AI helps Northern California health providers to improve care, lower costs

Alexa and Siri, Amazon and Apple’s respective digital voice assistants, are everyday examples of artificial intelligence at work. But the technology is all around us.

With artificial intelligence, computers and machines are programmed to have the ability to solve problems and make rational decisions, augmenting human intelligence.

In a health care setting, AI is improving diagnoses and patient care delivery for everything from surgery scheduling and spotting cancer in its earliest stages, to alerting caregivers if a patient suddenly takes a turn for the worse.

AI also is being developed for patients so they can use an app on their smartphones to put in their symptoms and find out what may be wrong. Digital health company Babylon is doing this kind of work. In April, it acquired Novato-based Meritage Medical Network with its 700 North Bay physicians who provide care for nearly a half-million residents.

Help yourself

Along those lines, Sutter Health uses an AI-powered tool called Ada that lets people type in their symptoms on Sutter’s website or a smartphone to, again, learn what may be ailing them. Ada identifies a potential diagnosis, along with recommended next steps, according to Dr. Albert Chan, vice president and chief of digital patient experience at Sutter Health.

Sutter two years ago partnered with Ada Health, a Berlin-based global company that designed the application. Ada is not meant to replace physician care, but rather enhance it, Chan noted.

“One of the subtle things Ada does is it queues up the questions that doctors may have so when you ultimately see the doctor, you’ve thought about it,” said Chan, who also practices family medicine at the Palo Alto Medical Foundation. “In time, we're looking to actually integrate those reports so they come in directly to the doctor.”

From the inside

AI in the medical arena is mostly being deployed for internal purposes.

One of the primary uses of AI at NorthBay Healthcare is for medical transcription, said Chris Timbers, vice president and chief information officer at the Fairfield-based health care system.

“Compared to the old method, it's dramatically and statistically provable how much more accurate it is,” Timbers said. “The other advantage is it's very timely. With the old-fashioned dictation and transcription, there was usually a 24-hour turnaround. With this, the notes are almost instantaneous.”

The technology, which NorthBay adopted about five years ago, has the ability to learn and recognize voice patterns, as well as medical vocabulary. That efficiency has a dual purpose: it eliminates the cost of an outside medical transcription service and the potential for human error, which could lead to further problems down the road.

“I know that we were able to cut our dictation and transcription costs in half,” Timbers said.

On schedule

NorthBay also adopted Houston-based Leap Rail’s AI technology that can accurately determine how long a surgery would take, potentially freeing up operating rooms sooner to make room for more procedures to be done on the same day, Timbers said.

Leap Rail’s technology does a deep dive through the patient’s electronic medical record to look for any comorbidities a patient may potentially have that aren’t obvious. That can make a difference in how much time a surgeon may need for a given procedure.

“It’s applying a statistical method to really large data sets that are just a little bit more accurate than even some of our most educated human beings,” Timbers said. “So, we've really been able to improve the accuracy of booking and scheduling our operating rooms, which is one of our most valuable resources.”

There's also a human value.

“We're able to give the family members a much better estimate of how long the surgery will take,” he said. “You can imagine if we tell them the surgery should take 90 minutes and it goes two-and-a-half hours, that creates a lot of anxiety.”

Researching AI

Kaiser Permanente, which operates four hospitals and numerous outpatient facilities in the North Bay, recently published a study that showed it’s possible to teach computers — using AI technology — to read written notes in medical records. The study focused on patients with aortic stenosis, a common but serious heart valve disease.

The study’s findings showed the computer was able to read through nearly a million electronic health records and echocardiograms, and identify 54,000 patients with aortic stenosis, according to Dr. Matthew Solomon, the study’s lead author.

“It would have taken years to go through the number of charts that we went through in minutes with the computer,” said Solomon, who is an adjunct researcher and cardiologist at The Permanente Medical Group.

Solomon said the authors’ research didn’t involve calculating a potential return on investment.

“It certainly is cost saving and a more efficient way to do the types of programs we're aiming to do,” he said. “I think we’re entering a brave new world where we're going to be able to do population management in a way that's never been done before.”

Kaiser's research team next plans to use the technology to identify other types of heart conditions. If successful, the next step would be to teach the computer how to analyze patterns in electronic medical records that could identify patients at risk for aortic stenosis, according to Kaiser.

On high alert

Back at NorthBay, the health care system is using another AI platform called the Rothman Index, which can quickly identify if a patient’s health condition suddenly worsens and requires a swift intervention, Timbers said.

The Rothman Index is an early-warning system, which is integrated into the electronic medical record, according to the company’s website. It continuously tracks more than two-dozen variables.

“These are usually pretty sick patients, and it's able to accurately identify when that patient may be heading toward becoming sicker and maybe having to go up to the ICU,” Timbers said. “It’s not replacing the clinician’s judgment, but it's looking at a lot of additional factors.”

The Rothman Index also can identify patients who are ready to leave the ICU, as well as those who aren’t improving and may instead benefit from palliative care, according to the company.

The cost for AI platforms like the Rothman Index and Leap Rail can range from $50,000 to $400,000 a year, depending on the size of the health care system, Timbers said.

The ROI, however, is hard to determine.

“When it (comes to) improving outcomes and quality, we believe very strongly that there's a cost savings associated with that, but it's a little bit difficult to draw a straight line to it,” Timbers said.

Finding cancer early

A personal experience led Chan to connect Sutter Health with Ferrum Health, a San Francisco-based developer of an AI-based patient-safety platform to help health systems prevent avoidable errors. Sutter in February 2020 partnered with Ferrum to use its AI technology to support radiologists in finding cancer at earlier stages.

“I helped us engage with Ferrum because a family member had a lung cancer that was missed,” Chan said, emphasizing he doesn’t blame anyone. “When I see a problem like this, I think it's not the individual people, it's the supporting systems that need to be improved and made better to avoid that error that we all could make.”

The Ferrum monitoring platform runs in the background and alerts doctors to potential missed diagnoses. If there is a potential discrepancy, such as a nodule found in a scan with no mention in the report, the scan is flagged and sent to a radiologist subject-matter expert for further review, according to Sutter.

Since its implementation at Sutter’s North Sacramento Valley facility, Ferrum’s AI technology has reviewed approximately 60,000 CT scans of patients containing lung tissue. About 200 of those scans were flagged and required further radiologist review and intervention. Some of those patients were confirmed to have findings that needed follow-up care, according to Sutter.

“There is a radiology professor I follow who says, ‘I don't think AI will replace doctors, but doctors who use AI will replace doctors that don't,” Chan said. “I think that was very, very wise.”

Cheryl Sarfaty covers tourism, hospitality, health care and education. She previously worked for a Gannett daily newspaper in New Jersey and NJBIZ, the state’s business journal. Cheryl has freelanced for business journals in Sacramento, Silicon Valley, San Francisco and Lehigh Valley, Pennsylvania. She has a bachelor’s degree in journalism from California State University, Northridge. Reach her at cheryl.sarfaty@busjrnl.com or 707-521-4259.

Show Comment