The Majority of Medicare Spending Variation Is Unexplained

From time to time, the Population Health Blog spouse finds that her husband is insufficiently attentive.  During a recent conversation about that very topic, things stopped when the PHB pointed out the window and exclaimed "Look! A squirrel!"

Naturally, the spouse is curious about the PHB's erratic attentiveness.  Is it how its brain is hardwired?  Too much sugar? Substandard parenting?  And of all those possibilities, how much do each contribute?

That introduction should help the PHB and its readers check out this just-published Health Affairs paper on Medicare's erratic spending habits.

As PHB readers know, Medicare's patient costs patient vary from one locale to another by thousands of dollars, with no discernible impact on survivorship or quality. One narrative is that the health system is being consciously or unconsciously manipulated by doctors and hospitals at a regional level.  Another is that poverty is causing patients in some areas of the country to have have more than their fair share of health problems.

Enter Laurence Baker et al, who wanted to know if patients' preferences are playing a role.

The answer is that it does. But, compared to hospitals and patient income, not that much.

The authors obtained Medicare claims data from 2005 and sorted it by Medicare Hospital Referral Region (or "HRRs," which can span several counties). They wanted to know if HRR costs correlated with 1) county and zip code-level median income, 2) self-reported health status, 3) the availability of doctors and hospital beds and 4) a six question survey that ascertained respondents' preferences for care based on scenarios like chest pain or cough.

The results?

The HRRs were grouped and sorted into low to high spending quintiles.  As the quintiles increased, so did the number of hospital beds per thousand (2.2 low to 2.5 high), which suggested that the supply of services increases health care utilization. Doctors were negatively correlated (the more docs, the lower the spending, 214 per 100K low vs. 193 per 100K high). 

Income was not correlated.

Patient preferences were correlated but only by a small amount (just over $100 across the quintiles).

It's one thing, however to have a correlation, it's another to know the strength of the correlation.  Using regression analytics, the authors found that the availability of hospital beds and doctors could independently account for 23% of the low to high variation across the quintiles. Health status and income seemed to drive another 12%. Patient preferences explained another 5%.

While that explains approximately 40% of the low to high variation across the quintiles, that means 60% remains a mystery. 

In other words, if hospital services, patient economic disparities and patient preferences were completely neutralized by very enlightened central planning, wholly just income redistribution and perfect patient education, only 40% of the cost variation across the United States would go away.  Boston would still cost more than Boise.
 
The PHB's take:

1) Squirrels abound: there is still a lot that we don't understand about the national swings in Medicare's costs.  Some areas are cheap, others aren't and the majority of that has little to do with the availability of hospital services, poverty or beneficiary preferences.

2) Any wonk, policymaker, politician, academic or blogger who offers "a solution" to Medicare's variation is kidding themselves.  The majority remains outside the reach of laws, regulations or payment reforms.

3) Compared to Medicare, PHB's variable attention span is a comparatively modest problem.  The spouse should take some comfort in that.

Image from Wikipedia