Are metrics in medicine a good thing?

The Washington post reports on new efforts by insurance companies to rate doctors performance and their policies that penalize doctors for performing poorly according to their metrics.

After 26 years of a successful medical practice, Alan Berkenwald took for granted that he had a good reputation. But last month he was told he didn’t measure up — by a new computerized rating system.

A patient said an insurance company had added $10 to the cost of seeing Berkenwald instead of other physicians in his western Massachusetts town because the system had demoted him to its Tier 2 for quality.

In the quest to control spiraling costs, insurance companies and employers are looking more closely than ever at how physicians perform, using computers, mountains of health claims and billing data and sophisticated software. Such data-driven surveillance offers the prospect of using incentives to steer patients to care that is both effective and sensibly priced.

Now, on the surface, many people might say this is a good thing. But I will argue, that this kind of superficial measurement of performance will not only demoralize doctors but adversely affect patient care:

As medical records are increasingly digitized, the ability to mine them for information about physician performance becomes more and more tempting. It’s an obvious and noble idea, but is a good one? Just think about it. If you could rate doctors based on simple data culled from their records – like how well they manage blood sugars in diabetics, blood pressure in hypertensive patients, etc. – you could figure out how good a job they are doing in patient outcomes, right?

Well, luckily it has been studied, and this article in the New England Journal (free) shows what happened when the British NHS, which has extensive electronic record infrastructure, rewarded physicians based upon metrics of quality of care. I blogged about this before and I’ll give a similar warning – don’t just read the abstract, read the data too because I think the authors misread their own data a little. Really what they showed when they rewarded doctors for maintaining good health metrics in their patients was that doctors that treated the young, healthy, and rich did well, whereas those that served more patients, the poor, the elderly, and difficult patients were paid poorly. Also, those who filed lots of “exception reports” to justify the exclusion of a patient from the data set did the best of all.

Basically, they show that rewarding (and I suspect this will apply to penalizing as well) doctors based on patient health metrics led to doctors serving “easy” populations to do well, and those serving “difficult” populations to do poorly (or try and cook the books). There is no evidence it led to a significant improvement in care. I’m all for getting paid more, but the problem with penalizing or rewarding doctors based on how the patients perform is that it rewards doctors for avoiding difficult patients.

Now, on top of the problems shown in a national health care system, imagine the insurance companies in the United States rating doctors based on similar metrics. On top of the problem of rewarding doctors for seeing easy patient pools, you add a lack of transparency or appeals process, no system for exception reporting, and a patchwork of systems for each insurer. Since it’s a penalty system, that one patient who’s still smoking despite being on oxygen might cause your co-pays for all your patients to go up by 100%. What do you do? You try to avoid the poor, the elderly, and the non-compliant, because they won’t just cost you more money, but they may ruin your practice, and you won’t be able to do anything about it. Further, insurance companies often encourage doctors to use specific drugs and use less expensive procedures. The next step will obviously be to make sure that doctors perform the cheapest procedures they can justify, and select the drugs the insurance companies prefer because of rebate deals they make with drug companies.

So before people jump on medical rating systems as some great new idea, realize that superficial data collection might actually hurt medical care, especially for the most at-risk and ill populations of patients. If data collection is going to be tied to medical reimbursement, or worse, penalties, the system needs to be designed in such a way that doctors can see how the calculations are determined, they must be able to appeal to change their rating based on errors or on characteristics of their patients that might not be in the records, and it must not discourage doctors from seeing the patients that most need medical care.