Medicine and Public Issues27 June 2017

The Relationship of Health Insurance and Mortality: Is Lack of Insurance Deadly?

FREE
    Author, Article, and Disclosure Information

    Abstract

    About 28 million Americans are currently uninsured, and millions more could lose coverage under policy reforms proposed in Congress. At the same time, a growing number of policy leaders have called for going beyond the Patient Protection and Affordable Care Act to a single-payer national health insurance system that would cover every American. These policy debates lend particular salience to studies evaluating the health effects of insurance coverage. In 2002, an Institute of Medicine review concluded that lack of insurance increases mortality, but several relevant studies have appeared since that time. This article summarizes current evidence concerning the relationship of insurance and mortality. The evidence strengthens confidence in the Institute of Medicine's conclusion that health insurance saves lives: The odds of dying among the insured relative to the uninsured is 0.71 to 0.97.

    Key Summary Points

    In several specific conditions, the uninsured have worse survival, and the lack of coverage is associated with lower use of recommended preventive services.

    The Oregon Health Insurance Experiment, the only available randomized controlled trial that has assessed the health effects of insurance, suggests that it may cause a clinically important decrease in mortality, but wide CIs preclude firm conclusions.

    The 2 National Health and Nutrition Examination Study analyses that include physicians' assessments of baseline health show substantial mortality improvements associated with coverage. A cohort study that used only self-reported baseline health measures for risk adjustment found a nonsignificant coverage effect.

    Most, but not all, analyses of data from the longitudinal Health and Retirement Study have found that coverage in the near-elderly slowed health decline and decreased mortality.

    Two difference-in-difference studies in the United States and 1 in Canada compared mortality trends in matched locations with and without coverage expansions. All 3 found large reductions in mortality associated with increased coverage.

    A mounting body of evidence indicates that lack of health insurance decreases survival, and it seems unlikely that definitive randomized controlled trials can be done. Hence, policy debate must rely on the best evidence from observational and quasi-experimental studies.

    At present, about 28 million Americans are uninsured. Repeal of the Patient Protection and Affordable Care Act (ACA) would probably increase this number, whereas enactment of proposed single-payer legislation (1) would reduce it. The public spotlight on how policy changes affect the number of uninsured reflects a widespread assumption that insurance improves health.

    A landmark 2002 Institute of Medicine (IOM) report on the effects of insurance coverage on the health status of nonelderly adults buttressed this assumption (2). The IOM committee responsible for the report found consistent evidence from 130 (mostly observational) studies that “the uninsured have poorer health and shortened lives” and that gaining coverage would decrease their all-cause mortality (2).

    The IOM committee also reviewed evidence on the effects of health insurance in specific circumstances and medical conditions. It concluded that uninsured patients, even when acutely ill or seriously injured, cannot always obtain needed care and that coverage improves the uptake of essential preventive services and chronic disease management. The report found that uninsured patients with cancer presented with more advanced disease and experienced worse outcomes, including mortality; that uninsured patients with diabetes, cardiovascular disease, end-stage renal disease, HIV infection, and mental illness (the 5 other conditions reviewed in depth) had worse outcomes than did insured patients; and that uninsured inpatients received less and worse-quality care and had higher mortality both during their hospital stays and after discharge.

    At the time of the IOM report, only 1 adequately controlled observational study had examined the effect of coverage on all-cause mortality. In this review, we summarize key evidence on this issue (Table 1), focusing on studies that have appeared since the IOM report and other previous reviews (3–6). Although not reviewed in detail here, more recent studies generally support the earlier reviews' conclusions that insurance coverage reduces mortality in several specific conditions (such as trauma [7] and breast cancer [8]), augments the use of recommended care (9), and improves several measures of health status (10, 11).

    Table 1. Summary of Studies on Relationship Between Insurance Coverage and All-Cause Mortality*

    Methods

    We searched PubMed and Google Scholar on 19 May 2017 for English-language articles by using the following terms: “[(uninsured) or (health insurance) or (uninsurance) or (insurance)] and [(mortality) or (life expectancy) or (death rates)].” After identifying relevant articles, we searched their bibliographies and used Google Scholar's “cited by” feature to identify additional relevant articles. We limited our scope to articles reporting data on the United States, quasi-experimental studies of insurance expansions in other wealthy nations, and recent cross-national studies. We contacted the authors of 4 studies to clarify their published reports on mortality outcomes.

    We excluded most observational studies that compared uninsured persons with those insured by Medicaid, Medicare, or the Department of Veterans Affairs because preexisting disability or illness can make an individual eligible for these programs. Hence, relative to those who are uninsured, publicly insured Americans have, on average, worse baseline health, thereby confounding comparisons. Conversely, comparing uninsured persons with those with private insurance (which is often obtained through employment) may be confounded by a “healthy worker” effect: that is, that persons may lose coverage because they are ill and cannot maintain employment. Nonetheless, most analysts of the relationship between uninsurance and mortality have viewed privately insured persons as the best available comparator, with statistical controls for employment, income, health status, and other potential confounders.

