Be careful what you measure, people's lives depend on it.

An example of unintended consequences when changing complex systems, and especially arising from poor selection of performance metrics.

From It Sure Looks Like This Obamacare Program Has Led to More People Dying by Peter Suderman.

To determine whether a government program is successful, it's often necessary to look not only at how well it does what it's supposed to do, but what it's doing that it isn't supposed to. For example, killing people.

Take the hospital readmissions program built into Obamacare. The program derived from a simple observation that hospitals were treating lots of people who would then return for more treatment within the month. Unnecessary readmissions cost Medicare an estimated $17.5 billion a year. If hospitals were treating people effectively, the thinking went, those people shouldn't need to return so soon.

So the health law instituted a Medicare payment penalty for hospitals with too many readmissions for pneumonia, heart failure, and heart attack. Since 2012, Medicare has assessed about $2 billion in penalties on hospitals with too-high readmissions rates.

Hospital groups have argued that these payments are punitive and unfair, particularly to so-called safety net hospitals that serve the poorest, sickest patients. These patients tend to have higher readmissions rates, and the hospitals that treat them were more likely to be hit with payment reductions. (Earlier this year, the Trump administration changed the penalty structure for safety net hospitals.)

But the program has often been labeled a success because it accomplished its primary goal. Readmissions dropped between 2.3 and 3.6 percentage points for the conditions targeted. Readmissions associated with other maladies dropped by 1.4 percent. The authors of one 2016 study suggested that the lower readmission rates "point to how Medicare can improve the care that patients receive through innovative payment models." It offered proof, and hope, that with the right incentives, Medicare could save money and provide better care.

A new study appears to dash that hope, at least as far as readmissions are concerned.

The study, published in the Journal of the American Medical Association (JAMA) and conducted by by researchers associated with Beth Israel Deaconess Medical and Harvard Medical School, looked at hospitalizations between 2005 and 2015. It found that "30-day post-discharge mortality"—the number of people who died within a month of leaving the hospital—increased for heart failure patients after the readmissions penalty program was implemented.

Although heart failure mortality was already on the rise, the rate of increase became more rapid after Medicare started penalizing readmissions. In addition, mortality rates amongst pneumonia patients, which had been stable, increased.

Fewer people were being readmitted to hospitals, but more people were dying.

This is not the first study to conclude that the program increased mortality. A separate JAMA study last year looked at about 115,000 Medicare patients and also found that although readmissions for heart failure were down, mortality had increased, with about 5,400 more people dying annually.