Four years later, David Himmelstein, Steffie Woolhandler, Elizabeth Warren, and Deborah Thorne are back. Last week they published a new analysis in The American Journal of Medicine. Now they say that 62.1 percent of bankruptcies in 2007 were “medical” in origin. But what they have really done is engaged in goal post shifting and misdirection.
Himmelstein et al. have extended the study from five states to the whole US but use a sample that was self-selecting. They sent questionnaires to 4976 debtors and received 2314 fully filled out (46.5 percent). A full 49.3 percent didn’t answer at all. Telephone interviews and analysis of court records then followed on some of these debtors.
Although the authors tried to check whether the respondents were representative, they could not control for the most important factor, medical debt. As the content of the questionnaire almost certainly betrayed their specific interest in health costs, debtors who were willing to participate may have been more likely to have such bills and/or believe they were a factor in the bankruptcy.
To expand their 2005 study, Himmelstein et al. applied the same criteria. This includes not only the 29 percent who cited health care costs (itself a self-report) but also those “reporting uncovered medical bills > $1000 in the past 2 years” (emphasis mine), losing two weeks of income from their job due to being sick or hurt (or had to leave their job, a criteria change from 2005), or mortgaging their house to cover medical debt. They consider this definition to be “conservative” (excuse me while I go laugh).
Further, “[d]ebtors who gave no answers regarding reasons for their bankruptcy were excluded from analyses,” i.e. the authors took out of the sample a large number of people who didn’t have medical debts or didn’t consider them a cause.
Himmelstein et al. do use a different threshold for uncovered costs of $5000 or 10 percent of family income for some analyses but do not distinguish clearly when each limit is being used. Further, estimates of the average health care costs of Americans in 2006 and 2007 put them at 6 to 10 percent of family income (based on a family plan and an income of $58,526). No surprise then that 34.7 percent of the debtors in the study met this threshold, especially since their sample has an average monthly income of $2676 or about $32,000 annually (median $2299/~$27,600), low for being considered middle class.
As for the two weeks of lost income criterion, well, it has always bothered me and not just because it over-represents certain types of jobs and doesn’t look at the debtor’s overall finances. Rather I wonder why this is considered a problem of the health care system. The authors haven’t put forth an argument that better care would have avoided or mitigated time out of work but instead seem to want a greater subsidization of those unable to work for whatever reason, an impression heightened by the fact that they then throw in income lost to care for someone who was sick as a criterion for medical bankruptcy. Whatever the value of such a policy, it is a question of broader social policies, not of the health system itself.
Himmelstein et al. go on to gloss over the many factors that distinguish so-called medical bankruptcies from non-medical ones, and which have obvious ramifications for increasing health costs. The debtors are slightly older (44.9 vs. 43.3 years), are more likely to be married (46.3 vs. 40.1 percent), and had a slightly larger average family size (2.79 vs. 2.63 people). All were statistically significant (p-values were .01, .02, .02). Also, they were more likely have had a break in insurance coverage in the last two years (40 vs. 34.1 percent, p=.005).
The authors claimed the debtors were middle-class based on “occupational prestige,” not income, and those with medical bankruptcies actually had lower scores in this area (86.1 percent with scores over 20 vs. 89.8 percent, p=.01). Further, they had lower incomes (mean/median monthly income = $2586/$2225 vs. $2851/$2478, p=.002) and were less likely to have a spouse who was employed (75.5 vs. 85 percent, p=.001). Their homes were also worth less ($141,861 vs. $159,145, p=.03).
Finally, Himmelstein et al. have again failed to take even a cursory look at the other bills and financial stresses or the level of overall debt.
None of this is to say that medical bills don’t play a role in some bankruptcies, perhaps even providing the final straw. Himmelstein et al. are coy, they never actually state that these bankruptcies are caused by medical bills but the authors are happy to opine to the press, knowing that the public will hear the scare story, not the reality. Oh, and guess what? There were fewer bankruptcies in 2007 than in 2001 – and fewer medical bankruptcies, however you define them. But shhhhh…it’s a secret.