    Finally, we focus primarily on nonelderly adults because most studies have been limited to this group, and this group is likely to experience large gains or losses of coverage from health reforms. Since the advent of Medicare in 1966, almost all elderly Americans have been covered, precluding studies of uninsured seniors. Although Medicare's implementation may not have accelerated the secular decline in seniors' mortality (12), the relevance of this experience, which predates many modern-day therapies, is unclear.

    Children have also been excluded from most recent analyses of the relationship of insurance to mortality. Death in this population beyond the neonatal period is so rare that studies would need to evaluate a huge number of uninsured children to reach firm conclusions, and high coverage rates make assembling such a cohort difficult. The few studies addressing the effect of insurance on child survival have found that coverage reduces mortality (13–15), and few policy leaders contest the importance of covering children.

    Randomized Controlled Trials

    Only 1 well-conducted randomized controlled trial (RCT)—the Oregon Health Insurance Experiment (OHIE)—has assessed the effect of uninsurance on health outcomes (10, 16). In 2008, the state of Oregon opened a limited number of Medicaid slots to poor, able-bodied, uninsured adults aged 19 to 64 years. The state held a lottery among persons on a Medicaid waiting list, with winners allowed to apply for a slot. The OHIE researchers took advantage of this natural experiment to assess the effect of winning the lottery on the 74 922 lottery participants.

    Many lottery winners did not enroll in Medicaid, and 14.1% of lottery losers obtained Medicaid through other routes (some also got private coverage). Hence, the difference in the “dose” of Medicaid coverage was modest: an absolute difference of about 25%. To adjust for this, the OHIE researchers multiplied outcome differences by about 4 (10).

    At 1 year of follow-up, the death rate among lottery losers was 0.8%, and the death rate among the winners was 0.032% lower, a “dose-adjusted” difference of 0.13 percentage points annually (17). This difference was not statistically significant, an unsurprising finding given the OHIE's low power to detect mortality effects because of the cohort's low mortality rate, the low dose of insurance, and the short follow-up.

    The findings on other health measures, obtained from in-person interviews and brief examinations on a subsample of 12 229 individuals in the Portland area, help inform the mortality results. Most physical health measures were similar among lottery winners and losers in the subsample. However, winners had better self-rated health, were more likely to have diabetes diagnosed and treated with medication, and were much less likely to screen positive for depression (10). Medicaid coverage was associated with a nonsignificant decrease of 0.52 (95% CI, 2.97 to −1.93) mm Hg in systolic blood pressure and 0.81 (CI, 2.65 to −1.04) mm Hg in diastolic blood pressure (10). In addition to the low dose of insurance, these wide CIs reflect the lack of baseline blood pressure data; this precludes analyses that take advantage of paired measures on each individual, which would reduce the variance of estimates.

    In sum, the OHIE yields a (nonsignificant) point estimate that Medicaid coverage reduced mortality by 0.13 percentage points, equivalent to a (nonsignificant) odds ratio of 0.84.

    Two older RCTs are also relevant to the effect of insurance and access to care on mortality, although neither directly compared insured and uninsured persons. In the RAND Health Insurance Experiment, random assignment to full (first-dollar) coverage reduced diastolic blood pressure by an average of 0.8 mm Hg (P < 0.05) relative to persons randomly assigned to plans that required cost sharing (18), an effect size similar to the blood pressure findings in the OHIE. Unlike the OHIE, the RAND Health Insurance Experiment obtained baseline blood pressure readings, allowing researchers to determine that for participants with hypertension at baseline, full coverage reduced diastolic blood pressure by 1.9 mm Hg, mostly because of better hypertension detection (19); the effect was larger among low-income (3.5 mm Hg) than high-income (1.1 mm Hg) participants (19).

    The Hypertension Detection and Follow-up Program also suggests that removing financial barriers to primary care in populations with high rates of uninsurance may reduce mortality. That population-based RCT carried out in the 1970s screened almost all residents of 14 communities, with oversampling of predominantly black and poor locations. Persons with hypertension were randomly assigned to free stepped care in special clinics or referral to usual care. Although the clinics' staff treated only hypertension-related problems, they provided informal advice and “friendly referrals” for other medical issues (20). Strikingly, all-cause mortality was reduced by 17% in the intervention group, with similar reductions in deaths due to cardiovascular and noncardiovascular conditions (21).

    Finally, a flawed RCT carried out by the Social Security Administration starting in 2006 bears brief mention. That study randomly assigned people who were receiving Social Security disability income and were in the waiting period for Medicare coverage to receive immediate or delayed coverage (22). Unfortunately, randomization apparently failed, with many more patients with cancer assigned to the immediate coverage than to the control group, precluding reliable interpretation of the mortality results (11). Interestingly, persons receiving immediate coverage had rapid and significant improvements in most measures of self-reported health (11).

