Pay for performance, data mining - Realities and Limitations of P4P and Data Mining

Pay for performance (health care), is an emerging movement in health insurance, in which providers are rewarded for quality of health care services.

Data mining is the process of sorting through large amounts of data and picking out relevant information. It is usually used by business intelligence organizations, and financial analysts, but is increasingly being used in the sciences to extract information from the enormous data sets generated by modern experimental and observational methods

Wikipeda

The September 8 American Medical News features a front-page piece “Practices Hit Medicare P4P Quality Targets, But Bonuses Still Fall Short.”

The Story

As part of a 4 year Medicare P4P project, the story reports in its second year, 10 large medical groups – Dartmouth-Hitchcock Clinic, Marshfield Clinics, University of Michigan Faculty Practice, Everett Clinics, St. John’s Health System, Geisinger Health System, Park Nicollete Health Services, Middlesex Health System, Billings Clinic, and Forsyth Medical Group – hit 93% to 100% of 27 quality measures, but only four received bonuses, and those bonuses didn’t cover the cost of participating.

Terms of Paying Groups

The idea behind the project is that participating medical groups will receive 80% of savings generated by using P4P to reduce complications and hospitalizations. But the groups will only be paid if savings exceed 2% of the results of a community control group.

The trouble is doctors take the financial risks, but financial rewards are usually lacking or inadequate. Robert Bennett MGMA commented, “It’s taking a lot of money to participate, and they’re not getting it back. It the large groups can’t do, it makes you wonder if the small guys can.”

Mining Data to See Who Gets Paid

Part of the project also involved data mining to see if doctors were meeting their targets and improving outcomes. All of this makes one wonder if the hoopla about P4P and data mining is worth it. Big groups may have the infrastructure to participate, but what about small practices?

If small groups, which provide 80% of health care, don’t measure up or participate will they be excluded or punished financially? In view of the looming physician shortage, the use of data to see who plays and how much they’re paid may be an exercise in futility. Small groups are already under financial pressure and are swamped with patients. Only 7 to 10% have complete EMR systems – which may be necessary to track P4P compliance.

The Numerati

There’s a book just out, Numerati, by Stephen Baker, reviewed in the week end edition of the Wall Street Journal on September 13, The reviewer offers these thoughts on data mining.

The world is buried in data, great banks and drifts of the stuff. In recent years, a new technology has emerged – compute programs that will drill through it all to pick out patterns and trends – information that may be useful to marketers, employers, doctors, matchmakers or national security analysts. Such programs are extraordinarily sophisticated, and their creators need to be very clever indeed. A doctorate in math or computer science is pretty much required. Stephen Baker calls such whizzes the “Numerati” Using “data mining” they can seek out veins of useful ore in the mounts of facts that computers accumulate every day.

Closing Comments

To which I say,

Put me down as deeply doubtful, that all P4P data is useful, even if provided by the snoutful. No matter what your volume and quality of data, you can’t document or erase all human errata. Nor can you discern all clinical patterns and trends, no matter how sophisticated your algorithmic lens. I’m sure if you’re one of those PhD numerati, you may think doctors are haughty and naughty, and data drilling will reveal all, beyond ordinary mortal recall. But doctors are smart, they will find their way around, all those “facts” embedded in that gigantic data mound. You can parse, analyze, aggregate data, and do your data mining, but unless you’re there at the point of care, which in case of data experts is mighty rare, numbers won’t help much in care redefining. Finally, if P4P costs exceed the reward, chances for P4P success are untoward.