10 ways North Carolina health care providers are harnessing AI

By Emily Vespa and Michelle Crouch

Co-published with The Charlotte Ledger

Faced with a growing health care worker shortage and provider burnout, some North Carolina health care leaders have said there is great promise in artificial intelligence systems. 

The state’s health care systems have been early adopters of AI tools — overt and invisible — that are shaping the patient and provider experience. Across North Carolina, AI is helping providers predict health risks, communicate with patients and manage administrative tasks. 

Here are 10 ways North Carolina health care providers are using AI.

1. AI helps doctors diagnose lung cancer

Some Atrium Health doctors are using AI to help catch lung cancer in its early stages. 

When pulmonologists find a lung nodule on a scan, they consider factors such as the patient’s medical history to assess lung cancer risk. Physicians might recommend a lung biopsy to a high-risk patient, but they avoid an unnecessary procedure if a patient is low risk. 

At Atrium Health Wake Forest Baptist, doctors are using an AI tool to help determine the likelihood that a nodule is cancerous. The tool, called the Virtual Nodule Clinic, scores a nodule selected on a scan from 1 to 10, with 10 being the highest cancer risk. 

Travis Dotson, a pulmonologist at Wake Forest Baptist, said the virtual clinic has helped him and his patients decide next steps when lung nodules seem to have an intermediate cancer risk. 

“It’s another level of support that gets added to the clinician’s feelings and their synopsis of how they feel about the nodule,” Dotson said. “The right thing to do is to just be conservative, which you can imagine could be pretty hard for a patient if they’re very concerned and there’s the uncertainty about what this nodule is.”

Dotson said recently, the tool assigned a high score to a nodule that other calculations said was a fairly low cancer risk. The score helped his patient decide to have a biopsy, and the nodule ended up to indeed be malignant. 

The tool doesn’t replace a clinician’s judgement, but Dotson said he has found that patients are very receptive to AI use in augmenting clinical decision making. Wake Forest Baptist was the first academic medical center in the country to start using the technology in March 2023, it said in a news release.

2. AI checks on surgical patients

At OrthoCarolina in Charlotte, an AI-based digital assistant called Medical Brain follows up with patients to see how they’re recovering after a hip or knee replacement.

The smartphone app “interacts with patients, asking questions about their recovery and giving them information about what they can expect,” according to Brian Krenzel, a hip and knee surgeon and OrthoCarolina’s chief quality and development officer.

Patients can also ask questions and get an instant answer, he said.

In the four months since OrthoCarolina began piloting the technology, it has interacted with about 200 patients, Krenzel said, averaging 30 to 60 messages per patient.  

A medical team manually reviews every interaction, he said, and so far, the system has made no mistakes. The tool relies on responses that OrthoCarolina has pre-written and loaded into their system. If it doesn’t know an answer, it directs patients to call the practice triage line.

Krenzel said Medical Brain has reduced the volume of traditional messages and phone calls coming into the clinic after surgery by about 70 percent. OrthoCarolina is rolling out the technology to all of its hip and knee patients and is adapting it to work for spine surgery patients as well. 

3. AI spots danger on patient X-rays and scans

Across the state, AI is already changing the game when it comes to medical screening by delivering fast and possibly more accurate results that can help doctors focus on the most urgent cases.

For example, in Novant Health emergency rooms, AI scans patient images to spot serious conditions, such as a broken neck, brain bleed or blood clot, ensuring that those patients are treated first, according to a Novant spokeswoman Caroline Arey.

The technology serves as a “second set of eyes for radiologists,” Arey wrote in an email. She said it helps them identify the most acute cases first and eases the burnout “often experienced when working against the clock to ensure fast diagnosis and treatment.” 

Another AI platform called Viz.ai specifically analyzes the CT scans of suspected stroke patients, Arey said, looking for a particularly dangerous type of blockage in the brain’s arteries. 

Within seconds, the tool sends results to stroke specialists’ smartphones. That’s important because for every minute a stroke goes untreated, millions of brain cells die, increasing the risk of permanent brain damage.

4. AI may be drafting messages you get from your doctor

Like many health systems across the country, Atrium Health and WakeMed have turned to AI to help manage the flood of messages their clinical teams receive through patient portals.

Since the pandemic, the number of messages to doctors has skyrocketed, adding to clinicians’ already packed workloads and increasing stress.

WakeMed has reduced patient portal messages by 12 to 15 per provider per day by using AI to draft responses, filtering out unnecessary messages and forwarding some messages to other staff members, according to an article in Becker’s Hospital Review.

At Atrium, AI drafts “initial responses to patient messages that the team then edits before sending,” a spokesman said in an email.  