    Mortality Follow-up of Population-Based Health Surveys

    Several routinely collected federal surveys that include information about health insurance coverage have been linked to the National Death Index, allowing researchers to compare the mortality rates over several years of respondents with and without coverage at the time of the initial survey. One weakness of these studies is their lack of information about the subsequent acquisition or loss of coverage, which many people cycle into and out of over time. This dilutes coverage differences and may lead to underestimation of the effects of insurance coverage.

    Sorlie and colleagues (23) analyzed mortality among respondents to the 1982–1985 Current Population Survey, with follow-up through 1987. In analyses limited to employed persons, the relative risk for death associated with being uninsured was 1.3 for white men and 1.2 for white women (neither overall figures nor those for minorities were reported) (23). The study's lack of data on important determinants of health, such as smoking, and its reliance on employment status as the only proxy for baseline health status weaken confidence in its conclusions.

    Kronick used data from the 1986–2000 National Health Interview Surveys, with mortality follow-up through 2002 (24). The mortality hazard ratio for uninsured versus insured individuals was 1.10 (CI, 1.03 to 1.19) after adjustment for demographic variables, smoking, and body mass index. The hazard ratio fell to 1.03 (CI, 0.95 to 1.12) after additional adjustment for baseline health, defined by using self-reported disability and self-rated health. Although the self-rated health scale is known to be a valid predictor of mortality (25), it may introduce inaccuracies in comparisons of uninsured versus insured persons. Recent data (10, 11, 16, 26) indicate that gaining coverage improves self-rated health, before improvements in objective measures of physical health are detectable (or plausible). This suggests that uninsurance may cause people to underrate their health, perhaps because of anxiety or the inability to gain reassurance about minor symptoms. Analyses, such as Kronick's, that rely on self-rated health for risk adjustment therefore may inadvertently compare relatively sick insured persons to relatively healthy uninsured persons, obscuring outcome differences caused by coverage. Studies that include more objective measures of baseline health should be less subject to any such bias.

    Mortality Follow-up of Population-Based Health Examination Surveys

    Two studies have analyzed the effect of uninsurance on mortality using data from the National Health and Nutrition Examination Survey (NHANES), which obtains data from physical examination and laboratory tests among participants.

    Franks and colleagues (27) analyzed the 1971–1975 NHANES, with mortality follow-up through 1987. They compared mortality of uninsured and privately insured adults older than age 25 years, adjusted for demographic characteristics, self-rated health, smoking, obesity, leisure time exercise, and alcohol consumption. In addition, their models controlled for evidence of morbidity determined by laboratory testing and medical examinations performed by NHANES staff. By 1987, 9.6% of insured persons and 18.4% of uninsured persons had died. After adjustment for baseline characteristics and health status, the hazard ratio for uninsurance was 1.25 (CI, 1.00 to 1.55).

    Wilper and colleagues' study (which we coauthored) used data from the 1988–1994 NHANES, with mortality follow-up through 2000 (28). The study assessed mortality among uninsured and privately insured persons aged 17 to 64 years, controlling for demographic characteristics, smoking, alcohol consumption, body mass index, leisure time activity, self-rated health, and physician-rated health after the NHANES physician completed the medical examination. The study also included sensitivity analyses adjusting for the number of hospitalizations and physician visits within the past year, limitations in work or activities, job or housework changes due to health problems, and number of self-reported chronic diseases, which yielded results similar to those of the main model. In the main model, being uninsured was associated with a mortality hazard ratio of 1.40 (CI, 1.06 to 1.84).

    Quasi-experimental Studies of State and Provincial Coverage Expansions

    In 2 similar studies (29, 30), Sommers and colleagues compared mortality trends in states that expanded coverage to low-income residents (before implementation of the ACA) with trends in similar states without coverage expansions.

    Their analysis of Medicaid expansions in Maine, New York, and Arizona during the early 2000s found that adult mortality rates fell faster in those states than in neighboring ones (a relative reduction of 6.1%, or 19.6 deaths per 100 000), coincident with a decline in the uninsurance rate of 3.2 percentage points (29). Mortality reductions were largest among nonwhites, adults aged 35 to 64 years, and poorer counties. Sommers and colleagues' subsequent reanalysis using data that allowed better matching to control counties yielded a slightly lower estimate of the mortality effect (30). As the authors note, the large mortality effect from a relatively modest coverage expansion may reflect the fact that Medicaid enrollment often occurred “at the point of care for patients with acute illnesses,” leading to the selective enrollment of those most likely to benefit from coverage.

    A study of the effect of Massachusetts' 2006 coverage expansion compared mortality trends in Massachusetts counties with those in propensity score–matched counties in other states. Mortality decreased by 2.9% in Massachusetts relative to the comparison counties, a difference of 8.2 deaths per 100 000 adults, with larger declines in poorer counties and those with lower coverage rates before the expansion (31).

    Other Quasi-experimental Studies

    Several researchers have used data from the Health and Retirement Study (HRS)—a longitudinal study that has followed cohorts enrolled at age 51 years or older—to assess the effect of insurance coverage on mortality. The HRS periodically surveys respondents and their families and has been linked to Medicare and National Death Index data.

