University of Pittsburgh Medical Center (UPMC) is committing to a five-year agreement with Microsoft that will modernize and expand this infrastructure, all with the goal of continuing to improve patient care.
Particularly during the COVID-19 pandemic, UPMC has proven the importance of its clinical analytics effort, started in 2012. Backed by more than 30 years’ worth of clinical and financial data, UPMC’s clinicians and researchers were able to quickly assess and adjust COVID-19 therapies, resulting in a measurable decrease in in-hospital mortality month after month since early 2020.
“We’re on a quest to become a true data-driven organization, a ‘learning health system,’” said Dr. Oscar Marroquin, chief healthcare data and analytics officer. “We can do this only if analytics are embedded in everything that we do – from the executive suite to our clinicians at the bedside.”
To continue its progress toward this vision, UPMC is partnering with Microsoft to provide its latest cloud computing, artificial intelligence and machine-learning tools to the clinical analytics team. Working with UPMC clinicians, these experts are helping to mine more than 13 petabytes of structured clinical data and 18 petabytes of imaging data to create new insights that improve patient care.
“This agreement with Microsoft will allow Oscar and his team to do their work in a faster, more scalable and more sustainable way in the years to come, transforming care for patients,” said Chris Carmody, chief technology officer at UPMC. “By modernizing our analytics platform, we’ll also be better able to support expansion and innovation in our Health Plan and other divisions,” he added.
“Organizations like UPMC, which have a long and successful track record of using analytics to measurably improve patients’ lives, are at the forefront of the digital revolution in healthcare,” said Patty Obermaier, vice president, U.S. Health and Life Sciences at Microsoft. “By using the Microsoft Cloud to analyze vast amounts of data, UPMC routinely develops better insights, better experience and ultimately better care.”
Beyond its importance in COVID-19 care, UPMC’s clinical analytics effort has produced other notable successes for the health system and its patients, and it has done so by being embedded within the clinical operations of UPMC. Using hourly data feeds, the analytics team supports activity at most of the health system’s 40 hospitals, and this effort is the backbone of various quality, finance, innovation and other improvement efforts.
Explained Marroquin: “Our clinical analytics team enables, with data, the work of all our service lines, service centers and departments. They partner in a ‘hand and glove’ relationship that starts with data-driven identification of opportunities for continuous improvement of our clinical programs, regardless of where the care is delivered.”
For example, diabetes mellitus is a common and complex medical condition that is associated with increased risk for other conditions and adverse outcomes, particularly in those who have difficulty controlling their disease. UPMC’s endocrinologists wanted to enhance the care that physicians can provide by adding diabetes educators to the team. But to effectively use those limited resources, the endocrinologists asked the analytics team to help identify those at the highest risk of having uncontrolled diabetes before it happens, since those patients are most likely to benefit from diabetes education services.
The UPMC analytics team used historical data from more than 170,000 diabetic patients to build a machine-learning model that allows them to predict who is most at-risk with high levels of accuracy. After the model was developed, trained and prospectively validated, it was embedded within the workflow of UPMC’s endocrinologists. “While still early in the process, the use of this model has produced significant improvements in diabetes control for those who engage in the program as compared to those who do not,” said Marroquin.
Similarly, UPMC’s perioperative service line worked with the analytics team to develop a predictive model identifying patients who are at high risk of having adverse outcomes following surgical procedures. This is important because identification of these patients well before surgery allows clinicians to help them improve their health status through measures like smoking cessation, weight loss, better management of chronic diseases and other interventions offered through the Center for Pre-Operative Care.
The machine-learning model was built by looking at more than 1 million surgeries that have occurred at UPMC to identify patients most likely to have bad outcomes within 30 days after surgery. Now, when any surgical procedure is scheduled at UPMC, the model runs automatically, and clinicians use the information to thoroughly evaluate and manage high-risk patients before surgery to give them the best chance for a good post-operative outcome. “Our perioperative service line has seen significant improvement in outcomes since this approach has been embedded within clinical operations,” added Marroquin.
Successes like these have created “not only the desire, but also the expectation, that our analytics program will continue to advance by acquiring more data, to do it faster and to distribute it more seamlessly to our users,” said Carmody. “By partnering with Microsoft, we’ll now have the ability to modernize our processes and technologies to meet those expectations.”