How AI Makes Health Care Better at Northwestern Medicine

Advances in augmented intelligence (AI)* and machine learning (ML) are happening at an accelerated pace, and their adoption in health care is widening. At Northwestern Medicine, teams work to make sure patients and ethics are always top of mind, no matter the innovation.

“We have teams that carefully evaluate challenges and consider whether AI is the right solution for them,” explains Shruti Cruz, director, Information Services, who helps oversee digital solutions at Northwestern Medicine. “Sometimes, there are easier and lower-tech ways to solve problems.”

For example, Cruz says, when care teams at Northwestern Medicine Bluhm Cardiovascular Institute discovered that patients were being referred to heart failure specialists too late, they turned to an AI solution. Teams used existing historical patient data to train an AI model to recognize signs of advanced heart failure. The model is now used to monitor current patient data and identify patients with advanced heart failure before they progress too far or once they are too ill to qualify for life-saving therapies, leading to more timely evaluation and treatment.

To build a solution, whether it be clinical or nonclinical, teams go through “stages” of AI development, explains Danny Sama, vice president, Information Services, and Chief Digital Executive at Northwestern Medicine. These stages include data, models and workflows.

To build and implement the advanced heart failure AI model:

• The team built a robust data bank of demographics, lab results, medications, and billing information for patients with heart failure.
• The team then trained a model with that information to make predictions about whether a patient might have advanced heart failure.
• The model’s output is embedded into the Northwestern Medicine electronic medical record workflow, where patient charts are flagged for nursing coordinators to cross-check and note if they agree or disagree with the model.

“The model is not making any clinical decisions,” explains Cruz. “It helps us identify patients earlier and improve the clinical team’s ability to coordinate appropriate care.”

Northwestern Medicine teams working on AI solutions always incorporate social factors that may influence a patient’s health. The technology can also help reduce health disparities by factoring in additional risk factors that may affect a patient.

Teams also incorporate the perspectives and feedback of expert clinicians throughout the development and training of models, not just at the end for a final review or approval.

“We look to keep the research world and the healthcare operations world in sync while we develop new models here at Northwestern Medicine,” says Sama. “We have an integrated perspective, stitching data together that helps care professionals identify patients at risk and prevent them from falling through the cracks.”

In health care, Sama says innovation might move more slowly than in other industries. However, he says that the endless opportunities to improve the field of health care will continue to inform how Northwestern Medicine uses AI to drive operational efficiency as well as improve patients’ lives.

“We analyze what we can and can’t do, and what we should and shouldn’t do with this technology,” Sama explains. “Really, this is not about replacing care teams. It’s about helping them provide better care for patients.”

*Considered a subset of artificial intelligence, augmented intelligence uses information to support human intelligence rather than replace it.