    McWilliams and colleagues found significantly higher mortality rates among uninsured compared with insured HRS respondents, even after propensity score adjustment for multiple predictors of insurance coverage (32). Baker and colleagues found that respondents who were uninsured (compared with those who had private insurance) had higher long-term but not short-term mortality (33). After adjustment for multiple baseline characteristics, including instrumental variables associated with coverage (such as a spouse's union membership), Hadley and Waidmann found a strong positive association between insurance coverage and survival before age 65 years (34). Black and colleagues suggested, on the basis of a “battery of causal inference methods,” that others overestimated the survival benefits of insurance and that uninsured HRS respondents had only slightly higher (adjusted) mortality than those with private coverage (35). Finally, studies have reached conflicting conclusions regarding whether the health of previously uninsured persons improves (relative to those who were previously insured) after they reach age 65 years and become eligible for Medicare (26, 36). Overall, the preponderance of evidence from the HRS suggests that being uninsured is associated with some increase in mortality.

    Some studies using other data sources suggest that death rates drop at age 65 years, coincident with the acquisition of Medicare eligibility (37, 38), whereas others do not (39).

    Finally, several studies have assessed the relationship between insurance coverage and hypertension control, a likely mediator of any relationship between coverage and all-cause mortality. Lurie and colleagues (40) followed a cohort of 186 patients who lost Medicaid coverage because of a statewide policy change and a control group of 109 patients who remained eligible. Among those who lost coverage, 5 died within 6 months (compared with none in the control group; P = .16), and the average diastolic blood pressure of those with hypertension increased by 10 mm Hg (compared with a 5–mm Hg decrease in controls; P = 0.003) (40). At 1 year, 7 patients who had lost Medicaid and 1 control patient had died; blood pressure differences were slightly less marked than those seen at 6 months (41). A similar study of patients terminated from Veterans Affairs outpatient care because of a budget shortfall found marked deterioration in hypertension control among the terminated patients relative to controls who maintained access (42). These clinic-based findings accord with cross-sectional, population-based analyses of data from NHANES, which have found worse blood pressure control among uninsured than insured patients with hypertension (43–45).

    Evidence From Other Nations and From Cross-National Studies

    The United States has lower life expectancy than most other wealthy nations and is the only one with substantial numbers of uninsured residents (46). Although many factors confound cross-national comparisons, a recent study suggests that worse access to good-quality health care contributes to our nation's higher mortality from medically preventable causes (so-called amenable mortality) (47). Similarly, a recent review of studies from many nations concluded that “broader health coverage generally leads to better access to necessary care and improved population health” (48).

    Quasi-experimental studies assessing newly implemented universal coverage in wealthy nations have reached similar conclusions. For instance, Taiwan's rollout of a single-payer system in 1995 was associated with an accelerated decline in amenable mortality, particularly in townships where coverage gains were larger (49, 50). In Canada, a study exploiting the different dates on which provinces implemented universal coverage estimated that coverage expansion reduced infant mortality by about 5% (P < 0.03) (51).

    Finally, a recent study of cystic fibrosis cohorts also suggests that coverage reduces mortality. Such patients live, on average, 10 years longer in Canada than in the United States. Among U.S. patients, those without known coverage have the shortest survival; among privately insured persons, life expectancy is similar to that of patients in Canada (52).

    Discussion

    The evidence accumulated since the publication of the IOM's report in 2002 supports and strengthens its conclusion that health insurance reduces mortality. Several newer observational and quasi-experimental studies have found that uninsurance shortens survival, and a few with null results used confounded or questionable adjustments for baseline health. The results of the only recent RCT, although far from definitive, are consistent with the positive findings from cohort and quasi-experimental analyses.

    Several factors complicate efforts to determine whether uninsurance increases mortality (Table 2). Randomly assigning people to uninsurance is usually unethical, and quasi-experimental analyses rest on unverifiable assumptions. Death is rare and mortality effects may be delayed, mandating large studies with long follow-up. Many people cycle into and out of coverage, diluting the effects of insurance. Statistical adjustments for baseline health usually rely on participants' self-reports, which may be influenced by coverage; hence, such adjustments may under- or overadjust for differences between insured and uninsured persons.

    Table 2. Why the Causal Relationship of Health Insurance to Mortality Is Hard to Study

    Inferences about mechanisms through which insurance affects mortality are subject to even greater uncertainty. In some circumstances, coverage might raise mortality by increasing access to dangerous drugs (such as oral opioids) or procedures (such as morcellation hysterectomy). On the other hand, coverage clearly reduces mortality in several serious conditions, although few are common enough to have a detectable effect on population-level mortality. The exception is hypertension, which is prevalent among uninsured persons and seems to be a likely contributor to their higher death rates. Although uncontrolled hyperlipidemia is also more common among the uninsured (44), the OHIE—the only RCT performed in the statin era—found no effect of coverage on cholesterol levels.

