Mount Sinai Health System improves patient outcomes, engages providers, and scales for 3 TB of annual data growth using AI on Azure

With eight hospitals, globally recognized medical and nursing schools, and an extensive ambulatory care network, the Mount Sinai Health System (Mount Sinai) needed scalable tools that could maximize the impact its providers have on patients. It was among the first medical systems to apply data science to healthcare in a drive to reimagine the patient experience through the creation of a dedicated Clinical Data Science (CDS) team in 2015. In the years since, Mount Sinai has developed numerous cutting-edge projects that deploy machine learning and AI to reduce the risks associated with predictable events, such as malnutrition and falls, as well as with delirium and other conditions.

As Mount Sinai targeted ambitious projects, it realized that it needed to increase compute capacity. It migrated its apps to Microsoft Azure to accelerate the pace and scale of innovation. Now, Mount Sinai and Microsoft are in a strategic partnership to advance medical innovation and save lives. “Mount Sinai’s physicians, nurses, and scientists collaborate seamlessly to implement clinical decision support tools that are both accurate and helpful to our clinicians in their workflows,” says David L. Reich, MD, President of The Mount Sinai Hospital and Professor of Anesthesiology, Pathology and AI. “Creating ‘augmented intelligence’ at scale is our goal, and the partnership with Microsoft Azure has been a key element in our successes thus far.”

Scaling to support more than 3.7 million visitors and an increase of 3 TB of data each year

Mount Sinai’s providers receive more than 3.7 million visits annually. When the organization envisioned using big data to improve patient care in 2015, all compute capacity was hosted in on-premises datacenters. “Mount Sinai has developed a robust portfolio of AI products that are improving patient safety and bringing hospital operations to the next level,” says Robbie Freeman, Vice President of Digital Experience and Chief Nursing Informatics Officer of Mount Sinai. However, as it sought to make CDS tools available for use in an increasing number of clinical workflows, the organization recognized that using only on-premises computing resources would limit scale. Because the health system generates 3 terabytes of data growth each year, scalability is critical.

Along with its pioneering use of data science, Mount Sinai takes a collaborative and user-centric approach to development. Data scientists and engineers work closely with nurses, doctors, dietitians, and other providers to understand clinical problems, current workflows, and desired outcomes. Then they design easily adoptable workflows to help ensure that their providers benefit from innovation, transforming challenges into solutions. Mount Sinai also draws on its partnership with Microsoft to refine its products. “As chief nursing informatics officer, I sat down to have a peer-to-peer conversation with the chief nursing information officer at Microsoft,” says Freeman. “The fact that Microsoft has these leaders in house shows that it understands healthcare from our side of the table.”

Improving patient outcomes with 14 responsible AI solutions on Azure

One of the first projects that Mount Sinai delivered used predictive machine learning to identify patients who are at risk of falling in the hospital. The team relied on electronic medical record (EMR) data to build its model. After testing, the new solution significantly outperformed the existing fall-scale assessment. Not only is the new fall predictor improving patient outcomes, but it is also saving significant resources. “The average cost of treating a patient who falls in the hospital is $30,000, and we have more than 3,000 beds,” says Freeman. “Quality outcomes are often tied to a strong financial return on investment.”

Another especially successful use case was identifying malnourished patients. “If you can address underlying nutrition needs for patients, you move the needle on important quality metrics,” says Freeman. “Improved nutrition is linked to wound healing, preventing hospital readmission, reducing infection rates, and lowering the risk of death.” Previously, it was only possible to predict with about 20 percent accuracy which patients were at risk of developing malnutrition. Mount Sinai’s new predictive tool can indicate which patients a dietitian should visit with over 70 percent positive predictive value, greatly reducing the number of unnecessary visits. The health system is piloting another AI product, Nuance DAX, to provide voice-to-text transcription so that providers can focus more on patients and less on paperwork. Getting the right resource to the right patient at the right time has reduced mortality rates while boosting employee morale and engagement.

With 14 different AI products using Azure, Mount Sinai is tackling a remarkable array of use cases. Predicting the risk of delirium is a prime example. “Some patients can get really confused, so we built a tool to identify patients at risk of delirium,” says Freeman. “Based on the data, we can get the right resource to the bedside.” The information that Mount Sinai uses is multimodal, ranging from structured EMR data to written notes and images. The team uses AI to analyze this information and empower providers with lifesaving insights.

Another major area of focus has been the responsible use of AI. Mount Sinai has created tools to make sure that it applies AI responsibly to counter inequality in healthcare delivery. It built a tool to help ensure that when patient reps go out for rounds, they see a representative sample of the demographic composition of the hospital at large. “We have an extensive infrastructure put in place to ensure the safe and ethical use of AI,” says Freeman.

Building a foundation for sustained healthcare innovation

Mount Sinai is only getting started. The health system will use new AI tools to tackle industry-wide challenges in the years to come. “We don’t want to use our team’s time and resources for hardware maintenance,” says Freeman. “We’d much rather have our data scientists and engineers operating at the top of their capabilities.”

Mount Sinai hopes that others will follow in its footsteps. It regularly fields calls from peer organizations that want to do similar work. “We hope we can highlight ways that innovative teams like ours can help move the needle through collaboration with Microsoft,” says Freeman. “We’ve made so much progress in a short period of time.”