How the Utah-based medical group uses AI to identify medical wastage and make huge savings


“In the past, most health systems would save money by cutting out devices or procedures that cost the most,” observes Dr. David Skarda, Medical Director for Center for Value-Based Surgery at Intermountain Healthcare, a Utah-based, not-for-profit medical group with 22 hospitals and over 1600 physicians and advanced practice clinicians at 180 clinics. “By analyzing total medical costs over 120 days, we get a clearer picture of what gives us the best surgical outcomes, which also tends to lower the total cost of care.”

Dr. Skarda is referring to the surgical care process model he created to change the way the medical group analyzes and codes surgeries. The model leverages AI to evaluate data from electronic health records, claims and anything associated with the cost of care from 30 days before to 90 days after surgery. As of now, the tool was deployed for two procedures across Intermountain Healthcare and is projected to be saving more than $8 million for the first year.

Little wonder, last October, Dr. Skarda was named one of 2020’s top 25 Innovators by Modern Healthcare magazine.

Dr. Skarda’s interest in AI began with his involvement in using data to track opioid use and their potential for abuse in patients. He designed a series of surveys back in 2016 to find out how many doses of an opioid pain prescription they would take after their operations and disseminated them to 80,000 surgical patients at Intermountain Healthcare since January 2017.

Of the 30% who responded, Dr. Skarda found that on average, patients only take about 20-40% of the number of doses that are prescribed to them. More than 60% of the doses usually go unused and end up in their medicine cabinet. This increases the risks for misuse by others in the household, from diversion or theft. This confirms a critical risk for chronic opioid users. From then on, Dr. Skarda was drawn into the world of using data science to combat non-essential prescriptions, tests, and other medical wastage.

The surgical care process model was first used last May to help correct inefficiencies in the system before, during and after surgeries. Almost immediately, the model discovered that some pre-surgery tests were unnecessarily ordered. “Some of the patients had done a lot of blood tests,” said Dr. Raymond Price, Vice President of the Department of Surgery at Intermountain Healthcare. “We were not able to justify some of the tests identified by the AI. Eventually, we had to sit down as a surgical team to assess and reorganize our resources to minimize waste and value add the medical journey of pre-operative patients.”

The AI model was also used to analyze the costs of knee replacement, a common surgical procedure by accounting for medications, imaging, physical therapy, and possible complications over 120 days. “We found that price does not guarantee outcomes,” said Dr. Price. “If a device is slightly more expensive but results in fewer complications and rapid recovery, the AI will still favor it even if its initial cost is high.”

The AI model is now part of the medical group. All surgeons at Intermountain Healthcare will receive a report card that tells them where they can reduce costs and how other physicians in the same domain are improving outcomes. Dr. Skarda believes this will provide doctors with data to demonstrate what works and assists them in making better decisions. “I think doctors are more willing to change their practices, provided the correct information is there to prove what works best.”

On top of AI, Intermountain Healthcare also uses blockchain to support the management and storage of a large volume of data and securing underlying data. This allows the medical group to harness granular data ownership, data sovereignty, privacy, and data accessibility across various stakeholders ranging from patients, government agencies, insurers, to researchers and digital health solution providers in a secured manner. The medical group had saved more than $90 million in surgical costs alone over two years between 2017 and 2019.

“AI, machine learning and natural language processing facilitate data analyses to single out differences in care provided to patients,” Dr. Skarda added. “They connect each treatment or surgery to their corresponding outcomes so that doctors can make smarter decisions and prescribe more accurately to enhance care delivery. This, in turn, will bring down any avoidable spending.”