    Finally, our focus on mortality should not obscure other well-established benefits of health insurance: improved self-rated health, financial protection, and reduced likelihood of depression. Insurance is the gateway to medical care, whose aim is not just to save lives but also to relieve suffering.

    Overall, the case for coverage is strong. Even skeptics who suggest that insurance doesn't improve outcomes seem to vote differently with their feet. As one prominent economist (53) recently asked, “How many of the people who write such things . . . choose to just not bother getting their healthcare?”

    References

    • 1. Expanded & Improved Medicare For All Act, H.R. 676, 115th Cong. (2017). Google Scholar
    • 2. Institute of Medicine; Committee on the Consequences of UninsuranceCare Without Coverage: Too Little, Too Late. Washington, DC: National Academies Pr; 2002. Google Scholar
    • 3. McWilliams JMHealth consequences of uninsurance among adults in the United States: recent evidence and implications. Milbank Q2009;87:443-94. [PMID: 19523125] doi:10.1111/j.1468-0009.2009.00564.x CrossrefMedlineGoogle Scholar
    • 4. Hadley JSicker and poorer—the consequences of being uninsured: a review of the research on the relationship between health insurance, medical care use, health, work, and income. Med Care Res Rev2003;60:3S-75S. [PMID: 12800687] CrossrefMedlineGoogle Scholar
    • 5. Levy HMeltzer DThe impact of health insurance on health. Annu Rev Public Health2008;29:399-409. [PMID: 18031224] CrossrefMedlineGoogle Scholar
    • 6. Freeman JDKadiyala SBell JFMartin DPThe causal effect of health insurance on utilization and outcomes in adults: a systematic review of US studies. Med Care2008;46:1023-32. [PMID: 18815523] doi:10.1097/MLR.0b013e318185c913 CrossrefMedlineGoogle Scholar
    • 7. Rosen HSaleh FLipsitz SRogers SOGawande AADownwardly mobile: the accidental cost of being uninsured. Arch Surg2009;144:1006-11. [PMID: 19917936] doi:10.1001/archsurg.2009.195 CrossrefMedlineGoogle Scholar
    • 8. Hsu CDWang XHabif DVMa CXJohnson KJBreast cancer stage variation and survival in association with insurance status and sociodemographic factors in US women 18 to 64 years old. Cancer2017. [PMID: 28440864] doi:10.1002/cncr.30722 CrossrefMedlineGoogle Scholar
    • 9. Fox JBShaw FEOffice of Health System CollaborationOffice of the Associate Director for Policy, CDCRelationship of income and health care coverage to receipt of recommended clinical preventive services by adults—United States, 2011-2012. MMWR Morb Mortal Wkly Rep2014;63:666-70. [PMID: 25102414] MedlineGoogle Scholar
    • 10. Baicker KTaubman SLAllen HLBernstein MGruber JHNewhouse JPet alOregon Health Study GroupThe Oregon experiment—effects of Medicaid on clinical outcomes. N Engl J Med2013;368:1713-22. [PMID: 23635051] doi:10.1056/NEJMsa1212321 CrossrefMedlineGoogle Scholar
    • 11. Weathers RRStegman MThe effect of expanding access to health insurance on the health and mortality of Social Security Disability Insurance beneficiaries. J Health Econ2012;31:863-75. [PMID: 23000873] doi:10.1016/j.jhealeco.2012.08.004 CrossrefMedlineGoogle Scholar
    • 12. Finkelstein AMcKnight RWhat did Medicare do? The initial impact of Medicare on mortality and out of pocket medical spending. J Public Econ2008;92:1644-68. CrossrefGoogle Scholar
    • 13. Morriss FHIncreased risk of death among uninsured neonates. Health Serv Res2013;48:1232-55. [PMID: 23402526] doi:10.1111/1475-6773.12042 CrossrefMedlineGoogle Scholar
    • 14. Currie JGruber JSaving babies: the efficacy and cost of recent expansions of Medicaid eligibility for pregnant women. J Polit Econ1996;104:1263-96. CrossrefGoogle Scholar
    • 15. Currie JGruber JHealth insurance eligibility, utilization of medical care, and child health. Q J Economics1996:431-66. CrossrefGoogle Scholar
    • 16. Baicker KFinkelstein AThe effects of Medicaid coverage—learning from the Oregon experiment. N Engl J Med2011;365:683-5. [PMID: 21774703] doi:10.1056/NEJMp1108222 CrossrefMedlineGoogle Scholar
