The impact of nearly universal insurance coverage on health care utilization: evidence from Medicare. American Economic Review. 2008; 98:2242-2258.
 
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ.  Use of health services by previously uninsured Medicare beneficiaries. N Engl J Med. 2007; 357:143-53. PubMed
 
Decker SL.  Medicare and the health of women with breast cancer. Journal of Human Resources. 2005; 40:948-68.
 
McWilliams JM, Zaslavsky AM, Meara E, Ayanian JZ.  Impact of Medicare coverage on basic clinical services for previously uninsured adults. JAMA. 2003; 290:757-64. PubMed
 
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ.  Health of previously uninsured adults after acquiring Medicare coverage. JAMA. 2007; 298:2886-94. PubMed
 
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ.  Differences in control of cardiovascular disease and diabetes by race, ethnicity, and education: U.S. trends from 1999 to 2006 and effects of medicare coverage. Ann Intern Med. 2009; 150:505-15. PubMed
 
Ward L, Franks P.  Changes in health care expenditure associated with gaining or losing health insurance. Ann Intern Med. 2007; 146:768-74. PubMed
 
Hadley J, Holahan J, Coughlin T, Miller D.  Covering the uninsured in 2008: current costs, sources of payment, and incremental costs. Health Aff (Millwood). 2008; 27:w399-415. PubMed
 
Sheils J, Chen YJ.  Medicare Buy-In Options: Estimating Coverage and Costs. New York: The Commonwealth Fund; 2001. Report no. 441.
 
Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults.  Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA. 2001; 285:2486-97. PubMed
 
Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr., et al. National Heart, Lung, and Blood Institute Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure.  The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA. 2003; 289:2560-72. PubMed
 
UK Prospective Diabetes Study (UKPDS) Group.  Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet. 1998; 352:854-65. PubMed
 
UK Prospective Diabetes Study Group.  Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes: UKPDS 38. BMJ. 1998; 317:703-13. PubMed
 
Gaede P, Lund-Andersen H, Parving HH, Pedersen O.  Effect of a multifactorial intervention on mortality in type 2 diabetes. N Engl J Med. 2008; 358:580-91. PubMed
 
Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA.  10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med. 2008; 359:1577-89. PubMed
 
Holman RR, Paul SK, Bethel MA, Neil HA, Matthews DR.  Long-term follow-up after tight control of blood pressure in type 2 diabetes. N Engl J Med. 2008; 359:1565-76. PubMed
 
Health and Retirement Study.  1992-2006 final release public use datasets. Ann Arbor: University of Michigan; 2009.
 
Newhouse JP.  An iconoclastic view of health cost containment. Health Aff (Millwood). 1993; 12:Suppl152-71. PubMed
 
.  Economic Report of the President, Transmitted to the Congress February 2008. Washington, DC: U.S. Government Printing Office; 2008.
 
Baker DW, Sudano JJ.  Health insurance coverage during the years preceding medicare eligibility. Arch Intern Med. 2005; 165:770-6. PubMed
 
Baker DW, Sudano JJ, Albert JM, Borawski EA, Dor A.  Lack of health insurance and decline in overall health in late middle age. N Engl J Med. 2001; 345:1106-12. PubMed
 
Sudano JJ Jr, Baker DW.  Intermittent lack of health insurance coverage and use of preventive services. Am J Public Health. 2003; 93:130-7. PubMed
 
Baker DW, Sudano JJ, Albert JM, Borawski EA, Dor A.  Loss of health insurance and the risk for a decline in self-reported health and physical functioning. Med Care. 2002; 40:1126-31. PubMed
 
Hernán MA, Brumback B, Robins JM.  Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology. 2000; 11:561-70. PubMed
 
Robins JM, Hernán MA, Brumback B.  Marginal structural models and causal inference in epidemiology. Epidemiology. 2000; 11:550-60. PubMed
 
Rubin DB, Little RJ.  Statistical Analysis With Missing Data. 2nd ed. Hoboken, NJ: J Wiley & Sons; 2002.
 
Heeringa SG, Connor JH.  Technical description of the Health and Retirement Survey sample design. Vol. 2009. Ann Arbor, MI: Univ of Michigan; 1995. Accessed athttp://hrsonline.isr.umich.edu/sitedocs/userg/HRSSAMP.pdfon 23 July 2009.
 
Lohr SL.  Sampling: Design and Analysis. Pacific Grove, CA: Duxbury Pr; 1999.
 
