When in time we started contemplating the elimination of tips from the Linkery, Ilooked for actual research on the subject, and found Michael Lynn’s then-recently-published“Tip Levels and Service”. This paper shows that in spite of what people think motivates theirtipping calculations, the quality of service has only a tiny effect on how much a restaurant customer actually leaves as a tip. In fact, the percent tip left by a guest is as much influenced by whether the server (if female) draws a smiley face on the check, or predicts good weather, as by the guests’ happiness with the quality of service…if you’re a server who wants to maximize your income, service quality should not be your focus….
The strength of their intervention is that they made it incredibly simple for adoctor to get a payout: 1) Measure blood pressure, 2) Treat accordingly. But that strengthmorphs into a weakness when it comes to what we’re supposed to do with this study. As the authors note, it didn’t require complex diagnosis—really any diagnosis, for that matter, since blood pressure is measured at every physician visit. And it didn’t require treating the blood pressure successfully; everyone who tried got an A for effort. So in this particular condition(hypertension), with this particular incentive structure, pay for performance worksbeautifully. The RCT was a success.
These two things are related.
Simple monetary incentives are pretty good at motivating people to do check box activities. And these incentives have their place in medicine. For instance, it is a good thing that everyone who can safely take a flu shot gets a flu shot, or every child in their 4 year old well visit gets a hearing and vision screening or the body part that is to be operated on is marked with an “X” with a Sharpie pen before surgery. These things are simple, direct and rely on a vast literature which supports doing these actions 10,000 times will produce significantly better results than not doing these actions 10,000 times.
The entire set of activities that actually promote good health are not check box activities.
They are complex, multi-variate problems with massive interdependencies. What if the blood pressure problem is due to a complex underlying root cause that the basic treatment that was rewarded masks for a few years? How does the incentive structure prevent proper care?
Most pay for performance systems are multi-tiered. The first tier are direct payments for following guidelines and best practice recommendations (flu shots, aspirin for people complaining of chest pains/tightness, handwashing protocals etc). These are the low hanging fruit which should produce strong signals of quality and reward. However, the second level is the more complex and higher pay-off levels. How do we promote health of a population instead of encouraging providers to only “treat” health patients and pocket the health incentives?
This is a complex problem with plenty of opportunities to royally jack things up (that is the highly technical insurance jargon). The solution has to been to create metrics that attempt to measure health care quality given a population’s initial starting condition. A population that is older and has more problems will be normalized in a different manner than a population that consists of 23 year old tri-atheletes. And even then, there will be odd results and perverse incentives that will need to get ironed out in multiple iterations.
(BTW — this is similar to the complex problem of actually measuring teacher value add — it is a highly complex problem where the simple solution of just taking test scores and looking at the yearly delta is amazingly wrong but appealing)