Algorithm spots patients’ potential breast cancer risk

Sanford Health develops AI to look for clues in provider notes on family history

Algorithm spots patients’ potential breast cancer risk

For even the most experienced health care providers, it’s never easy to tell someone they have cancer. For Sanford Health radiologist Chris Johansen, M.D., one situation is especially perplexing.

“Women would come in, sometimes for the first time, their very first mammogram, and they’d already have an advanced cancer. That, in particular, was deeply frustrating — young people that were following screening recommendations but simply hadn’t turned 40 yet,” he said.

Sanford Health recommends that women age 40 and above get an annual mammogram.

“And then we’d look at their medical record, and we would see that all the information had already been collected to indicate that this person was at high risk for breast cancer,” Dr. Johansen said. “A hardworking primary care provider and their staff had taken the time to ask the person about their medical history, about their family history and put all of these things into the record. We had all the information there.”

It was there but wasn’t displayed in a manner to flag high-risk patients.

Until Dr. Johansen, who practices at the Edith Sanford Breast Center, connected with a Sanford Health team that develops augmented intelligence tools to solve such problems. Cancer is the second leading cause of death in the U.S., and breast cancer is responsible for the highest number of new cancer cases, according to the Centers for Disease Control and Prevention.

“We decided that we should come up with an automated way to search the medical record, scan through it, looking for these indications of high risk,” Dr. Johansen said.

Series of filters

Dr. Johansen worked with Sanford Health advanced business intelligence analyst Austin Amdahl. He developed a system of filters, an algorithm, that sifts through doctor and patient notes in the medical records. It looks for keywords and displays up to seven risk indicators in a spreadsheet, based on National Comprehensive Cancer Network guidelines.

“(Dr. Johansen) had specific criteria he was looking for,” Amdahl said. “For instance, patients with known cancer susceptibility genes in their family history or family members that had early incidences of breast cancer or the number of those family members, how close they are to those family members. Whether it’s an aunt or a sister or something more distant like a great-grandparent or something like that.

“If the geneticist and the radiologist start going through charts it would take them forever,” he added.

Some relationships are obvious and require no digging. If a woman’s mother had breast cancer at a young age, there’s a good chance she might as well, for example. The augmented intelligence spots those not-so-obvious connections. A sister diagnosed at 40 is more of a concern than an aunt diagnosed at 60.

“The closer the relative and the younger the age of breast cancer onset, the more likely it is that that cancer was genetically underpinned and that genetic risk could have been passed on to you as that relative,” Dr. Johansen said. “If it was your mom, well you share 50% of your genetic material with her. As we get further out, you can still have an elevated risk, but the rules get more complicated.”

That’s where the AI excels.

“The algorithm can be programmed to take into account anything we ask it to,” Dr. Johansen said.

High-risk clinic

Besides developing the algorithm, the Sanford Health team also expanded its high-risk breast clinic to reach out to women flagged at elevated risk and then coordinate care with the patients.

It’s staffed by nurse manager Laurie Kruse, M.S., B.S.N., R.N.; nurse practitioner Colleen Schmidt, A.P.R.N.-C.N.P.; senior genetic counselor Larissa Risty, C.G.C.; breast health specialist Heather Harvey, B.S.N., R.N.; and clinic nurse Bev Keogh, L.P.N.

“Our hope is that we can get them into the clinic for a new patient visit and they can visit with myself and the genetic counselor,” Schmidt said. “If they are confirmed to be at elevated after a formal risk analysis, then we will develop an individualized breast imaging plan for them.”

Sanford Health Foundation supports the genetic counseling portion of the visit. If the patient is confirmed to be at elevated risk, most insurance companies will cover the cost of the breast screening exams.

“For the most part, patient screening exams are getting covered. And if not, we will assist the patient to file an appeal explaining why the insurance should pay for these exams,” Schmidt said. “Sometimes it is a fight. But we’re willing to do it because they need to get this important imaging.”

The clinic, in the Edith Sanford Breast Center in Sioux Falls, South Dakota, is also a resource for women not identified through the algorithm but who believe they may be at high risk. Other physicians who suspect a patient might have some of the factors also may refer women.

“If that risk is confirmed to be high, now Colleen is going to say, ‘OK let’s start looking at a personalized screening plan for you.’ Or she might talk about chemo prevention,” Kruse said. “We’re going to get them on the right path for breast cancer prevention and/or early detection.”

Work with physicians

In May, Sanford Health started using the algorithm and the workflow at the high-risk clinic. The early results from more than 50 patients are very promising, Kruse said.

Three physicians at the 26th Street and Sycamore Avenue clinic in Sioux Falls are part of a pilot program that helped further refine the workflow. All Sioux Falls region physicians have since agreed to allow the clinic staff to reach out to their patients in an effort to evaluate lifetime risk of breast cancer.

Women are alerted through their My Sanford Chart app, followed by a phone call from Harvey, the breast health specialist. From there the patient can decide whether she would like to schedule a visit at the high-risk clinic, where she will receive care from both a genetic counselor and Schmidt.

“The algorithm doesn’t actually tell you if a person is or is not high risk. It identifies factors that mean the person should be evaluated through a formal visit,” Dr. Johansen said. “The providers will be the ones that determine the patient’s actual risk.”

The approach ensures primary care physicians are involved from the start.

“You don’t want to tell someone you may be at high risk for breast cancer and do that to a whole bunch of people and all of a sudden, our primary care docs are just being inundated. So we wanted to make sure we have a high-risk clinic that is also ready to take these people and partner with primary care docs to make sure the workload stayed reasonable,” Dr. Johansen said.

In response to the pandemic, the high-risk clinic also provides telemedicine visits, so women can do video appointments.

“At Sanford Health, one of our core values is we want to bring high-quality, specialized medical care to rural areas,” Dr. Johansen said. “Telemedicine allows us to bring an evaluation by a breast cancer risk specialist to all our patients without them having to physically come here to Sioux Falls.”

Learn more

Posted In Cancer, Cancer Screenings, Foundation, Imaging, Innovations, Rural Health, Sioux Falls, Virtual Care, Women's