Tuesday, June 22, 2010

HC Value in Mass: AG vs. Partners

In late January of this year, the attorney general's office (AGO) released a report that was supposed to show that
  • charges to insurances vary significantly between providers within the same geographic area;
  • variations in charges are not explained by quality, case severity and complexity, Medicare or Medicaid proportion of the case mix, or teaching hospital status;
  • variations in charges are correlated to market leverage (relative market position compared with others);
  • variations in charges are not correlated to the payment scheme - expenses were "sometimes" higher for globally paid providers compared with fee-for-service basis; and that
  • increase in price, not in utilization, caused "most of the increases in health care costs during the past few years in Massachusetts".
A couple of days ago, Partners Healthcare released an analysis by Paul Dreyer, PhD, a consultant to Partners Healthcare who had worked at the Mass Department of Public Health for over twenty years before he retired. His reported has also been released via the internet. Dreyer finds that
  • the AGO's methods are inappropriate as they use cross-sectional studies and uni (bi) variate [regression] analyses;
  • using appropriate uni (bi) variate [regression] analyses, you get different results;
  • in a multivariate [regression] model about half of the variance (R2=0.53) is explicable by the following variables (directly copied from report):
    • "DSH Dummy - A dummy coded variable indicating a hospital’s DSH [disproportionate share (of patients paid for by government programs plus free care)hospitals] Status
    • GEO Dummy - A dummy coded variable indicating a hospitals status as small and geographically isolated (ADC <20, 20 miles or more to next hospital)
    • # Specialty Teaching Programs - From the ACGME, the number of accredited specialty teaching programs, or 0 if the hospital has no teaching program
    • Cost Per Day - From the RSC 403, total patient expenses including capital divided by total patient days
    • Case Mix Index - From the Report";
  • the leverage argument cannot sustained when looking at all hospitals (rather than only six selected ones);
  • a shift to hospitals with higher charges only explains 0.6% of the cost increase in 2005-8; and
  • there is a variation in quality between the hospitals, and that this variation "is comparable to the variation in price presented in the [AGO's] Report".
So let's take a closer look at how the two sides assess the value of health care services in Massachusetts.

This is a very interesting story and reminds me of the peer-review process for a scientific journal! I witness this quite a bit as I am a co-editor for a journal called "Value in Health". Only that that their review is usually single- (the authors don't know the reviewers' identity) or double-blinded (neither side knows who the other is, until the final decision is, of course, to publish the manuscript - the reviewers will know then if they bother to find out). 
However, a major difference here is that the original report and the "reviewer critique" are public! This makes the arguments for us all transparent, which I highly welcome.

What I like about the Dreyer report is that it is very quantitative and shows most of the results. Therefore, I cannot comment much on the AGO's report (also, I'm not a lawyer).
Generally, I think it's a great idea to built multivariate models rather than assessing the relationship between two parameters at a time. However, this is only half-way! Have all variables that would a priori make sense as hypothesis been assessed as independent variables? And what about interaction (i.e., is there an association between the variables considered "independent") terms? As an example, geography could have influenced the case mix, and the case mix surely influences the DSH status. This is all cost stuff.
As for the quality, Dreyer assesses "Serious Reportable Events" and 30-day mortality rates for myocardial infarction (heart attack), heart failure, and pneumonia. In contrast to the other sections of the report, he does not use scatter plots but just summarizes the number of hospitals above or below national average. He argues that these differences are statistically significant; however, the number of national hospitals is huge (and therefore the standard error small), so it would have been more interesting to assess the difference between tertiles, quartiles, or quintiles in Massachusetts alone.

What is missing in both reports is a real assessment of how quality can (or cannot) explain differences in hospital and physician charges, and how quality relates to the other variables that, according to Dreyer, might explain variation in charges.

The ball is in the AGO's court again.


Post a Comment