The Medical University of South Carolina embraces Microsoft tools and technologies to transform patient care

Building a modern-day ecosystem through Azure

In 1992, MUSC launched an extensive initiative to collect data from its healthcare, education, and research arms and store it in a clinical data repository. According to Mark Daniels, MUSC’s Chief Technology Officer, “We had lots of datasets that are like gold. Unfortunately, they were very disparate. They were not linked or mapped together.” MUSC worked with Microsoft to restructure its enterprise data warehouse onto the Microsoft Azure cloud. “In collaboration with Microsoft, we've been able to bring many of our datasets together, creating what we call a modern-day ecosystem that we’ve been able to leverage across the board.” 

To move the system’s decades of patient records and research data from the disparate storage locations into one accessible knowledge repository, MUSC's technical team chose to use Azure Databricks, a software as a service (SaaS) application that runs on Azure. The cloud-based software stores, cleans, and sorts huge volumes of data so it can be analyzed and used in other applications. 

Transforming patient care

The digital transformation at MUSC provides clinicians with more raw data, in addition to more knowledge and actionable information. As the wealth of data grows, the tools used to gather the data will continue to improve and become increasingly valuable to clinicians.  

One significant area of opportunity is in diagnostic information for identifying symptoms and characteristics. According to Dr. Danielle Scheurer, Chief Quality Officer at MUSC Health, technology “helps our clinicians make smarter and more efficient decisions at the bedside. It helps with diagnostics, it helps with matching diagnostics with therapeutics, and it helps us predict things that we otherwise would be guessing. So, for example, we’ve used technology enablers and artificial intelligence to try to figure out which patients are at risk for getting readmitted to the hospital. We can actually very precisely predict that risk and then layer on evidence-based interventions to try to reduce that risk so that it doesn’t happen again.”

MUSC has applied AI functionality to medical record data in the detection of sepsis, a dangerous, full-body infection that can be fatal. “Sepsis can be very difficult to identify in its early stages, and the early stages are when the clinical interventions work best,” Dr. Scheurer explains. “Within our EMR [electronic medical records], we have systems built so that the bedside nurse gets an alert when a patient is meeting or triggering those criteria for early sepsis.” 

Once that system is triggered, a cascade of communication and interventions begins, making sure that the patient gets early care, which has been shown to improve sepsis outcomes. “It builds in what we call ‘decision support’ and triggers the exact clinician who can act on behalf of that patient,” Dr. Scheurer says.

An AI model is also used to proactively assess a patient’s risk of readmission, notifying clinicians while the patient is still hospitalized. Analytics of population-level data can help identify which interventions are most likely to help an individual patient. According to Daniels, “AI analytics help clinicians make diagnoses faster, identify problems earlier, and predict outcomes more accurately. Administrators identify cost-saving operational efficiencies. And students are taking advantage of the modern technology they expect, both in the classroom and for remote learning.”

Using Azure Machine Learning, a cross-platform client that enables AI-empowered data applications, Daniels and his team have found innovative ways to apply AI analytics to their data stores. The data team developed the pipelines that allowed MUSC to pull EMR data into the data lake within days. This allows AI models to be updated frequently and to adapt and adjust constantly as new data is acquired from patients in the clinical setting or through primary research.

“Azure Machine Learning allows us to quickly create and validate models, but it also allows us to use those models with our EMR and interject them into the clinical workflow. This means clinicians are seeing the byproduct of these models at the right time—while they're treating the patients,” Daniels says. 

“Users today for the most part don’t want you to hand them a report," he adds. "They want you to hand them access to the data. And this is both in the healthcare setting as well as in the university and research setting. I really think it’s a shift in the way our users want to deal with the data that they need. They no longer want it to be handed to them through someone else’s eyes, they want to actually have access to the raw data as well as to all the metrics that the hospital and health system have approved for them to use in all of their measurements.”

However, MUSC expects data-related challenges as healthcare moves into the future. According to Dr. Cawley, “The hardest thing about digital transformation is getting leaders and even people at the frontline to understand how, with these new tools, the constraints they were under before no longer exist. We need to stop thinking about the old way of doing things and get rid of some of those old constraints, which is exactly what digital transformation does. It’s a mindset change.”

Another issue, in Daniels' words, is that, “We also have some disparity in our connectivity and device availability, especially in some rural areas of South Carolina. We saw this with the setting up of some of our COVID testing. We have to transform in a way that we’re not making it so that certain people cannot participate in where we’re taking this.”

Using AI to improve COVID-19 screening

Early in the COVID-19 health crisis, MUSC developed an online service to remotely screen patients with possible coronavirus symptoms to minimize the risks associated with in-person interactions. 

With Azure Databricks, MUSC data scientists trained an AI model using natural language processing and applied it to patient data collected from telehealth visits. The model flagged the strong correlation between positive COVID tests and patients who had lost their sense of smell and taste. These symptoms were integrated into the MUSC screening algorithm before they were officially recognized as symptoms by the Centers for Disease Control and Prevention (CDC). The model was trained in less than two weeks. Out of hundreds of callers a day, providers were able to identify a manageable number of several dozen daily cases that required follow-up calls.

Building the future of healthcare

In classrooms, MUSC is continuing to explore the use of new technology for better ways to teach medical concepts, including 3D printing, mixed reality, and virtual reality. The remote teaching infrastructure put in place during the health crisis will likely continue to shape and influence instruction.

MUSC intends to continue expanding its capabilities through digital transformation. Along with providing clinicians with more raw data, guidance, and actionable information, new tools and technology will help the medical system meet and anticipate the needs of its population. “Patients really want to embrace this digital transformation. They want to see us doing it,” Daniels says. Looking ahead, MUSC is continuing to build out genomic stores and AI predictive modeling, and is looking for new ways to apply IoT interfaces for smarter medical devices and improved patient care—all advances built using the power of the Azure cloud.

“Digital transformation is something that we can't achieve in small steps,” Daniels says. “We have to be more disruptive. We have to take a step back and look at new ways to approach our problems.”