For years I have railed against P4P.  The concept seems so right to the wonks and politicians. The concept seems right if you have shallow thinking.  If you ask practicing physicians to do a pre-mortem analysis on how P4P would fail, they would have understood.

Last week in my favorite thought provoking blog (Farnam Street), they published a wonderful piece that explains the problem well – The Distorting Power of Incentives.  While I hope you read that post, I will provide the essence.

Understanding incentives comes through second-and-third-level thinking. Many incentive systems have backfired because people failed to consider other interests and incentives.

P4P rewards physicians or systems for meeting certain measurements.  First level thinking involves identifying a guideline, learning that adherence to that guideline varies, and then providing rewards for certain percentage of adherence.  Second and third level thinking looks at the entirety of patient care.  It asks whether shining the light on a set of measures might detract from other aspects of patient care.

Good incentives acknowledge recognition, public perception, and the value of pursuing work that we can be proud of. So yes, if we want to persuade, we should appeal to interests not reason. But when it comes to interests, appeal not just to net worth but also to self-worth.

While money has great importance, most physicians value more than money.  We want to do what is best for the patient.  We have our own concepts of what a good physician does for their patients.

We know that P4P plans have some influence on the care patients receive.  Occasionally they improve targeted care, but they can detract from care that does not have a measurement.

P4P is the wrong idea.  It represents shallow thinking.  The concept ignores what we know about motivation.  It ignores what we know about the complexity of patient care.

It ignores an understanding of the impact of incentives.

During my India trip, one physician told me this story. I will try to get the gist of the story correct. He told the story in response to my lecture on the dangers of guidelines and performance measures.

He was practicing in England at the time this incident occurred. England had just started their P4P project. An older woman was going to see her physician. She was having symptoms of uterine prolapse. Her daughter accompanied her, but she did not let the daughter come into the examination room because of embarrassment.

The physician comes into the room and starts reviewing each of her known medical problems with a focus on those issues that would impact his performance measures. He never asks for her agenda, and abruptly finishes the meeting.

The woman goes back to the waiting room, and has to tell the daughter that she never had a chance to seek help for her concern.

Performance measures can change the physician patient interaction. We are told that medical care should become more patient centered, while focusing on performance measures changes the physicians priorities. We do not have a good measure this concept.

Since quality has many dimensions, we must worry that focusing on some dimensions will decrease our attention to other important dimensions. This story describes a bad medical visit. But the physician likely scored perfectly on his measures.

Readers know how I would vote on this debate.  Should Physician Pay Be Tied to Performance?

The con side is well done, however, it does miss a major point.  Performance is too complex to summarize in measures.  We have a responsibility to individualize therapy – be patient centric. Some patients deserve different goals than others.  Trying to assume that we should always treat each disease in the same way ignores the co-morbidities, the social situation, the patient’s age and many other issues.  Performance measurement does not apply to every patient, because we will never have performance measures for every disease and every manifestation of the disease.  Performance measurement does not value diagnosis.  It does not value bedside manner – patient satisfaction is NOT a proxy for bedside manner.

P4P is a horrible idea for many reasons.  But the wonks will not give up this idea.

Rarely does a policy paper stimulate excitement. I have to admit that I did not expect RWJ to produce such a document. So bravo, kudos, and congratulations are in order.

Bob Berenson writes about this new paper – Seven Policy Recommendations To Improve Quality Measurement. You can find the policy paper at the RWJ site – Achieving the Potential of Health Care Performance Measures

Here are their summary recommendations:

1. Decisively move from measuring processes to outcomes.

2. Use quality measures strategically, adopting other quality improvement approaches where measures fall short.

3. Measure quality at the level of the organization, not the clinician.

4. Measure patient experience with care and patient-reported outcomes as ends in themselves.

5. Use measurement to promote the concept of the rapid-learning health care system.

6. Invest in the “basic science” of measurement development.

7. Task a single entity with defining standards for measuring and reporting quality and cost data, similar to the role the Securities and Exchange Commission (SEC) serves for the reporting of corporate financial data, to improve the validity, comparability, and transparency of publicly-reported health care quality data.

