Truth is not only the first casualty of war, it is also the first casualty of serious public policy debate.
“An estimated 17,000 children in the United States might have died unnecessarily over nearly two decades because they didn’t have health insurance,” said U.S. News and World Report. “The authors estimated that at least 1,000 hospitalized children died each year simply because they lacked insurance,” said The New York Times.
They’re talking about a Johns Hopkins Children’s Center study [gated, but with abstract]. But between the media hype and the actual study is an enormous chasm that separates fact from fiction. In truth, the authors of the study did not establish that anybody, anywhere, died of any cause whatsoever because of a lack of health insurance.
This is only the latest in a series of ridiculous claims that have been injected into the health insurance debate. What follows is a brief review, some of which has appeared earlier at the Health Affairs blog.
It’s life’s illusions I recall
Last year, a report by Families USA made the astounding claim that 6 people die every day in Florida because they are uninsured. Seven die every day in Texas, 8 in California, and 25 in New York.
How was Families USA able to tally up all that carnage with such pinpoint precision? As one of us previously explained, these claims are based on a 15-year cascade of studies — each repeating the errors and misinterpreting or mischaracterizing the findings of the previous one and ultimately relying on data that is 37 years old.
It begins with a paper by Peter Franks et al. published in Journal of the American Medical Association in 1993, estimating that being uninsured increased the probability of death by 25%. Although the subjects were interviewed only once, for the study’s inference to be meaningful, one is forced to make the unverified assumption that the uninsured stayed uninsured for a full 19 years!
Continuing the saga, the Institute of Medicine (IOM) uncritically used the Frank’s result to claim that 18,000 deaths a year in the U.S. are attributable to a lack of health insurance. The Urban Institute updated the IOM report, and Families USA updated that report.
Not to be outdone, the Physicians for a National Health Program (PNHP) repeated the exercise (with all its methodological sins) and boosted the tally to a 40% increase in the probability of dying for the uninsured. That produces a whopping 45,000 premature deaths every year — almost as bad as the Vietnam War. And, yes, we get a state-by-state breakdown. There will be 5,302 deaths attributed to uninsurance in California this year. There will be 75 in Wyoming, etc., etc. There is even a minute-by-minute tally: “The Institute of Medicine, using older studies, estimated that one American dies every 30 minutes from lack of health insurance,” says David Himmelstein, one of the authors. “Now one dies every 12 minutes.”
As in the previous incarnations, the researchers interviewed the uninsured only once — and never saw them again. A decade later, the researchers assumed the participants were still uninsured and, if they died in the interim, lack of insurance is blamed as one of the causes.
Yet, like unemployment, uninsurance happens to many people for short periods of time. Most people who are uninsured regain insurance within one year. The authors of the study did not track what happened to the insurance status of the subjects over the decade examined, what medical care they received or even the causes of their deaths.
Also, before you go into mourning too quickly, be aware that when former Director of the Congressional Budget Office (CBO) June O’Neill and her husband Dave used a similar approach they found that the involuntarily uninsured (low-income people) were only 3% more likely to die over a 14-year period than those with health insurance. There was no statistically significant effect on the “voluntarily uninsured” (higher-income people).
That’s not too surprising in light of a RAND study finding. People are receiving appropriate care a little better than half the time when they see doctors. According to RAND, the care patients receive is not affected by whether they are insured or uninsured or by the type of insurance they have. People who are uninsured, of course, may delay seeing a doctor in the first place — because of their lack of insurance. But this problem is unlikely to be solved by enrolling them in Medicaid programs that routinely ration by waiting.
The uninsured in the PNHP study are disproportionally poor, minority, high school dropouts and unemployed. Although the study ignores public insurance, bills before Congress would try to insure this group by expanding Medicaid. The O’Neills found that people enrolled in Medicaid have a much higher mortality rate than the uninsured. In other words, Medicaid enrollment reduces life expectancy rather than increasing it!
The Johns Hopkins study at least has one advantage over all the other studies: The authors knew the insurance status of the children at the time they died. After combing hospital records over a 20-year period, they determined that although children admitted to hospitals rarely die, those who did were 60% more likely to have been uninsured. But does that mean that uninsurance is the reason? The authors offer no theory connecting uninsurance with health and no specific theory is tested.
Of children admitted to hospitals who did not die, the average length of stay (two days) is the same for the insured and the uninsured. The cost of care is about 90% the same. This is consistent with the RAND results. However, among children who died, the length of stay for insured children is longer and more varied, and the cost of care is 2½ times higher than for the uninsured. In fact, more than half of the uninsured children who died did not remain in the hospital for even a single day — which suggests that many died in the emergency room or were DOA. It seems likely, therefore, that among those who died, the insured and uninsured were very dissimilar.
The most common reasons for children being hospitalized were complications from birth, pneumonia, and asthma. The study found that the reasons did not differ by insurance status.
What did differ was the fraction of deaths in the older uninsured population, particularly those aged 11-18. This matters because there is an extensive literature chronicling the fact that child accident and homicide rates, which have the largest effects on teenagers, are inversely related to parental characteristics like income, and educational attainment, directly related to parental levels of substance abuse, and a function of marital status and living arrangements.
In the early 2000s, motor vehicle accidents, drowning, burns, and poisoning, cause more than a third of all deaths among children aged 1-19. Homicide was the 4th leading cause of death in 1-4 year olds, 5-9 year olds, and 10-14 year olds. It was the 2nd leading cause of death in 15-19 year olds. [link, link] Lack of insurance is unlikely to contribute to deaths by homicide, motor vehicle crash, drowning, burns, or poisoning.
Remarkably, this paper is silent on whether people living in conditions that make their children more prone to accidental death or homicide are also less likely to have health insurance.
Richard Kronick has a review of the literature on what we know about those who lack health insurance in the August issue of Health Services Research. He writes:
There would not be much change in the number of deaths in the United States as a result of universal coverage, although the difficulties in inferring causality from observational analyses temper the strength of this conclusion.
As for the public policy implications of all this, the PNHP authors are on record as favoring Canada’s system of single-payer national health insurance. Yet insured patients in Canada fare worse than the uninsured in the United States. The O’Neills find that:
- Among US women age 40 to 64, 87% of those with insurance had a mammogram within 5 years.
- The rate for Canadian women is 65% — the same as for uninsured women in the US.
- Canadian women also have the same rate of screening for cervical cancer as uninsured US women (80%), over five years. Among insured US women, the rate is 92%.
- Among uninsured US men, 31% were screened for prostate cancer, compared with 16% of Canadian men. For insured US men, the rate is 52%.