Using AI and big data to transform health care

The Big Data and Artificial Intelligence (AI) Innovation Hub at Sheba Medical Center in Ramat Gan, Israel, harnesses technologies such as the cloud and AI to improve patient outcomes, develop and deploy multinational projects with global impact, and create new sources of revenue.

The challenge

Sheba Medical Center has an ambitious innovation program called ARC (Accelerate, Redesign and Collaborate), a global ecosystem that now includes more than 100 members around the world, including other leading hospitals, industry and academic partners, and start-ups. Within ARC are six specialist hubs, one of which is the Big Data and AI Innovation Hub.

The Hub was set up to aggregate Sheba’s data into a data lake and deploy tools, such as AI and data analytics to improve patient care, develop new projects and build revenues. It needed to tackle a number of challenges: like many health care organizations, Sheba had highly siloed, dispersed data (held across more than 40 sources) and insufficient infrastructure to make sense of this information. This was hampering its analytics work and reducing productivity.

Robert Klempfner, Clinical and Scientific Director of the Hub, said the project involved a change of mindset for stakeholders: “If you’re dealing with AI, you learn to value data, and the curation of data, and the management of data, and the structuring of data – and similar to other hospitals, I think we found out that we were lacking in a lot of the infrastructure, physical but also conceptual. We were not organized enough.”

There were a range of issues to overcome. Regulation restricted the degree to which cloud technology could be used, making projects more costly and less flexible. Another big issue was interoperability and the need to get all of a patient’s data in one place so a clinician could access it through a single interface. “Doctors want everything that we can tell them about the patient in the same screen. They don’t want a pop-up, they don’t want to open another application, they want everything integrated,” said Sigal Sina, ARC Chief Data Scientist. 

The solution

To overcome barriers and ensure the right environment for the Hub’s activities, Sheba has worked hard to put the right enablers in place. It introduced commercial partnerships, allowing it to develop cutting-edge digital innovations without the backing of million-dollar budgets. Projects involve clinicians in the early design phase – in some cases, pairing them with start-ups or specialist analytics providers. They establish a mutually beneficial relationship in which Sheba shares the data and clinical expertise and the industry partner brings the technical skills. To further promote an entrepreneurial mindset, Sheba incentivizes staff innovations in big data and AI by handing out grants of up to US$50,000 each.

It has also overcome the regulation issue by lobbying for change. Klempfner explained: “We are working with the Ministry of Health and the justice department on domains such as patient confidentiality, on secondary use of medical data, and every couple of months or so, these regulations inch very slowly ahead. A couple of months ago, it wasn’t legal in health care to use cloud computing. Now, we have a cloud computing committee that can approve, in certain conditions, the use of public clouds, so things can change.”

The impact

Sheba’s ultimate goal is to improve the quality of care it delivers. The Big Data and AI Innovation Hub has allowed it to achieve this via analytics solutions, including predictive models. A recent example of this is its response to COVID-19. At the beginning of the pandemic, Sheba did not have a system to manage patients suffering from the virus and track their progress. It was able to combine multiple sources from the data lake to provide hospital managers with a dashboard that allowed them to see information for each patient and their condition, and track the number of tests performed, the number of deaths and other metrics.

It was also able to create an AI model to predict what the status of its patients would be in the next six hours, enabling better resource management and demand planning across departments. Finally, it could prioritize patients who were at risk of rapid deterioration and send those not at risk home to recover.