While we might have a few discussions at the margins, this policy paper goes directly against some poorly considered mandates in the Affordable Care Act.

Berenson writes:

…I work mostly inside the Beltway in a world of policy makers who, despite good intentions, by their actions often display a lack of understanding of the challenges associated with measures, measurement, public reporting, and pay-for-performance. For example, the physician value-based modifier, which was mandated as part of the Affordable Care Act and now must be implemented by CMS, cannot produce a valid snapshot of an individual physician’s “value” but will be imposed nevertheless, unfortunately feeding those within the physician community who resist all efforts to improve accountability and transparency of performance.

As much as I have written against P4P at the physician level, I do support P4P at the hospital level, if the measure has good documentation. Central line infection rate should decrease – and incentives should drive the cultural changes necessary to achieve that safety outcome measure.

So I applaud Bob Berenson, Peter Pronovost and Harlan Krumholz for looking objectively at this problem. Their recommendations deserve widespread adoption. We who blog should work to make knowledge of this paper widespread. It is very important, nay, critical at this time.

Yesterday, I received a comment that assumes that without P4P physicians will not work to improve their performance.  That comment shows an incomplete understanding of the behavioral economics.

Physicians get to medical school partly because they strive.  We strive for excellence through high school, college and medical school.  We need not link pay and performance, but we should carefully provide performance measures to physicians.  This distinction is important.

A good friend is the chief medical officer for an insurer.  He describes it well.  The vast majority of physicians live in the 25-75% range on any performance measure.  They are not the problem.  Small differences do not matter.

He worries about the lowest 10% and tries to learn from the top 10%.

If you have a vaccination rate that is statistically lower than your peers, then you should know and most physicians will try to remedy that deficiency. 

The problem with Pay is that it changes the motivation process, as the blog post I cited yesterday explains. 

Physicians have enough internal motivation and that trumps external motivation.  Attempts at external motivation almost always have unintended consequences.

I hope this post makes clear my thoughts.  One other caveat.  We still must be certain that our performance measures are meaningful and valid.  We have many that are, but unfortunately we still have many that are not.  But that is a story for another day.

The Health Affairs Blog has a wonderful post – Will Pay For Performance Backfire? Insights From Behavioral Economics

This blog post describes the behavioral economics about P4P.  The co-author, Dan Ariely, is a well known professor of psychology and behavioral economics.  In the post, he and his co-authors provide the research basis that argues against any benefit from P4P.

As I have blogged repeatedly, we have no evidence that P4P improves care, while we do have evidence that P4P can harm.

We must always understand that P4P is an intervention.  It has the properties of any other intervention.  Many of the same people who want to restrict pharmaceutical use and device use until we have solid evidence still support P4P.  Why do they not demand the same evidence of improvement and lack of harm from this intervention?

P4P is no different.  Please read the article.  It expands on the ideas that Daniel Pink has include in his book Drive.

I spent yesterday at ACP's Performance Measurement Committee meeting.  We spent hours discussing performance measures both specifically and generally. 

I concede that measuring some performance can provide useful feedback to physicians. More than that performance measures might evaluate systems of care.

I particularly like the concept of performance measures directed at overuse – CT scans without indications, over treatment, etc.

But I continue to argue against paying incentives to physicians to improve their performance.  This 2009 column from the Heath Brothers says it loudly – Why Incentives Are Irresistible, Effective, and Likely to Backfire

We also are cursed with too many performance measures that have no good evidence base.  We must speak loudly and tell everyone that the Emperor Has No Clothes.

Readers know that I have preached about an unclothed movement for many years.  Proponents, who claim to support evidence-based medicine, ignored the lack of evidence that a P4P plan would improve care.  Proponents KNEW that this strategy must work.  But it does not, and it likely hurts overall patient care (although this paper does not make that point).

