The delusional pursuit of metrics

Two points:

  1. Measuring something is not the same as improving it. 
  2. Improving something requires thinking of quality as a process rather than a product.

Before we think about how these two principles apply to medicine, let’s consider two approaches to coaching basketball: one using incentives tied to outcome metrics, and one using interventions designed to identify and address process problems.

Imagine you own a basketball team and want to improve its performance. You look at the stats for the most successful basketball teams over the last 20 years and find the following correlations:

• >25 points from 3-point shots per game

• 2-point shooting percentage > 60%

• Free throw shooting percentage > 80%

• Points on turnovers > 15

• Points against on turnovers < 5

You set up an incentive plan to pay players a bonus for shots made, for points scored on turnovers, and impose penalties for lost balls. What happens? The coaching literature is clear: players will take more shots and commit more fouls, but the team will not play better or win more games.

Alternatively, you could watch game films and study your team’s play. You find some or all of the following problems: poor fitness causing defensive breakdowns in the fourth quarter; poor inbounding against 1:1 defense; one or more players with poor dribbling, passing, offensive rebound, ball handling or shooting skills; one or more players with poor motivation, identified by poor attendance and lack of effort.

Based on this information (and the hypothesis that improving these could improve performance) one could: set up a conditioning program; practice 1:1 inbounding in tight spaces; practice shooting, passing, dribbling, offensive rebounds, ball handing and shooting (as a team or individually, depending on the identified deficits); incentivize those with poor motivation, based on attendance and effort ratings during practices and games.

The coaching literature is clear that this approach works. Improving team performance requires identifying and specifically addressing weaknesses. Incentives only work if one or more of the weaknesses are related to poor motivation. Paying more for three-point baskets can motivate a skilled shooter who doesn’t bother, but does not teach a player how to shoot better, or put the ball in his hands at the right time and in the right place. And it may cause the unskilled shooter to attempt 3 point shots rather than pass to the skilled shooter.

In medicine one sees the same two approaches, and with the same results. Some institutions try to ‘buy’ quality by tying payment to metrics. Others try to improve quality by improving the process of care. The results of both approaches are predictable.

Successful quality improvement programs focus on improving the actual process of care. They use outcome metrics, but only to assess progress and NOT to apportion rewards. They start by evaluating systems to identify weaknesses and deficits. They then set up interventions targeted at the identified issues. Metrics are used to see if interventions make a difference. (The metrics are used to evaluate the intervention, not the individuals working in the system.) In the case of diabetes, comparing populations of patients with well controlled diabetes to those with poorly controlled diabetes shows thatpopulations with fewer diabetic complications have better insurance, fewer co-morbidities, better education, and lower A1cs. Poor insurance, co-morbidities, poor education and lower A1cs do not occur because the clinician is underpaid or under motivated. Paying clinicians more if their diabetics are better insured, have fewer co-morbidities, have better educations or have lower A1cs are all equally senseless approaches

What WILL work is examining the barriers to good control in those diabetics who have poor control and are at greater risk for complications, and then specifically addressing those barriers. Some interventions that have been shown to work include: longer appointments, financial support for medication purchase, use of point-of-care evidence based decision support, more intense education from a diabetic team, follow-up calls from a trained nurse clinician after visits and at regular intervals between visits, social work interventions to address transportation, shopping, diet and exercise issues. A few studies have shown that paying for better A1cs results in etter A1cs, but there are no studies showing that paying clinicians more for lower A1cs in their patients improves clinical outcomes or reduces complications. None

Paying clinicians more for improved control in their diabetic patients would only improve care if two things were true: (1) the care was poor because the clinicians were not motivated; and (2) the incentive is great enough to motivate the clinician. 

Setting up a program that pays bonuses to clinicians based on quality metrics and calling it a quality improvement program (despite the absence of any actual quality improvement activities) is foolish, insulting, and sadly common.

  • It is foolish, because there is no evidence that it works, and considerable evidence for negative unintended consequences.
  • It is insulting because it is based on the premise that patients get sub-standard care (and have sub-standard outcomes) because their clinicians are not motivated, and that offering a small financial reward will motivate clinicians to care enough to do a better job. 
  • It is common because the ‘pay for metrics’ approach is cheap and easy, and because those tasked with QI processes either do not understand the principles of human behavior change, or because their institutions are too lazy or short sighted to do the work involved in real quality improvement.

A delusion is fixed false belief that is resistant to reason or confrontation with actual fact. The idea that paying clinicians for improved A1cs, colonoscopy screening rates, or check boxes that claim that smoking cessation advice was offered will improve quality is delusional.

Full. Stop.


Links to more on this topic::