Predictive modeling goes to work for Medicare

The law sets in place a fast race among database software vendors for some very big contracts that could grow even larger within a few years.

The new law says Medicare claims in 10 states will go through a predictive modeling process, something many insurers already use to spot patterns in large databases. Predictive models are also used to compile credit scores and flag transactions for investigation.

For instance, while driving to Texas early this year I got a call from my credit card company asking about purchases of gasoline along my route. Such calls have become more common than new credit offers over the last years as banks tightened their credit policies.

As early as January, the Centers for Medicare and Medicaid Services will take bids on predictive modeling contracts covering the 10 biggest fraud states. (Larry Ellison just got real excited.) Implementation is due in six months, and after a year a report is due on potential savings, after which the program could be expanded.

Chances are good one of those 10 states will be Florida, where the issue is big in the state's governor's race and where some of the biggest recent prosecutions have taken place.

One Miami doctor has seen a 30-year sentence for Medicare fraud upheld. More multi-year sentences were recently handed out there. Increasingly, bail is being denied in these cases.

There is some concern that delays in payment will cause more doctors to reject Medicare patients, but all the political pressure is coming from the other side, especially now with international organized crime involved. And as I noted in my credit card example, modeling can work quite quickly even on large databases.

Besides, insurers are already using this technology, and are certain to increase their use of it.