Getting to the Core of Geographic Variation

On the heels of the American Hospital Association’s recent demonstration of gross discrepancies in the Dartmouth group’s data, MedPAC released its December 2009 report to Congress showing the same. Confirming data for 2000 (reported in their 2003 report), MedPAC demonstrated much less variation among states and metropolitan statistical areas (MSAs) than described by Dartmouth for states or hospital referral regions (HRRs). Closer scrutiny of MedPAC’s data reveals even more.

Adjustments. First, the Dartmouth group has claimed that they adjusted their measures of Medicare spending for age, sex, race, mortality, disease incidence and prices. But Dartmouth’s adjusted data are indistinguishable from MedPAC’s  unadjusted data, both among states (in 2000) and HRRs (in 2006), as shown to the left. This confirms suspicions that Dartmouth’s “adjustments” are all shadows and mirrors, or just malarkey.

Adjusted variation. Second, as reported by MedPAC and consistent with the above, MedPAC found much less variation in Medicare spending among MSAs after adjusting for prices, health status and special payments than Dartmouth found among HRRs after supposedly adjusting for prices and a host of other factors. The two figures below demonstrate that. One need only look at the broadly dispersed bars in the illustration of Dartmouth variation and the more tightly packed ones in MedPAC’s.








Sociodemographic realities. But, despite finding much less geographic variation, MedPAC pointed out that there still was plenty. The greatest likelihood is that most of this residual variation is related to differences in income, which MedPAC does not account for (except as it correlates with health status). The final graphic supports that view. It distinguishes a cluster of eight southern states, which house 89 MSAs, from the other forty states (Alaska and Hawaii were excluded), which house 312. Compared to the other forty, the “southern eight” has a poverty rate that is 55% higher (89% higher for blacks); its rate of uninsurance is 31% higher (89% higher for blacks); and its mortality rate is 16% higher (30% higher for blacks). Excluding the “southern eight,” Medicare spending is within 10% of the mean in 87% of the MSAs in the other forty states. Only five MSAs out of 312 fall beyond +15%.

While no one would deny that the practice of medicine varies among practitioners for a host of reasons, both good and bad, such variation is not responsible for geographic variation, or at least not usually. Rather, geographic variation in health care spending reflects geographic variation in prices and variation in two patient-related factors: income and burden of disease. As I said before, the Dartmouth Atlas is the “Wrong Map for Health Care Reform.” Getting to the core of geographic variation can help us get health care reform right.


  1. Pingback: Myth 33. Reducing geographic disparities will reduce spending without sacrificing quality. « AAPS News of the Day
  2. Robert W. Geist MD

    Why do I find few colleagues or local policy gurus to believe that geographic variation is due povertyand disease burden, not “poor” medical care and FFS avarice? It’s crazy out there even with many academics paranoid about “culprits” without bothering to look at tax-subsidized demand inflation and poverty and cultural status problems as explnations for medicine’s ailments? What locks up minds? Bob

  3. Pingback: The Road Back from Dartmouth Deception Will Be Difficult, but We Must Now BeginDartmouth is Dead « PHYSICIANS and HEALTH CARE REFORM Commentaries and Controversies
  4. Pingback: MedPAC, Poverty and Geographic Variation in Health Care | PHYSICIANS and HEALTH CARE REFORM Commentaries and Controversies
  5. Pingback: MedPAC, Poverty and Geographic Variation in Health Care | Federal Healthcare Reform

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s