    • 17. Finkelstein ATaubman SWright BBernstein MGruber JNewhouse JPet alOregon Health Study GroupThe Oregon health insurance experiment: evidence from the first year. Q J Econ2012;127:1057-1106. [PMID: 23293397] CrossrefMedlineGoogle Scholar
    • 18. Newhouse JPFree for All: Lessons from the Rand Health Insurance Experiment. Cambridge, MA: Harvard Univ Pr; 2003. Google Scholar
    • 19. Keeler EBBrook RHGoldberg GAKamberg CJNewhouse JPHow free care reduced hypertension in the health insurance experiment. JAMA1985;254:1926-31. [PMID: 4046121] CrossrefMedlineGoogle Scholar
    • 20. Kass EHSpecial clinics for hypertension—the role of the hypertension detection—and follow-up programme. Br J Clin Pharmacol1982;13:81-6. [PMID: 7066158] CrossrefMedlineGoogle Scholar
    • 21. Five-year findings of the hypertension detection and follow-up program. I. Reduction in mortality of persons with high blood pressure, including mild hypertension. Hypertension Detection and Follow-up Program Cooperative Group. JAMA1979;242:2562-71. [PMID: 490882] MedlineGoogle Scholar
    • 22. Weathers RRSilanskis CStegman MJones JKalasunas SExpanding access to health care for Social Security Disability Insurance beneficiaries: early findings from the accelerated benefits demonstration. Soc Secur Bull2010;70:25-47. [PMID: 21261168] MedlineGoogle Scholar
    • 23. Sorlie PDJohnson NJBacklund EBradham DDMortality in the uninsured compared with that in persons with public and private health insurance. Arch Intern Med1994;154:2409-16. [PMID: 7979836] CrossrefMedlineGoogle Scholar
    • 24. Kronick RHealth insurance coverage and mortality revisited. Health Serv Res2009;44:1211-31. [PMID: 19453392] doi:10.1111/j.1475-6773.2009.00973.x CrossrefMedlineGoogle Scholar
    • 25. DeSalvo KBBloser NReynolds KHe JMuntner PMortality prediction with a single general self-rated health question. A meta-analysis. J Gen Intern Med2006;21:267-75. [PMID: 16336622] CrossrefMedlineGoogle Scholar
    • 26. Polsky DDoshi JAEscarce JManning WPaddock SMCen Let alThe health effects of Medicare for the near-elderly uninsured. Health Serv Res2009;44:926-45. [PMID: 19674430] doi:10.1111/j.1475-6773.2009.00964.x CrossrefMedlineGoogle Scholar
    • 27. Franks PClancy CMGold MRHealth insurance and mortality. Evidence from a national cohort. JAMA1993;270:737-41. [PMID: 8336376] CrossrefMedlineGoogle Scholar
    • 28. Wilper APWoolhandler SLasser KEMcCormick DBor DHHimmelstein DUHealth insurance and mortality in US adults. Am J Public Health2009;99:2289-95. [PMID: 19762659] doi:10.2105/AJPH.2008.157685 CrossrefMedlineGoogle Scholar
    • 29. Sommers BDBaicker KEpstein AMMortality and access to care among adults after state Medicaid expansions. N Engl J Med2012;367:1025-34. [PMID: 22830435] doi:10.1056/NEJMsa1202099 CrossrefMedlineGoogle Scholar
    • 30. Sommers BDState Medicaid expansions and mortality, revisited: a cost-benefit analysis. Am J Health Econ2017. doi:10.1162/AJHE_a_00080 CrossrefGoogle Scholar
    • 31. Sommers BDLong SKBaicker KChanges in mortality after Massachusetts health care reform: a quasi-experimental study. Ann Intern Med2014;160:585-93. [PMID: 24798521]. doi:10.7326/M13-2275 LinkGoogle Scholar
    • 32. McWilliams JMZaslavsky AMMeara EAyanian JZHealth insurance coverage and mortality among the near-elderly. Health Aff (Millwood)2004;23:223-33. [PMID: 15318584] CrossrefMedlineGoogle Scholar
    • 33. Baker DWSudano JJDurazo-Arvizu RFeinglass JWitt WPThompson JHealth insurance coverage and the risk of decline in overall health and death among the near elderly, 1992-2002. Med Care2006;44:277-82. [PMID: 16501400] CrossrefMedlineGoogle Scholar
    • 34. Hadley JWaidmann THealth insurance and health at age 65: implications for medical care spending on new Medicare beneficiaries. Health Serv Res2006;41:429-51. [PMID: 16584457] CrossrefMedlineGoogle Scholar
    • 35. Black BEspín-Sánchez JAFrench ELitvak KThe long-term effect of health insurance on near-elderly health and mortality. Am J Health Econ2017. doi:10.1162/AJHE_a_00076 CrossrefGoogle Scholar
