When you think of artificial intelligence in healthcare, you may think of the technology diagnosing and curing diseases. While AI does play a growing role in diagnostics, much of its most immediate impact goes on in the background, relieving doctors and nurses of paperwork and repetitive tasks.
The upshot? Caregivers are gaining more time to focus their attention where it’s most needed: on patients.
“If we can accelerate that task … it allows them more time to focus on their true purpose, which is diagnosing disease,”
Jesse Tetreault, NVIDIA
“If we can automate, if we can accelerate that task that they are forced to do, it allows them more time to focus on their true purpose, which is diagnosing disease,” said Jesse Tetreault, senior solutions architect at NVIDIA.
Tetreault delivered the keynote at AI4Health, an event that brought together clinicians, researchers and health system leaders at University of Delaware this month.
A theme ran throughout many of the panels and discussions at the daylong symposium: When AI works in healthcare, patients may never notice it directly. But data shows clinicians do, with less burnout and more time at the bedside.
“These mechanisms can help teams be energized and patients well served, even as demand patterns shift,” said Tim Gibbs of the Delaware Health Force.
Some inefficiencies can be potentially eased by applying AI to hardware. Susan Smith, a nurse researcher with ChristianaCare, shared lessons from a humanoid robot pilot program at Christiana Hospital that was designed to handle deliveries and logistics. In practice, the technology exposed how tightly clinical work is tied to urgency and timing.
“When you’re taking care of a patient, you want something now,” she said at AI4Health. The robots were often slowed down by basic things, like the inability to exit an elevator if people were blocking them.
The pilot ended with pain points identified and hopes for future, more efficient robots. Ultimately, Smith said, AI and automation must be judged not by novelty, but by whether they meaningfully save clinician time and support patient-facing work.
AI analytics for retention and clinician workload
The stress of healthcare often leads to staff turnover, and some drivers are potentially preventable. Healthcare analytics startup Atalan is using machine learning to uncover hidden causes of burnout in clinician workflows, particularly those tied to electronic health records.
“People think that burnout is because you’re not resilient enough,” Atalan cofounder Tiffany Chan told Technical.ly last spring. “It’s not fair and it’s not true. Most burnout is because you were overworking.”
The company’s software analyzes passively collected data, such as time spent charting or messaging, to identify patterns that may slip by managers. That insight can translate directly into reclaimed time.
“We detected that these doctors’ turnover risk dropped,” Chan said. “They’re spending less time on the [electronic health record] and more time with their patients without affecting their patient access.”
Another way AI can help physicians is by helping pinpoint the right information at the right moment. Electronic health records contain years of patient history, but accessing it quickly remains a challenge.
“I don’t have the time to go through the [records] from five years ago,” Tom Schwaab, a urologist and health system leader at ChristianaCare, said during his AI4Health talk.
“If I have AI algorithms to help me,” he said, “my clinical recommendation will be completely different.”
AI-assisted summaries, Schwaab said, could prevent unnecessary tests and improve decision-making. “We have the data,” he said. “We just need a way to analyze.”
And the amount of patient data is ever-increasing. That’s not always good for AI, Connor Callahan, cofounder of AI oncology platform Acellus Health, said. “More data doesn’t necessarily mean more intelligence,” he said. “As you add more data, you’re adding more noise, so you have to filter down to what’s actually important.”
The solution, Callahan said, lies in structuring data so AI supports clinical judgment. “Structure doesn’t necessarily remove clinical judgment,” he said. “It preserves it.”
Thoughtful adoption, faster science
Even when clinicians see the potential, enterprise adoption of AI requires caution and validation, Adam Dakin, CEO of Keriton, which develops neonatal feeding management software for NICUs, told Technical.ly. This is especially true when it comes to clinical use.
“AI holds tremendous promise with regard to clinical decision support within our neonatal feeding management platform,” Dakin said.
“However, we are taking a thoughtful stepwise approach. Enterprise health systems are being appropriately judicious and requiring robust validation before adopting AI enabled platforms,” he said, adding that developers need to be mindful of the associated regulatory requirements.
The future of healthcare may include robots gliding through the halls of hospitals, but, as Gibbs of the Delaware Health Force put it, AI’s promise lies in helping doctors and nurses do what they’re trained to do best.
“The goal isn’t automation,” Gibbs said. “It’s augmentation.”
