UPMC Chief Data Officer Highlights Integrated Health System's Role in Reducing Hospital Readmissions

Pamela

The University of Pittsburgh Medical Center (UPMC) describes itself as the second-largest integrated provider network, following Kaiser Permanente. At the Healthcare Information and Management Systems Society (HIMSS) Data and Analytics Forum, Pamela, UPMC’s Chief Data Officer, discussed how UPMC’s dual role as both payer and provider enables it to address readmission rates in a unique manner.

Pamela said, “Healthcare providers are responsible for treating diseases, while insurers need to manage financial risks. Asking providers to manage risk is like asking a person to place a stent inside their own body. It is unscientific to expect providers, who have not been trained for this, to manage financial risks; insurers can handle these tasks very effectively. This is precisely why integrating payers and providers is so powerful.”

According to data from insurance companies and hospitals, the University of Pittsburgh Medical Center reduced its readmission rate from 16.5% to 13% between 2008 and 2015.

In the first round of modeling, hospitals relied solely on claims data. Using this dataset, hospital systems aimed to identify patients at the highest risk of readmission. They sought to reduce readmission rates by having physicians conduct follow-up visits with patients within five days.

However, the true breakthrough came when UPMC, in its hospital operations, leveraged data modeling strategies based on electronic health record (EHR) data—such as hemoglobin and sodium levels—to reduce readmission rates. Pamela was surprised to find that claims data and EHR data, though originating from different domains, yielded the same conclusion.

“Our method for predicting readmission rates is unacceptable because our risk prediction for readmissions is based on markedly different inputs during hospitalization, yet the performance of the two is nearly identical,” she said.

Almost identical, but not entirely so: the claims data model indicated that a small number of patients were at high risk of readmission, whereas the hospital model classified them as low risk. Pamela believes that reconciling these discrepancies is the most critical factor in determining whether to prioritize specificity (to avoid false positives) or sensitivity (to avoid false negatives). To conserve resources and funds, they prioritized the former.

Pamela stated, “We do not wish to misclassify individuals without a high risk of readmission as being at high risk, as we aim to avoid the inefficient allocation of resources. Therefore, we require both a hospital-based model with high specificity and a claims-based model that may underperform compared to the hospital model; a hybrid approach combining the two can outperform either model alone.”

The integration of payer and provider data has helped UPMC reduce readmission rates. Pamela believes that health systems can allocate funds toward innovative spending.

She said, “There is a prime example in our psychiatric department: when a person stays in a psychiatric hospital for several months, they inevitably face complications. Health insurance does not cover their transitional period. What happens when they return home after months of hospitalization? The cat has died, the food in the refrigerator has molded, and they may even receive an eviction notice. Who will bury the cat, clean out the refrigerator, and reconcile with the landlord?”

“These crises can trigger another mental breakdown, leading to patient readmission. Therefore, this creative spending is another way for integrated systems to save money and help patients—a feat that traditional hospital systems may not be able to achieve,” said Pamela.

Translation | Zhang Xiaohao

Editor: Huang Jia