HC Statistics

State Health Care Rankings

New rankings are available detailing the quality of care received in different states. The Agency for Healthcare Quality and Research (AHQR) gives some State Snapshots based on National Healthcare Quality Report. My home state of Wisconsin is ranked #1 according to the Milwaukee Journal Sentinel (“State is No. 1…“). The Commonwealth Fund also has a report ranking each state. The Commonwealth Fund Rankings are as follows:

  • Top 10: HI, IA, NH, VT, ME, RI, CT, MA, WI, SD.
  • Bottom 10: OK, MS, TX, AR, NV, LA, KY, WV, FL, GA.

How seriously should these rankings be taken? In my opinion, not seriously at all.

Let us say that each American individual living in country X would have a health level of x. If an individual would move to another state j, then their health level would be (x + aj). Ideally, states would by ranking by ordering them from the ones with the highest level of aj to the lowest. Empirically, however, we only view (x + aj). Thus, some states may have poorer, less educated, non-English speaking residents who would have worse health regardless of which state they are in. Thus, scoring well on the dimension “Cancer deaths per 100,000 population per year,” which is measured by the AHQR’s quality report , does not tell the reader whether treatment is of a high quality or if the state’s population is just healthier to begin with.

Let us look at the state GDP levels according to the Bureau of Economic Analysis’ (“Regional Economic Accounts“):

  • Top 10: DE, CT, MA, NY, NJ, AK, CO, VI, CA, MN
  • Bottom 10: ID, ME, KY, AL, SC, OK, MT, AR, WV, MI

We see that 4 of the states with the worst health rankings also placed in the bottom quintile in terms of state GDP. Only Maine ranks in the top 10 in the health rankings and has a low GDP. None of the richest states rank in the bottom 10 Commonwealth Fund health care rankings. Finding a correlation between wealth and health is not surprising, but the exact relationship between the two is difficult to pin down.

In order to try to counter this selection bias based on the initial health levels of residents living in each state, most rankings include process measures in addition to actual health outcomes.

For instance, looking at individuals who have acute myocardial infarctions, if a high percentage of these individuals receive beta blockers within 24 hours of hospital admission or are prescribed aspirin at discharge, then a state will rank highly on these dimensions. While many of these process measures seem to be a good reflection of medical care quality, they may also reflect the characteristics underlying population, as well. For instance, ranking states based on the percentage of women age 40 and over who report they had a mammogram within the past 2 years (as does the AHQR), may indicate good health care policies by providers and health plans, or may indicate that the state’s population is more concerned with preventative care than residents of another state.

Also, states with high levels of health insurance usually gain extra points in the ranking, but it has not been conclusively shown that high levels of health insurance increase health levels.

So what can we use these reports for? While the overall rankings are not too instructive, policy makers can look at specific dimension to help improve care. For instance, why does New Mexico rank below average for “Receipt of recommended care for acute heart failure among Medicare patients?” Policy Makers could try to improve care by increasing the percentage of “Heart failure patients with left ventricular systolic dysfunction prescribed ACE inhibitor/angiotensin receptor blocker at discharge.” Most of the quality rankings, however, only take into account simple procedures, likely are not an accurate representation of a state’s medical care quality.


  1. Interesting analysis. But it misses the fact that that healthcare portion paid by employer in actually decreasing for poor Paul and potentially Median as opposed to Rich. There clear data indicating that the percentage covered by american employers dropped from the mid 60% to the 50%. And most of drop affects Poor and some Median.

    I would rewrite your little model to account for this fact to be credible…


Comments are closed.