Let me suumarize some of the concerns about pay for performance, quote some of the skeptics interviewed by the American Medical News, then quote responses from the P4P advocates, and add my comments.
P4P May Use Outcome Measures Without Adjusting Adqeuately for Patients' Characteristics, Leading to Perverse Incentives
Dr Randall Maxey said, referring to patients' ability to comply or to afford care:
It's going to be a lot easier to treat a little old lady from Beverly Hills. Some communities are more compliant and more health-literate and have more resources to influence outcomes than others. I may treat you exactly correctly and give you the right pills, but if you have to choose between buying pills and giving your baby milk, that drug may lose out and my performance may be judged as poor because of it.Dr Roy M. Poses (that's me) said:
Outcomes are determined not just by what the physician does but by how sick the patient is, what his or her other characteristics are, and to some extent, by chance. If you don't control for patient characteristics, you can have a perverse system.In response, Janet Corrigan PhD, CEO of the National Quality Forum, said,
It's recognized by everyone that these are not measures that are under the control of an individual clinician, but there are important things a primary-care provider can to do to encourage patients to adopt the right behaviors.Then what sense does it make to use the measures to assess physicians? Outcomes that are beyond an "individual clinician's" control do not reflect that clinician's performance.
Furthermore, Dr Greg Pawlson, executive vice president of the National Committee for Quality Assurance, said,
Not everybody's patients can be sicker.No, but a system that gives poorer grades to physicians whose patients actually are sicker would be perverse.
Process-Based Measures Should Reflect Processes Actually Under the Control of Physicians
Dr Chuck Kilo said,
Do I get dinged if a diabetic chooses not to have their A1c tested, or is my recommendation sufficient to get credit? If so, then that leaves a lot of room for gaming the system.Measures Based on Administrative Data are Suspect Because the Quality of the Data May be Poor
Dr Kilo also said,
Health plans have been measuring practices and sending data back for a long time, and most doctors would throw them in the circular file. Somewhere between 10% or 50% of the patients they have listed as mine are not mine. It doesn't take a whole lot of erroneous data built into it for doctors to write off the whole thing.Measures Meant to Control Costs Will Not Measure Quality
If the measurement systems or incentives are really designed to save costs for health plans, they may push physicians not to do that which would be the best for the patient in terms of clinical care and clinical outcomes. The devil is in the details.Summary
It's nice to see some balanced coverage of this issue, given all the hype it's getting from managed care, government agencies, and payers.
Remember that there are problems with P4P as it's currently formulated that were not addressed by this article. In particular, most pay for performance measures so far are about primary care, or primary and secondary prevention. Few are about specialty care, diagnosis, or management of acute illnesses. Furthermore, most measures are targeted at single diseases, and developed from studies of patients with only one disease. Few take into account management of patients with multiple diseases (who are not rare), or management of patients with ill-defined complaints. Measures focused on only a fraction of medical care can lead to another kind of perverse effect. Pushing doctors only to improve their performance in very limited areas may reduce their time and resources to even maintain performance in other areas. Since performance in those other areas is not measured, no one may notice it declining.