Just released in the BMJ, “Association between the Value-Based Purchasing pay for performance program and patient mortality in US hospitals: observational study“. Let’s not waste time. We’ll dive right in.
Everyone loves “pay for performance”. Medicare’s version to incentive quality is known as the Hospital-Based Purchasing Program. Basically, it provides financial inducements to reduce 30-day mortality for acute myocardial infarction, heart failure, and pneumonia. The program began in 2011, and is intended to be budget neutral. Medicare withholds some payments to hospitals in the program, and then gives it back based on their performance. In 2015, for instance, 1360 hospitals were penalized and 1700 received bonuses.
The program is instituted nationally, and is not voluntary. Performance is judged either against a national benchmark, or against improvement from the baseline period.
For this study, 4257 acute care hospitals were involved. Of those, 2919 participated in the program; another 1348 weren’t eligible for the program and served as controls. In these hospitals, more than 2.4 million patients were admitted from 2008 through 2013. The main outcome of interest was the 30-day risk adjusted mortality rate for acute MI, heart failure, and pneumonia. Comparisons were made to mortality from non-incentivized causes. Researchers were also interested in how the program affected those at the low end of the performance spectrum – those with the most to gain from improvement.
In the pay-for-performance hospitals, the mortality rates of the incentivized conditions dropped 0.13% in each quarter in the pre-intervention period in the study hospitals, compared to a drop of 0.14% in the control hospitals.In the post-intervention period, study hospitals dropped 0.03% each quarter compared to 0.01% in the control hospitals. This was not a statistically significant difference. In fact, there was no difference in any subgroup of hospitals.
I gave a talk last week to a bunch of hospital executives on how policy often fails to be evidence based. My last example was pay for performance. They seemed least likely to accept that example as correct.
It’s not that I think we can’t incentivize physicians to practive better. I’m sure we can. My problem is that we assume that we can pick an easy to measure metric (30-day mortality), tell everyone that this is the one to measure, that it translates into improved quality, and then expect results.
It still feels like the drunkard’s search. It still feels like it ignores other differences outside the health care system’s control. Most important, it still doesn’t seem to work.
But, you know, full steam ahead!