    • 36. McWilliams JMMeara EZaslavsky AMAyanian JZHealth of previously uninsured adults after acquiring Medicare coverage. JAMA2007;298:2886-94. [PMID: 18159058] CrossrefMedlineGoogle Scholar
    • 37. Lichtenberg FRThe Effects of Medicare on Health Care Utilization and Outcomes. Frontiers in Health Policy Research. Vol. 5. Cambridge, MA: MIT Pr; 2002. Google Scholar
    • 38. Card DDobkin CMaestas NDoes Medicare save lives? Q J Econ2009;124:597-636. [PMID: 19920880] CrossrefMedlineGoogle Scholar
    • 39. Card DDobkin CMaestas NThe impact of nearly universal insurance coverage on health care utilization: evidence from Medicare. Am Econ Rev2008;98:2242-58. [PMID: 19079738] CrossrefMedlineGoogle Scholar
    • 40. Lurie NWard NBShapiro MFBrook RHTermination from Medi-Cal—does it affect health? N Engl J Med1984;311:480-4. [PMID: 6379458] CrossrefMedlineGoogle Scholar
    • 41. Lurie NWard NBShapiro MFGallego CVaghaiwalla RBrook RHTermination of Medi-Cal benefits. A follow-up study one year later. N Engl J Med1986;314:1266-8. [PMID: 3517642] CrossrefMedlineGoogle Scholar
    • 42. Fihn SDWicher JBWithdrawing routine outpatient medical services: effects on access and health. J Gen Intern Med1988;3:356-62. [PMID: 3404297] CrossrefMedlineGoogle Scholar
    • 43. Christopher ASMcCormick DWoolhandler SHimmelstein DUBor DHWilper APAccess to care and chronic disease outcomes among Medicaid-insured persons versus the uninsured. Am J Public Health2016;106:63-9. [PMID: 26562119] doi:10.2105/AJPH.2015.302925 CrossrefMedlineGoogle Scholar
    • 44. Wilper APWoolhandler SLasser KEMcCormick DBor DHHimmelstein DUHypertension, diabetes, and elevated cholesterol among insured and uninsured U.S. adults. Health Aff (Millwood)2009;28:w1151-9. [PMID: 19843553] doi:10.1377/hlthaff.28.6.w1151 CrossrefMedlineGoogle Scholar
    • 45. Egan BMLi JSmall JNietert PJSinopoli AThe growing gap in hypertension control between insured and uninsured adults: National Health and Nutrition Examination Survey 1988 to 2010. Hypertension2014;64:997-1004. [PMID: 25185135] doi:10.1161/HYPERTENSIONAHA.114.04276 CrossrefMedlineGoogle Scholar
    • 46. Organization for Economic Cooperation and Development. OECD health statistics. 2017. doi:10.1787/health-data-en Google Scholar
    • 47. GBD 2015 Healthcare Access and Quality CollaboratorsHealthcare Access and Quality Index based on mortality from causes amenable to personal health care in 195 countries and territories, 1990-2015: a novel analysis from the Global Burden of Disease Study 2015. Lancet2017. [PMID: 28528753] doi:10.1016/S0140-6736(17)30818-8 CrossrefMedlineGoogle Scholar
    • 48. Moreno-Serra RSmith PCDoes progress towards universal health coverage improve population health? Lancet2012;380:917-23. [PMID: 22959388] doi:10.1016/S0140-6736(12)61039-3 CrossrefMedlineGoogle Scholar
    • 49. Lee YCHuang YTTsai YWHuang SMKuo KNMcKee Met alThe impact of universal National Health Insurance on population health: the experience of Taiwan. BMC Health Serv Res2010;10:225. [PMID: 20682077] doi:10.1186/1472-6963-10-225 CrossrefMedlineGoogle Scholar
    • 50. Wen CPTsai SPChung WSA 10-year experience with universal health insurance in Taiwan: measuring changes in health and health disparity. Ann Intern Med2008;148:258-67. [PMID: 18283203] LinkGoogle Scholar
    • 51. Hanratty MJCanadian national health insurance and infant health. American Econ Rev1996;86:276-84. Google Scholar
    • 52. Stephenson ALSykes JStanojevic SQuon BSMarshall BCPetren Ket alSurvival comparison of patients with cystic fibrosis in Canada and the United States: a population-based cohort study. Ann Intern Med2017;166:537-46. [PMID: 28288488]. doi:10.7326/M16-0858 LinkGoogle Scholar
    • 53. Krugman P. America: equity and equality in health. YouTube video. Posted by Roosevelt House Public Policy Institute at Hunter College, 22 May 2017. Accessed at www.youtube.com/watch?v=TEti8sVAbcA&feature=youtu.be on 7 June 2017. Google Scholar

