Does the VA Cost Less Than Private Health Care?
Mark Hall | Leave a Comment
Taking a break from law, this post is about whether the Veterans Health Administration provides care more efficiently than the private sector. Paul Krugman and others have held the VA out as a shining example of the government’s ability to provide high quality care efficiently, as well as the private sector’s need to lower costs and improve quality through electronic medical records, comparative effectiveness research, reduced overhead, salaried physicians, and integrated delivery systems – all issues that are central to current health reform debates. It would be a huge blow if it turns out none of this is true – but that’s precisely what’s suggested by an article published by VA researchers earlier this year.
Wm. Weeks, MD (at Dartmouth and the VA’s regional center) reported that, from 2001-2006, the VA cost 33% more than equivalent care in the private sector, and its quality was not notably better. Here, I focus on the cost findings, since they diverge dramatically from the prior, state-of-the-art, study by Nugent, Hendricks et al (also from the VA), which found that, in 1999, the VA cost about 20% less than Medicare. Since Medicare itself costs 25-30% less than the private sector, Dr. Weeks reports the VA costs about twice what Dr. Nugent and other VA colleagues previously reported. What makes this discrepancy even more remarkable is that Weeks did not even cite this prior work of VA colleagues, published in multiple articles in leading journals.
What gives? I’m not expert, but its clear their methods differed sharply. Nugent et al. took all care delivered at 6 VA centers and valued the services at actual Medicare fee-for-service rates, comparing the total with costs borne by the 6 VA centers. Thus, the measures and comparison are direct, apples-to-apples. Weeks, on the other hand, compared total VA medical costs (excluding nursing homes) per user with per person costs reported by VA users in the Medical Expenditure Panel Survey (MEPS), which values those services at private sector rates. MEPS is a national survey that contains only a small subsample of 500 VA users, about 1 of every 50,000 VA user. Extrapolating from such a small sample is a much more indirect comparison, so merits closer scrutiny, which reveals many potential flaws:
- The 500 VA users in MEPS are probably not an accurate reflection of 5 million total users. MEPS surveys people living at home who respond to surveys. This entirely excludes people who are homeless, institutionalized, or have died earlier in the year, and it under-represents mentally ill or substance abusers. All of these categories have worse health, and regrettably are prevalent among vets, so MEPS almost certainly omits vets who reflect the highest burden of illness.
- This sample may lacks much statistical validity, even for the vets it does include. MEPS weights responses to make them nationally representative for demographic characteristics, but not for veteran status. Without this weighting, the chance of random error is much greater. This is suggested, for instance, by the fact that the value of VA care reported over this six-year study ranged two-fold from year to year, with no discernible pattern (the sixth year was twice the fifth year, which was half the third year, etc.)
- Even for those whom MEPS does represent, it underreports actual health care costs. Exactly how much and why is somewhat unsettled, but what seem to be the most recent studies conclude that MEPS underreports by 14% – 19%, in large part because reports of both utilization and costs are understated. Weeks acknowledges these possible flaws, but asserts that studies he and others have done show MEPS is reasonably accurate – again without citing any of the leading studies to the contrary.
Moreover, even Weeks’ self-selected cites do not fully support his accuracy claim. For instance, he says a RAND study reports that “MEPS expenditure estimates ‘agree quite well’ with estimates from other databases.” But, the RAND study (p. 34) spoke in that phrase only to utilization, not to expenditures, and even for utilization it said MEPS underreports by 85% for outpatient hospital use. For expenditures (use X price), RAND (on the very next page) said that MEPS underreports hospital costs by 21% and physician costs by 54%.
What is this Journal of Health Care Finance that would publish a flawed use of MEPS? It is hardly a leading health research journal. According to its website, it is
devoted solely to helping you meet your facility’s financial goals. . . . Make easier, better decisions, with advice from industry experts. . . . Experts in the field share their experiences on successful programs, proven strategies, practical management tools, and innovative alternatives, . . . including hospital/physician contracts, alternative delivery systems, generating maximum margins under PPS, improving productivity, taxation management, health care insurance, employee benefit cost-containment, joint ventures, mergers and acquisitions, employee incentive systems, and more.
An e-mail from its editor states that most articles are reviewed only internally, by its editorial board whose members are drawn primarily from industry.
It appears the Weeks article did not receive peer academic scrutiny, but what about scrutiny from the study’s own coauthors, who are affiliated with Dartmouth and Washington & Lee? The second author happens to be Weeks’ wife, and the third appears to be their son. As for Weeks himself, he is deeply embroiled in two serious legal controversies with the VA.
On balance, the Nugent, Hendricks et al. study remains unrebutted. In my view, the Weeks study suffers from too many serious flaws, and is too lacking in objective critique, to hold much or any credence in this important debate.