Laupacis A, Bourne R, Rorabeck C, Feeny D, Wong C, Tugwell P. et al.  The effect of elective total hip replacement on health-related quality of life. J Bone Joint Surg Am. 1993; 75:1619-26. PubMed
 
NIH Consensus Development Panel on Total Hip Replacement.  NIH consensus conference: Total hip replacement. JAMA. 1995; 273:1950-6. PubMed
 
Hawker G, Wright J, Coyte P, Paul J, Dittus R, Croxford R. et al.  Health-related quality of life after knee replacement. J Bone Joint Surg Am. 1998; 80:163-73. PubMed
 
Heck DA, Robinson RL, Partridge CM, Lubitz RM, Freund DA.  Patient outcomes after knee replacement. Clin Orthop Relat Res. 1998; 93-110. PubMed
 
van Essen GJ, Chipchase LS, O'Connor D, Krishnan J.  Primary total knee replacement: short-term outcomes in an Australian population. J Qual Clin Pract. 1998; 18:135-42. PubMed
 
Hamel MB, Toth M, Legedza A, Rosen MP.  Joint replacement surgery in elderly patients with severe osteoarthritis of the hip or knee: decision making, postoperative recovery, and clinical outcomes. Arch Intern Med. 2008; 168:1430-40. PubMed
 
Centers for Medicare & Medicaid Services, Office of the Actuary, National Health Statistics Group.  Personal health care spending by age group and source of payment, calendar year 2004. Accessed atwww.cms.hhs.gov/NationalHealthExpendData/downloads/2004-age-tables.pdfon 23 July 2009.
 
McWilliams JM, Zaslavsky AM, Meara E, Ayanian JZ.  Health insurance coverage and mortality among the near-elderly. Health Aff (Millwood). 2004; 23:223-33. PubMed
 
Baker DW, Sudano JJ, Durazo-Arvizu R, Feinglass J, Witt WP, Thompson J.  Health insurance coverage and the risk of decline in overall health and death among the near elderly, 1992-2002. Med Care. 2006; 44:277-82. PubMed
 
Card D, Dobkin C, Maestas N.  Does Medicare save lives? Quarterly Journal of Economics. 2009; 124:531-96.
 
Lubitz J, Cai L, Kramarow E, Lentzner H.  Health, life expectancy, and health care spending among the elderly. N Engl J Med. 2003; 349:1048-55. PubMed
 
Dor A, Sudano J, Baker DW.  The effect of private insurance on the health of older, working age adults: evidence from the health and retirement study. Health Serv Res. 2006; 41:759-87. PubMed
 
Hadley J, Waidmann T.  Health insurance and health at age 65: implications for medical care spending on new Medicare beneficiaries. Health Serv Res. 2006; 41:429-51. PubMed
 
Monheit AC, Vistnes JP.  Health insurance enrollment decisions: preferences for coverage, worker sorting, and insurance take-up. Inquiry. 2008; 45:153-67. PubMed
 
Levy H, DeLeire T.  What do people buy when they don't buy health insurance and what does that say about why they are uninsured? Inquiry. 2008; 45:365-79. PubMed
 
Cole SR, Hernán MA.  Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008; 168:656-64. PubMed
 
Bang H, Robins JM.  Doubly robust estimation in missing data and causal inference models. Biometrics. 2005; 61:962-73. PubMed
 
Cassel CM, Sarndal CE, Wretman JH.  Some uses of statistical models in connection with the nonresponse problem. Madow WG, Olkin I Incomplete Data in Sample Surveys III. Symposium on Incomplete Data, Proceedings. New York: Academic Pr; 1983.
 
Czajka JL, Hirabayashi SM, Little RJA, Rubin DB.  Projecting from advance data using propensity modeling: an application to income and tax statistics. Journal of Business and Economics Statistics. 1992; 10:117-131.
 
Kalton G, Piesse A.  Survey research methods in evaluation and case-control studies. Stat Med. 2007; 26:1675-87. PubMed
 
Lung Health Study Research Group.  Effect of inhaled triamcinolone on the decline in pulmonary function in chronic obstructive pulmonary disease. N Engl J Med. 2000; 343:1902-9. PubMed
 
Anthonisen NR, Connett JE, Kiley JP, Altose MD, Bailey WC, Buist AS. et al.  Effects of smoking intervention and the use of an inhaled anticholinergic bronchodilator on the rate of decline of FEV1. The Lung Health Study. JAMA. 1994; 272:1497-505. PubMed
 
Burge PS, Calverley PM, Jones PW, Spencer S, Anderson JA, Maslen TK.  Randomised, double blind, placebo controlled study of fluticasone propionate in patients with moderate to severe chronic obstructive pulmonary disease: the ISOLDE trial. BMJ. 2000; 320:1297-303. PubMed
 