Little Evidence for or Against Pay-for-Performance Plans

Six of the 7 studies in the review showed "positive but modest effects on a minority of the measures of quality of care included in [each] study," the authors write. This faint praise was further weakened by design flaws identified in the Cochrane review. For example, performance bonuses paid by private insurers in 6 of the studies went to the group practice, not individual physicians, and these studies did not report how the groups distributed or otherwise used the bonus money.

"If tasks are delegated to a physician assistant or nurse, then the issue is raised as to whether they were paid a share of the financial reward," the authors write, noting that who receives what has a bearing on effective teamwork.

A more serious design flaw, the authors point out, is the failure of all of the studies to address selection bias — that is, how did physicians come to participate in financial incentive programs?

"Physicians who provide poor quality of care may withdraw from the health plans providing these schemes, choose to contract with health plans that do not have incentive schemes, or choose not to participate in the study, thus leaving the ‘better performing’ physicians in the study," the authors explain. "Observed improvements in performance may therefore be due to selection rather than an actual change in physician behavior."

Future studies of pay-for-performance programs, the authors conclude, should be more carefully designed to address selection bias and other issues with study design.

Why do the proponents not understand that adopting P4P means that you are inducing a therapeutic intervention?  They would quickly demand good data prior to adopting a new drug or device.  Why would anyone think that P4P would not have the same potential for unintended consequences?

I believe that the proponents have a paternalistic view of practicing physicians.  These "do gooders" believe that practicing physicians provide substandard care.  If only those physicians would listen to the experts!  

But one cannot assess performance just from numbers.  Numbers exclude context, and in medicine context always trumps numbers.  We must provide the best care for our individual patient, given all their personal constraints, beliefs, other diseases, and expectations.  

Bravo to Cochrane for doing this study and publishing it.  Perhaps we can start to control the madness, and refocus medical care on the patient rather than the numbers.

I am giving a grand rounds on the Dangers of guidelines and performance measurement in about 10 days.  In revising an old talk, I have spent much time focusing on performance measurement.  Therefore I will propose these assumptions.  Please critique this formulation:

1. Making the correct diagnosis is not a problem
2. We can treat one problem without any impact on the patient's other problems
3. Adding drugs to meet a goal (e.g., HgbA1c, BP, LDL) carries no risks
4. Intermediate goals lead to better quality
5. Dichotomous goals are satisfactory
6. Only common diseases define quality

This list likely is incomplete.  Therefore, I need your help.  Challenge me.

Reactions to Atul Gawande's commencement address (and my rant challenging some of his assumptions) attracted many comments.  In reading the comments, I had a minor epiphany.  Checklists work will under the same conditions as external incentives.

Daniel Pink writes in Drive (I should get royalties as often as I cite this book) that performance incentives work for tasks that do not require complex cognitive involvement.  For example, you can incent bricklayers to lay more quality bricks through incentives.  However, external motivators do not help for problems having intrinsic internal motivation.  

Following this logic, I would argue that checklists are perfect for repetitive tasks at risk for error.  The prevention of central line infections is a perfect example.  Placing central lines is really a manual task.  The checklists reminds the operator of all the things one must do to decrease line infections.

In outpatient medicine, one could devise a routine diabetes visit checklist – describe the foot exam, describe the eye exam and record the date of the last eye exam, record the LDL and specify any treatment, record the BP and specify any treatment, record the HgbA1c and specify the treatment.  If the values are not at "target", specify any medication adjustments or why you do not plan any adjustments at this time.

However, if the diabetic patient comes to your office with chest pain, the checklist is no longer relevant.  Chest pain does not have a simple checklist because deciphering chest pain is a complex cognitive task.

In Gawande's terms, we can use a pit crew to address checklist problems, but we need a conductor (my term) to decipher diagnostic questions.  Diagnosis requires complex thinking and interpretation.  Such tasks are not improved through external rewards.  The task itself is intrinsically motivating.

For the same reasons that I oppose P4P for physicians, I urge the checklist devotees to understand the conditions that make checklists valuable.  Checklists help us perform well in routine situations.  The likely will not help us with complex decision making.  They may encourage mistakes, similar to the problem I ranted about 2 weeks ago – The danger of diagnostic labels