The Medicare Advantage Stars Program: Who Keeps the Money, Is it Evidence-Based and What About Variation?

The Population Health Blog has been practically banished from our living room when the spouse is watching TLC's Say Yes to the Dress.  This reality show follows the travails of young brides as they search for that perfect wedding gown while also dealing family expectations, tight budgets and dubious body art. 

When the PHB stops in and pauses to watch the bridal drama unfold over more than just a few minutes, it naturally wants to share.  Mysteriously, the spouse treats its helpful outreach like some sort of provocation.
 
Little wonder.  Years ago, the PHB medical director learned that the best way to make itself unwelcome at dreary management meetings dealing with the Medicare Advantage STARS Program was to inconveniently offer various insights, such as:

"If we hit the STARS target, let's not keep the millions in bonus payments; they should be distributed to our enrollees or used to lower premiums in the form of P4P4P."
 
"Too bad there is no randomized trial that shows that any of this help patients live longer.  Should we tell them that this isn't evidence-based?"

"If there is a small improvement, it could be the result of case mix or statistical variation."

And its innocent Yes to the Dress insights?

"All those dresses look pretty much the same!"

"I hope the groom takes a look at that mom, because if that's what the bride is going to look like years from now, he may want to reconsider."

"Seems to me she's trying to fit 50 pounds of potatoes into a 30 pound bag!"

The Tipping Point for Desktop Analytics: A Watershed Moment in the History of Health Care

Germ theory in 1860. The Flexner Report of 1910.  Zombie immortality in 2012  There are only a few watershed moments like these in the history of U.S. health care and, after hearing the other speakers at the recent Star Ratings Congress in Las Vegas, the Disease Management Care Blog thinks it's found another one.

 It calls it "desktop analytics."

In its early health services research career, the DMCB's studies consisted of creating study protocols that included data collection and storage, very high end computing, statistical planning and a carefully contrived reporting format. The timeline typically spanned over several months, required high end computing, involved fussy Ph.D. level statisticians unaccustomed to exceeding customer expectations and ultimately having to convince a narrow, highly educated, and skeptical audience of the veracity of the DMCB's conclusions at a scientific meeting.

While that is still necessary in traditionally funded research studies, the story is now far different in mainstream health care and insurance settings.  Tapping electronic record or insurance claims data bases are now far easier. Statistical software packages are do-it-yourself and 'walk' users through the basics. Ph.D-level statisticians are unnecessary. Mainstream health workers have a working appreciation of measurement as well as trending and the folks inhabiting the C-suites use their in-house research conclusions in core business planning.  And it can all be done using desktops that cost a few hundred bucks.

At the Star Ratings Congress, the DMCB listened to speaker after speaker who presented highly polished insights about quality and cost that were developed thanks to in-house information systems and analytics resources that would have been unthinkable a decade ago. This advance in data management has enabled providers and payers to spot trends on a month-to-month basis, compare local performance to historical as well as national benchmarks and report outcomes to external agencies on a regular basis.  The research efficiency was astonishing.

It was also so taken for granted. It shouldn't be.  Compared to 10 years ago, the industry has gone from the wheel and fire to the internal combustion engine and automatic transmission.

The DMCB thinks its going to get better too.  While the electronic health record vendors have been notoriously inept at supporting data analytics, it's going to just be a matter of time until community-based providers can hit a function key on their keyboards and scan (for example) mammography rates by age, race, zip code and months since last visit.  Insurers will be able to project which enrollees with diabetes on three or more prescription drugs are least likely to take their medicines after controlling for co-pay and weather.

When we finally do figure out how to increase quality and reduce costs, it'll be because desktop analytics had finally reached the tipping point.

Coda: This has important implications for the Affordable Care Act's Coordinating Council for Comparative Effectiveness Research.  The Council may find that by the time a prospective CER study is complete that desktop analytics had already found the answer and the much of the industry had moved on.  Stay tuned.