Calverley PM, Anderson JA, Celli B, Ferguson GT, Jenkins C, Jones PW, et al. TORCH Investigators.  Salmeterol and fluticasone propionate and survival in chronic obstructive pulmonary disease. N Engl J Med. 2007; 356:775-89. PubMed
 
Pauwels RA, Löfdahl CG, Laitinen LA, Schouten JP, Postma DS, Pride NB. et al.  Long-term treatment with inhaled budesonide in persons with mild chronic obstructive pulmonary disease who continue smoking. European Respiratory Society Study on Chronic Obstructive Pulmonary Disease. N Engl J Med. 1999; 340:1948-53. PubMed
 
Reilly JJ.  COPD and declining FEV1—time to divide and conquer? [Editorial]. N Engl J Med. 2008; 359:1616-8. PubMed
 
Tashkin DP, Celli B, Senn S, Burkhart D, Kesten S, Menjoge S, et al. UPLIFT Study Investigators.  A 4-year trial of tiotropium in chronic obstructive pulmonary disease. N Engl J Med. 2008; 359:1543-54. PubMed
 
DeNavas-Walt C, Proctor BD, Smith J.  Income, poverty, and health insurance coverage in the United States: 2007. Washington, DC: U.S. Census Bureau; 2008. Report no. P60-235.
 
Meara E, White C, Cutler DM.  Trends in medical spending by age, 1963-2000. Health Aff (Millwood). 2004; 23:176-83. PubMed
 

Letters

NOTE:
Citing articles are presented as examples only. In non-demo SCM6 implementation, integration with CrossRef’s "Cited By" API will populate this tab (http://www.crossref.org/citedby.html).

Comments

Submit a Comment
Would Insuring the Near-Elderly Reduce Medicare Spending Beginning at Age 65?
Posted on December 8, 2009
Daniel Polsky
University of Pennsylvania and National Center for Health Statistics
Conflict of Interest: None Declared

McWilliams and colleagues (1) find that Medicare spending is higher for the previously uninsured and use this finding to suggest that nearly half of the cost of expanding health insurance coverage to uninsured near- elderly adults might be offset by reduced spending once the age of 65 has been reached. We do not believe this research supports the notion of cost savings from insuring the uninsured, because a large fraction of higher Medicare expenditures observed for the previously uninsured cannot be avoided by insuring people earlier. As pointed out by Bhattacharya (2), this may be particularly true if poor health caused them to be uninsured rather than poor health resulting from lack of insurance.

McWilliams and colleagues are unable to adequately control for this reverse causation. This problem is exacerbated by including in their sample individuals from the Health and Retirement Study who acquired public insurance after 1992 but before turning age 65. (Of these, individuals who are ever uninsured before turning 65 are included in the "uninsured" sample.) Adults who transition to public insurance before age 65 are likely to have had a health event that led them to become eligible for public insurance. This health event may have also caused them to be uninsured prior to obtaining public coverage. For example, individuals who quality for Medicare prior to age 65 due to participation in Social Security Disability Insurance (SSDI) qualify only after a 24-month waiting period following SSDI entitlement. Since they must be too disabled to work in order to qualify for SSDI, a substantial fraction are uninsured during the waiting period. (3) According to our estimates using McWilliams' sample definitions, 25% of the "uninsured" but only 11% of the "insured" transitioned to public insurance before turning 65. Because the disabled are more likely to use health care services and because their health characteristics are not adequately captured in the control variables, the McWilliams estimate includes spending that could not be avoided by insuring the uninsured. Our calculations suggest that the McWilliams estimate of the difference in Medicare spending between the uninsured and insured would drop by 50% if those ever public insured before age 65 are dropped from the sample.

Unrealistic expectations about the cost of health reform could lead to early dissatisfaction with reform efforts and result in their curtailment. The costs of insuring the near-elderly uninsured are unlikely to be offset by significant reductions in Medicare spending after age 65.

References

1. McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ. Medicare spending for previously uninsured adults. Ann Intern Med. 2009;151.

2. Bhattacharya J. Insuring the Near-Elderly: How Much Would Medicare Save? Ann Intern Med. 2009;151.

3. Riley GF. Health Insurance and Access to Care among Social Security Disability Insurance Beneficiaries during the Medicare Waiting Period. Inquiry. 2006. 43; 222-230.

Note: The views expressed in this letter reflect those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

Conflict of Interest:

None declared

Submit a Comment

Summary for Patients

Clinical Slide Sets

Terms of Use

The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.

Toolkit

Want to Subscribe?

Learn more about subscription options

Advertisement
Related Articles
Related Point of Care
Topic Collections
PubMed Articles

Want to Subscribe?

Learn more about subscription options

Forgot your password?
Enter your username and email address. We'll send you a reminder to the email address on record.
(Required)
(Required)