    Comments

    0 Comments
    Sign In to Submit A Comment
    Steffie Woolhandler, M.D., M.P.H David U. Himmelstein, M.D.7 November 2017
    Author's Response
    Some health care reforms cost more than expected, others less. Dr. Grey cites the early underestimation of Medicare's costs, but omits cases where costs were overestimated. For instance, the Congressional Budget Office initially projected that the Affordable Care Act's coverage expansion provisions would cost $187 billion in 2017 , its latest estimate is $66 billion lower . Similarly, Medicare's drug benefit has cost 35% less than predicted .

    Experience in nations with national health insurance (NHI) also indicates that universal, comprehensive coverage need not break the bank. All spend far less than we do, yet avoid the narrow networks and surprise bills that bedevil many patients. Almost all enjoy better health outcomes, and in the ten other countries included in recent surveys, even poor residents reported better access than the average American ; only Germany's primary care doctors were less satisfied than those in the U.S.

    Nonetheless, Grey is correct that single-payer reform would require tradeoffs. We cannot afford private insurers, who add nothing of value while charging overhead four-fold greater than Medicare's, or the complex payment systems that impose $200 billion in unnecessary paperwork on hospitals and doctors. Nor can we sustain drug firms' exorbitant prices and profits.

    Our current payment strategies also encourage providers to inflate their billings. Hospitals, HMOs and ACOs live or die based on their bottom line - their profit (or, for non-profits, "surplus"). Profitable institutions can expand and modernize, while unprofitable ones shrivel, even if they're providing excellent and much-needed care. The profit imperative - under both capitated and fee-for-service payment - drives providers to seek out lucrative patients and services, avoid unprofitable ones and portray all patients as sicker than they really are, boosting administrative and total costs..

    Payment strategies that decouple care from the prospect of profit have proven far less inflationary, and better at matching resources to community need. For instance, Canada and Scotland pay hospitals global operating budgets - like schools or fire departments - obviating the need for per-patient billing. There's little incentive to upcode or cherry-pick, since hospitals can't keep surplus operating funds; new investments are instead funded through separate government grants.

    Market-driven care is the root cause of America's health care dilemma. No law of nature decrees that costs must soar or patients must suffer; that MBAs should supervise MDs; or that the our nation can't match or exceed others' health care successes.

    Congressional Budget Office. Letter to Nancy Pelosi. March 20, 2010. Available at: https://www.cbo.gov/sites/default/files/111th-congress-2009-2010/costestimate/amendreconprop.pdf (accessed 10/29/2017).
    Congressional Budget Office. Federal subsidies under the Affordable Care Act for health insurance coverage related to the expansion of Medicaid and nongroup health insurance: Tables from CBO's January 2017 baseline. Available at: https://www.cbo.gov/sites/default/files/recurringdata/51298-2017-01-healthinsurance.pdf (accessed 10/29/2017).
    Elmendorf, D. The Accuracy of CBO’s Budget Projections. March 25, 2013. available at: https://www.cbo.gov/publication/44017 (accessed 10/29/2017)
    Osborn R, Squires D, Doty MM, Sarnak DO, Schneider DC. In new survey of eleven countries, US adults still struggle with access to and affordability of health care. Health Aff (Milwood) 2016; 35:2327 -2336,
    The Commonwealth Fund. 2015 International survey of primary care doctors TOPLINE. available at: http://www.commonwealthfund.org/~/media/files/surveys/2015/2015-ihp-survey_topline_11-20-15.pdf (accessed 10/29/2017)).
    Michael R. Grey, MD, MPH12 July 2017
    A Report from the Committee of Brutal Facts

    “If you look for truth, you may find comfort in the end; if you look for comfort you will not get either comfort or truth only soft soap and wishful thinking to begin, and in the end, despair.” C.S. Lewis


    Himmelstein and Woolhandler’s article (AIM June 27 2017) is as well-grounded in empirical evidence as we are likely to get and it clearly comes out in favor of what both authors have both long articulated with passion and reason: a single payer system.  Supporters of a single payer system have not been completely transparent about the simple truth: we cannot afford the system of health care that we would actually like to have unless we accept sacrifices that as yet we are unwilling to bear.  The purported savings by creating a national or state-based single payer system are almost certainly overstated and fraught with uncertainty.  As an example, in 1965 the government’s lead actuary predicted that Medicare Part A would reach $9 billion by 1990; yet the actual cost exceeded $66 billion.  To advocate that a single payer system can happen without substantial and long term tax-based subsidies and/or limiting choice and access to services that most Americans see as birthrights is disingenuous.    
    The deeply difficult task of paring down the growth of entitlement programs like Medicare, and de facto entitlement programs like Medicaid—the primary drivers of the nation’s fiscal woes-- require limiting access to some services, restricting choice, and the political will and managerial skills to ensure fiscal accountability.   These realities are rarely addressed by reformers and critics regardless of political persuasion and the facts on the ground are not encouraging: exhibit A, the Veteran’s Administration.  Until we accept that we cannot afford the health care that we want without starving other areas of the economy in need of investment—infrastructure, education—we will remain stuck in a fruitless argument that is fired more by passion and politics than hard truths.  
    Arguments that the root causes of our present dilemma comes down to self-serving doctors, hospitals, employers or private insurers—or to perfidious government for that matter—are simplistic.  In fact, it is hard to find credible and consistent villainy in this dilemma.   No one wants anyone to die for lack of medical care; accusations to the contrary are counterproductive. When our political and professional leaders start promoting honest discussions with the American people on health care, just maybe we will find the good will and self-sacrifice necessary to solve a dilemma nearly a century old.  The time is past for soft soap and wishful thinking in regards to health care. It is truth we need, not comfort.  

    Disclosures: I was an intern in the same program with Dr. Woolhandler was a junior resident and Dr. Himmelstein a faculty member.