I am uncertain about so-called "pay for performance." On one hand, offering incentives to doctors for better work seems a straightforward application of basic economic theory. On the other hand, the tactic of compensating "outcomes" rather than "processes" relies on assumptions and mechanisms that are not clearly understood, and might not always function as intended.
Unexpected consequences from introducing "P4P" are virtually guaranteed, but among the foreseeable ones is what I call the "catnip problem."
Getting their attention
If you toss catnip into a roomful of cats, they will theoretically drop whatever they are doing (not that it's always easy to know what cats are "doing") and head straight for the herb, setting aside whatever business they were formerly about. If this analogy holds for doctors, you would expect them to shift their efforts to whatever CPT codes bring the most reward, and devote less time to the rest. (Actually, catnip sensitivity is not universal among cats; it is an inherited trait, enjoyed by only about 66 percent of the feline population. I cannot find literature on the prevalence of financial sensitivity among physicians, but with the established similarities between cats and docs regarding "herding behavior," I would assume it is similarly distributed.)
There are lots of different schemes to pay physicians for meeting specific performance goals. Some are clever and possibly constructive. My worry is that many of the P4P plans I have seen only provide incentives for a measly handful of outcomes out of the galaxy of "things doctors should do." Might not the improvement gained in a few areas come at the expense of deterioration in everything else?
The ICD-9 codes we use for diagnoses comprise about 20,000 categories of diseases, and at least a couple of hundred thousand individual conditions. Compare this with the famous 2003 article on gaps in the quality of U.S. health care, where the authors used an assessment tool from Rand that listed 439 measures of quality for a mere 30 medical conditions. Their study showed, overall, that patients received 54.9 percent of the recommended care for these conditions.*
Or consider Medicare's P4P system for hospitals, aimed at just five clinical conditions. (Medicare has P4P demonstration projects for physician groups, but they mainly involve financial efficiency, not clinical performance.) Medicare also has initiatives for physicians that address three conditions: congestive heart failure, diabetes, and end-stage renal disease. I'm sure there are a few more I'm not aware of.
In the commercial insurance world, there are disease management programs for perhaps a dozen conditions. Their designs range from truly nifty to absurd. Without trying to be ironic (because it isn't necessary to try), these programs are mostly aimed at disorders for which we happen to have useful drugs, for which better outcomes can be represented by increased pharmacy utilization.
To be fair, the target diseases addressed are responsible for major morbidity, are overall poorly managed, and have treatments that work. OK, so it makes sense to start where the best returns will be.
But the point is that P4P programs reward doctors for only 3 or 12 or 30 (maybe 439) tasks. My math says that even if we provide catnip in 439 flavors, we could cause deterioration in 19,561 other "things that doctors should do."
If I had polycystic ovary syndrome, hepatitis C, or scoliosis, there would be no bonus for my doctors. Is it unreasonable to think that I might notice a shift in attention away from me and toward folks with more rewarding diseases? Would it be like the movement of obstetricians away from general OB and its annoying nighttime interruptions toward the more lucrative and controllable subspecialty of fertility medicine? Or like the defection rate among general dermatologists, coinciding with an upsurge in cosmetic spas offering expensive — and usually cash-based — treatments, ointments, and vitamins? Or like the tendency among orthopedic surgeons to avoid (again annoying and uncontrollable) trauma duty in favor of better compensated (and easier to schedule) sports medicine?
Paging Dr. Skinner!
Is it paranoid to imagine that many problems would experience a detriment in attention, management, and outcomes, if 30 were chosen for special reward? What would B.F. Skinner, the father of operant conditioning, say about doctors' susceptibility to behavior modification?
You might object, "Unfair! Simple utilitarianism tells us we have to devote our energies to improving conditions that promote the greatest good for the greatest number, and work down the list as resources permit." And you might argue that the innate altruism of physicians (OK, augmented by the innate threat of malpractice litigation) would ensure that people with disfavored conditions would receive the same quality of attention as those with favored ones. Just because some diseases happen to win celebrity status, this doesn't mean that prosaic disorders will be proportionately ignored.
Who are you kidding?
First of all, the utilitarian argument falls flat. The top 20 causes of death in the United States (2004, all ages) are heart disease, cancer, stroke, chronic pulmonary disease, accidents, diabetes, influenza/pneumonia, Alzheimer's disease, nephritis, infection, suicide, liver disease, hypertension, Parkinson's disease, homicide, pneumonia, aortic aneurysm, perinatal problems, HIV, and benign neoplasms. (The top 20 for infants are different, as are those for teens and young adults.) Why aren't we focusing on all of them?
And death, be not proud. Looking at disability, the top 10 causes worldwide are major depression, iron-deficiency anemia, falls, alcohol use, COPD (cigarettes), bipolar disorder, congenital anomaly, osteoarthritis, schizophrenia, and obsessive compulsive disorder. In the United States, top disability causes are: arthritis, spine problems, heart problems, lung problems, hypertension, stiffness or deformity in a limb, diabetes, blindness, deafness, and stroke. Depending on what you measure, you might add migraine, prostate cancer, dementia, auto accidents, heroin abuse and sinusitis.
Conservation of energy
How do you pick some kind of all-star team from this catalog? To me, lists like these illustrate the dubiousness of picking five, or 30, or even 1,000 conditions that deserve doctors' attention "first." Isn't this like the problem we have now with competing "disease-of-the-moment" charities and "program-funded" popularity contests in legislatures? This is the core problem that bothers me about pay for performance. I would think the law of conservation of energy would apply: A doctor can't put more effort into a few diseases without diverting attention away from the rest.
Proof in poor outcomes
That, in a nutshell, is the argument in favor of general "process" improvement over "outcomes." If you examine physician behavior and ask, "What are the reasons for the current shortfall in performance?" you will usually find the answer in systematic failures affecting every disease treated in a given location, rather than specific deficiencies in managing a handful of sentinel problems.
Take the tiresome example of diabetes. Everybody knows that doctors do a lousy job managing it. The proof is in poor outcomes, which seem intractable to medical education events no matter how lavish the restaurants where they are held. The remedy, according to a slew of professional organizations, pharmaceutical companies, specialty board examiners, and well-intentioned educators, is apparently to indoctrinate primary care physicians to the point of coma in the Byzantine intricacies of diabetes physiology.
The evidence for doing this seems to be nothing more than the fact that it has not worked so far. I guess the theory is that if something doesn't work, you need to do a lot more of it.
Keep it simple
If you look at the better disease management programs, you see a suggestive commonality. Typical elements include focused patient communication, easy access to counselors, simple goals, evidence-based guidelines, stratification, tracking, and feedback. Why not apply these techniques to medical practice in general, rather than to a mere handful of diseases?
If airplanes tended to crash a lot in fog, would the solution be to pay pilots more on foggy days? They would rather have radar.