More on Medicare Mortality
You can expect that I'll be blogging quite a bit about this topic over the next few days. A reader in Tyler Cowen's comments offers this 2009 study:
Health insurance characteristics shift at age 65 as most people become eligi- ble for Medicare. We measure the impacts of these changes on patients who are admitted to hospitals through emergency departments for conditions with similar admission rates on weekdays and weekends. The age proﬁles of admissions and comorbidities for these patients are smooth at age 65, suggesting that the severity of illness is similar on either side of the Medicare threshold. In contrast, the number of procedures performed in hospitals and total list charges exhibit small but statistically signiﬁcant discontinuities, implying that patients over 65 receive more services. We estimate a nearly 1-percentage-point drop in 7-day mortality for patients at age 65, equivalent to a 20% reduction in deaths for this severely ill patient group. The mortality gap persists for at least 9 months after admission
I referred to three earlier studies, including one by Card, as surveyed by Levy and Meltzer:
Card et al. use a regression discontinuity approach to estimate the impact of Medicare coverage on mortality. The basic idea is that if health insurance significantly affects mortality in the short run, the dramatic increase in health insurance coverage at age 65 as a result of Medicare should translate into a reduction in mortality at age 65 relative to the overall trend by age. In fact, the data show no such discontinuity: mortality changes smoothly with age. Card et al. do find discrete, significant increases in consumption of medical care. They also noted some improvements in self-reported health at age 65, although many of these effects are imprecisely estimated. One important exception is the result for Hispanics and low-income minorities, both of whome see significant increases in the probability of reporting good or better health at age 65.
Polsky et al. take a different approach to estimating the impact of Medicare on health. They analyze changes in the trajectory of self-reported health at age 65 adn find that receiving Medicare increases the probability that respondents report excellent or very good health. One surprising aspect of these findings is that these shifts are observed both for respondents who were uninsured prior to Medicare and for those who were otherwise insured. This result is surprising because one would have expected benefits to be concentrated in those who were not previously insured. The observed improvements are also surprising because Medicare often provides less comprehensive coverage than do most private insurance plans. Polsky et al. hypothesize that the effect may be due to the stability of Medicare coverage compared with private coverage.
Finkelstein & McKnight use data from the 1960s to see whether geographic areas with lower insurance coverage rates prior to the enactment of MEdicare experienced improvements in mortality following the enactment of Medicare relative to areas with higher pre-1965 coverage rates. Earlier work by Finkelstein using this same strategy documents significant increases on hospital spending and utilization, but the work of Finkelstein & McKnight finds no corresponding improvement in mortality. The authors conclude that in its first 10 years, the establishment of universal health insurance for the elderly had no discernable impact on their mortality." Of course, this result applies to Medicare circa 1970; advances in medical technology and in the scope of Medicare benefits since then may have greatly increased the marginal health benefits of Medicare coverage.
Taken together, these three studies of Medicare paint a surprisingly consistent picture: Medicare increases consumption of medical care and may modestly improve self reported health, but has no effect on mortality, at least in the short run. Whether there are long-term effects remains an open question; this uncertainty reflects the limited generalizeability of the natural experiment results.
The science is always evolving. Obviously, if we get a lot of results showing that there is a big effect at 65, I'll change my mind; but right now, the bulk of the evidence runs the other way. It's worth noting that the later Card paper itself notes that the aggregate figures show no mortality improvement:
As is true for health insurance more generally (see Levy and Meltzer ), it has proven more difﬁcult to identify the health impacts of Medicare.9 Most existing studies have focused on mor- tality as an indicator of health.10 An early study by Lichtenberg (2001) used Social Security Administration (SSA) life table data to test for a trend-break in the age proﬁle of mortality at age 65. Although Lichtenberg identiﬁed a break, subsequent analysis by Dow (2004) showed that this is an artifact of the interval- smoothing procedure used to construct the SSA life tables. Com- parisons based on unsmoothed data show no evidence of a shift at age 65 (Card, Dobkin, and Maestas 2004). Finkelstein and McK- night (2005) explore trends in state-speciﬁc mortality rates for people over 65 relative to those under 65, testing for a break around 1966--the year Medicare was introduced. They also ex- amine the correlation between changes in relative mortality after 1966 and the fraction of elderly people in a region who were unin- sured in 1963. Neither exercise suggests that the introduction of Medicare reduced the relative mortality of people over 65, though it should be noted that the power of these analyses is limited.
How could mortality improve at the micro level, and not at the macro level? Increasing utilization of health services is not all mortality improving. As I note in the column, health care can kill as well as heal--one estimates puts the death from nosocomial (treatment-induced) infections at 80,000 a year. So while it's entirely possible--indeed certain--that some number of people are saved by having insurance, it's also very likely that some number of people are saved by not having it, or having less generous insurance, because they don't go in for a treatment that would have killed them.
The 2009 paper was looking at a small subset of conditions that are urgent, and which we're relatively adept at treating. But it may be washed out by the people who die having knee surgery.
This is, of course, why comparative effectiveness research is very popular among wonks. But it's trickier than it sounds, because patients are very heterogenous. I actually expect this problem to go down in the next twenty years or so, as better genomics gives us more of a handle on which treatments work for whom.
One thing it does suggest is that if we want to maximize the benefits from expanding insurance coverage, we really need to wage a scorched-earth battle against nosocomial infection. Hospital hygeine has slipped massively from where it was in the thirties, because antibiotics have made health care workers lless urgent about it. We need to return to the OCD days of yore.