One study found that AI-generated messages outperformed those written by humans in understandability and tone, and were more than twice as likely to be considered empathetic. However, the same study found that the AI responses were also 38 percent longer and more likely to use complex language.

Critics worry that too much reliance on AI for communicating with patients might lead to missed nuances in patient concerns or that it could disrupt the doctor-patient relationship.

5. AI helps clinicians decide who will benefit from drugs or treatments

At Wake Forest University School of Medicine, developers created an AI-based screening tool to identify patients who may have cognitive impairment. Called the electronic Cognitive Health Index, it helps flag patients who may benefit from a program that provides specialized care and medications for those at risk of developing Alzheimer’s Disease. 

6. AI makes sure you don’t miss your follow-up

North Carolina health systems are using AI to identify patients overdue for follow-ups or regular procedures. Duke Health Chief Health Information Officer Eric Poon said the system uses AI to spot patients who are overdue for a mammogram.

At Wake Forest Baptist, the Virtual Nodule Clinic flags patients that missed their prescribed follow-up lung scans. Dodson said those scans are “one of the most important aspects of lung nodule management.” 

7. AI spots patients with sepsis symptoms

Poon said Duke Health has also collected enough patient care data through research and daily operations that it uses AI to evaluate the data and inform future clinical situations.

For example, Duke Health emergency rooms input data to an algorithm that is trained to detect patients at a high risk of developing a life-threatening condition called sepsis, Poon said. The system first rolled out the tool, called Sepsis Watch, in emergency departments in 2018. 

In 2021, a prominent AI sepsis prediction tool built by health software giant Epic faced scrutiny after researchers found it had a high rate of false alarms, but Poon said Duke Health thoroughly vets AI tools before implementing them to make sure they’re safe, effective and equitable. Duke’s Sepsis Watch has reduced the system’s sepsis mortality rate by 31 percent, according to a news release.

Sepsis Watch was trained on data from over 42,000 patient encounters, according to Duke. It uses information like a patient’s vital signs and medical history to spot a high sepsis risk and alert a rapid response team.  

“The rapid response team will take a second look at the chart and talk with the patient’s primary care team to see if the primary care team should be thinking about sepsis as a condition, and whether they are doing all the appropriate things to rule out sepsis or an underlying infection,” Poon said.

Providers have used Sepsis Watch to identify over 3,000 patients with a suspected infection, according to a Duke Institute for Health Innovation report.

UNC Health also developed its own sepsis detection model that’s more accurate than the model built into its electronic health record system, according to a presentation that system leaders shared at a 2024 Health Care Compliance Association conference.

8. AI flags patients with high suicide and other health risks

At Novant Health, a machine learning system analyzes patient medical record data in real time to identify patients at risk of suicide, according to a spokeswoman.

The Behavioral Health Acuity Risk model creates a simple, color-coded risk assessment that providers see when they pull up a patient’s electronic medical record, making it easy for them to act fast if a patient is at high risk of suicidal thoughts.  

The model was built by Novant experts in mental health, emergency medicine and psychiatry.

Duke Health uses an algorithm to spot oncology patients at higher risk of needing a hospital stay, Poon said. It’s “very much invisible to the patients, but it really helps us as a clinical team be smarter, so that we can focus on patients who need the most attention,” he said. 

9. AI helps plan operating room schedules

Duke Health is using AI to make operating room schedules more efficient. The AI model, developed by Duke Health researchers and implemented in system hospitals in 2023, predicts how long surgeries will take in the operating room. The researchers found it was 13 percent more accurate than human schedulers. 

The operating room is estimated to cost between $22 and $133 per minute, so seamless surgery scheduling can reduce expenses like overtime costs, the Duke study found. 

10. AI chatbot directs UNC Health staff to reference tools 

UNC Health is using generative AI to help providers spend less time on administrative tasks. The internal chat bot answers questions related to UNC Health and can direct users to resources in the system’s training libraries, UNC Health said in a news release. 

“This is just one example of an innovative way to use this technology so that teammates can spend more time with patients and less time in front of a computer,” said David McSwain, a pediatric critical care physician at UNC Children’s Hospital and UNC Health’s chief medical informatics officer, according to the news release.

UNC Health was one of several health systems that partnered with Epic in 2023 to pilot new generative AI tools developed with Microsoft.

This article is part of a partnership between The Charlotte Ledger and North Carolina Health News to produce original health care reporting focused on the Charlotte area. 

You can support this effort with a tax-free donation. 

The post 10 ways North Carolina health care providers are harnessing AI appeared first on North Carolina Health News.