(3 p.m. update: See italicized items with responses from the lead author of the Harvard study, Dr. David Himmelstein.)
President Obama’s kicking off his health care reform today in the worst possible way: with a mischaracterization of data.
“The cost of health care now causes a bankruptcy in America every thirty seconds," Obama said at the opening of his White House forum on health care reform. The problem: That claim, based on a 2001 survey, is simply unsupportable.
The figure comes from a 2005 Harvard University study saying that 54 percent of bankruptcies in 2001 were caused by health expenses. We reviewed it internally and knocked it down at the time; an academic reviewer did the same in 2006. Recalculating Harvard’s own data, he came up with a far lower figure – 17 percent.
A more recent study by another group, approaching it another way, indicates that in 2007 about eight-tenths of one percent of Americans lived in families that filed for bankruptcy as a result of medical costs. That rings a little less loudly than “one every 30 seconds.”
The extrapolation of Harvard’s data to “a bankruptcy every 30 seconds,” which Obama also mentioned in his address to a joint session of Congress last month, comes, per the White House, from a 2005 Washington Post op-ed by Prof. Elizabeth Warren, a co-author of the Harvard paper. Fact-check.org has noted that even using Harvard’s numbers, it’s more like a bankruptcy every minute; indeed if you add up all bankrputcies in a year you barely get one every 30 seconds. (I've e-mailed Warren for comment.) But more to the point is that the Harvard data are clearly inflated, or at best, mischaracterized.
Himmelstein tells me that the reason for the difference is a change in federal law that sharply reduced the number of bankruptcies. In 2005, the year he and Warren wrote their op-ed, there were just over 2 million bankruptcies. Data out just today say that in 2008 there were 1.1 million (up sharply, by the way, over 2007). So this error in the White House claim stems simply from the fact that it's using out-of-date information. The next question is whether the estimate of “medical bankruptcies” is reliable in the first place.
A good part of the problem is definitional. The Harvard report claims to measure the extent to which medical costs are “the cause” of bankruptcies. In reality its survey asked if these costs were “a reason” – potentially one of many – for such bankruptcies.
Beyond those who gave medical costs as “a reason,” the Harvard researchers chose to add in any bankruptcy filers who had at least $1,000 in unreimbursed medical expenses in the previous two years. Given deductibles and copays, that’s a heck of a lot of people.
Moreover, Harvard’s definition of “medical” expenses includes situations that aren’t necessarily medical in common parlance, e.g., a gambling problem, or the death of a family member. If your main wage-earning spouse gets hit by a bus and dies, and you have to file, that’s included as a “medical bankruptcy.”
When I asked the lead author, Dr. David Himmelstein, about his definitions of medical bankruptcy back in 2005, he said, “It’s a judgment call,” and added that any death, for example, “to our mind is a medical event.”
A last problem was sampling: The Harvard researchers surveyed bankruptcy filers in five federal court districts accounting for 14 percent of bankruptcies nationally; projecting this to the other 86 percent is sketchy. Said Himmelstein: “Obviously the extrapolation is rough.”
Of such rough extrapolations are presidential pronouncements made.
Himmelstein today told me that he’s comfortable saying medical costs, as his study defines them, are “a cause” but not “the cause” of bankruptcies. In his view, “It’s accurate to say medical problems cause half of bankruptcies. There may be other conditions as well but medical problems were causal. I wouldn’t be comfortable with it as the ‘only’ cause.”
Worth keeping in mind is the fact that no one (apparently) disagrees about the pain medical expenses can cause to uninsured Americans. Prof. David Dranove of Northwestern University, who wrote the 2006 paper picking apart the Harvard study, noted that he has a new paper in the works showing that uninsured people who have a severe illness lose a substantial portion of their retirement assets.
"There is general agreement: Being uninsured and getting sick in the United States is really a bad thing,” Dranove told me today. “But for academics the validity of the research matters.” In the Harvard paper, he says, "The methods were so poor they gave cover to those who want to dismiss the problems of the uninsured – they can say the only paper out there uses a suspect method."
There’s been a fair amount of academic back-biting about this issue. On one hand Himmelstein, the lead Harvard researcher, is a co-founder of Physicians for a National Health Program, created to promote a government-run single-payer health system. On the other, Dranove took $5,000 from the nation’s health insurance industry for his report, which he says he now regrets for the criticism of his impartiality it’s engendered. Both papers were peer-reviewed.
“It stinks to be uninsured. I don’t want to be quoted saying anything else,” Dranove says. “But there are correct studies, and incorrect studies. For academics, the validity of the research methods matters.”
It should for the rest of us, too.
Himmelstein’s referred me to a 2006 paper in which he replied to Dranove, whom he accuses of “several out and out errors.” Says Himmelstein: “They were paid by the insurance industry to make this critique… They were hired guns out to try and make a point, and used a variety of illegitimate techniques to make that point.”
Science marches on.
(…and a 4 p.m. update: Elizabeth Warren, Himmelstein's co-author, is serving as chair of the Congressional Oversight Panel on TARP. Per spokeswoman Shanan Guinn, she's not currently giving interviews on her previous research.)