The University of Pittsburgh Medical Center (UPMC) describes itself as the second largest integrated payor-provider network after Kaiser Permanente. At the HIMSS Big Data and Healthcare Analytics Forum on Thursday, UPMC Chief Analytics Officer Pamela Peele talked about how being both a payer and a provider enables UPMC to address readmissions in a unique way.

"Providers are trained to manage disease, and insurers are trained to manage financial risk," she said. "[Asking providers to manage risk] is like asking me to put a stent in you. We’re asking providers who are not trained to manage financial risk to manage financial risk, which is something insurers do extraordinarily well, which is another reason putting a payer and a provider together is so powerful."

UPMC was able to reduce its readmissions from 16.5 percent down to 13 percent between 2008 and 2015 by using big data from two different sources -- one from the insurance side and one from the hospital side.

The first round of modeling the hospital did used only the claims data. Using this data set, the hospital system attempted to pinpoint the patients most likely to be readmitted. They managed to cut the chance of any given readmission from a 28 percent chance to an 11 percent chance by mandating that physicians follow up with patients within five days.

But the real breakthrough came when the hospital side of the business implemented its own data modeling strategy for reducing readmissions, based on data from the EHR like hemoglobin and sodium levels. Peele was surprised to see how claims data and EHR data could come to the same conclusions, starting from such different places.

"We’re predicting readmission before they’re ever admitted and this is predicting the readmission risk based on drastically different inputs while they’re in the hospital, yet the two models perform almost identically," she said.

Almost identically, but not exactly the same: there were a small number of patients that the claims data model said were at high risk for readmission but the hospital model pegged as low. Reconciling those differences, Peele said, was a matter of figuring out whether specificity (or the avoidance of false positives) or sensitivity (the avoidance of false negatives) was most important. And for managing resources and saving money, they prioritized the former.

"We don’t want to identify people as having high readmission risks when they don’t because we don’t want to be dropping those resources," she said. "So we have the hospital model, which has pretty good specificity, and the claims model, which doesn’t perform quite as well as the hospital model, but the hybrid model beats them both."

The combination of payer and provider data helped UPMC to reduce readmissions. During the Q&A, Peele also shared that being both the payer and provider allows the health system to spend money in care-adjacent ways that normal payers wouldn't necessarily reimburse for.

"A good example is in our psychiatric setup, when someone’s been in a psychiatric hospital for months, they’re sick," she said. "Medicare won’t pay for a transition team to take them home. Guess what happens when they’re home after a couple of months in the hospital? The cat’s dead, the refrigerator’s moldy, and you have an eviction notice. Who’s going to bury the cat, clean out the refrigerator, and reconcile with the landlord?"

Those crises can trigger another mental breakdown, leading to a readmission. So that kind of creative spending is another way integrated systems can save money and help patients in ways that traditional hospital systems can't, Peele said.