The Medicare policy of financially penalizing hospitals for high rates of readmission within 30 days of discharge for heart attacks, heart failure and pneumonia is meant to encourage better inpatient care and hold hospitals accountable for how they treat certain common health conditions. It also has given rise to a new class of health IT tools.

But one prominent health policy researcher believes that notion is only half correct.

"Readmissions are not a good quality measure -- but they may be a very good way to change the notion of accountability within the healthcare delivery system," Dr. Ashish Jha, a professor of health policy and management at the Harvard School of Public Health, wrote last week in an opinion piece on The Health Care Blog.

Jha, who describes himself on his Twitter page as an "advocate for the notion that an ounce of data is worth a thousand pounds of opinion," said that the Centers for Medicare and Medicaid Services (CMS) Hospital Readmission Reduction Program has prompted hospitals to ask whether a 30-day readmission rate is a good measure of care quality. "Over the last three years, the evidence has come in, and to my read, it is unequivocal. By most standards, the readmissions metric fails as a quality measure," Jha wrote.

[ Want to know what analytics vendors are doing to improve healthcare quality? See 7 Big Data Solutions Try To Reshape Healthcare. ]

He noted that the job of a hospital is to offer the right treatments for a given ailment, attend to each patient's needs and prevent adverse events such as in-hospital infections. Quality can be quantified through structural measures such as whether there are enough staffers, process measures like whether patients get aspirin following heart attacks, or by looking at outcomes, according to Jha. Outcomes are particularly difficult to measure.

"Death is clearly a bad outcome (with the caveat that for someone with a terminal illness, death may be an expected outcome). Nosocomial infections are bad outcomes. Is being readmitted a bad outcome?" Jha wondered.

"If we had a sick patient in the ER who was last discharged from the hospital 28 days prior, we don't make them better by sending them home instead of admitting them. Yes, sending them home avoids a readmission -- but the goal is not to avoid readmission, the goal is to make people better," the Harvard health policy specialist explained.

Jha also referenced a paper he co-authored in the January 23-30 issue of the Journal of the American Medical Association, in which readmission rates did not correlate with hospital mortality rates, a key benchmark of outcomes.

"Readmissions seem to have little external validity as a quality measure. Readmissions are, however, correlated with two things: how sick your patients are, and how poor your patients are. We now have good data that the Hospital Readmission Reduction Program disproportionately penalizes big academic teaching hospitals (that care for the sickest patients) and safety-net hospitals (that care for the poorest)," he said.

While Jha is unconvinced that readmission rates are indicators of quality, he admitted changing his previous belief that the metric does not promote accountability. "In conversations with colleagues and friends, the readmissions penalty program seems to have gotten some hospitals to think outside of their four walls," Jha wrote. "Hospital leadership has started to rethink the role of the hospital. Hospitals are building relationships with community-based organizations. Some are creating follow-up clinics while others are calling all the patients who are discharged to make sure they are doing OK at home."

But this is only happening where providers have a financial incentive to keep patients healthy, not to render as many services as possible. "No one gets rich keeping patients healthy and well (and therefore, not needing to be hospitalized). The Accountable Care Organization (ACO) program begins to do that, and I am hopeful it will make a difference. But the truth is that most patients are not in an ACO and won't be anytime soon," Jha said.

As large healthcare providers test the limits, many smaller groups question the value. Also in the new, all-digital Big Data Analytics issue of InformationWeek Healthcare: Ask these six questions about natural language processing before you buy. (Free with registration.)