The hidden cost of AI for hospitals

Hospitals and health systems across the U.S. are bombarded with new artificial intelligence-driven applications promising to solve big challenges for a better tomorrow. While many provide a clearly needed service, they aren’t a magic wand.

Advertisement

“AI cannot solve for broken systems or broken workflows,” said Deepti Pandita, vice president of informatics and CMIO of UCI Health in Orange County, Calif., on an upcoming episode of the “Becker’s Healthcare Podcast.” “If you are designing AI to be the square background, that whole approach is not going to work. AI tools promise efficiency gains, but struggle when introduced into real world workflows. I have had several experiences of this. Physicians, nurses and other clinical care team members are already overwhelmed with EHR burdens, and if the AI tool doesn’t seamlessly integrate into the workflow, it’s going to meet resistance and it’s going to fail.”

Dr. Pandita sees her role as ensuring AI implementations are “almost invisible” so they feel like a natural extension of workflows rather than a burden. She set clear governance frameworks to provide explainability of the AI tool, accountability and ownership of its success to combat against companies overestimating the return on investment for AI tools.

“There are a lot of hidden costs of AI, including workflow, training, accuracy issues and data science evaluations,” she said. “Make sure the ‘ROI will come’ conversation is happening rather than [them] falsely touting that the ROI is here now.”

Many health systems are considering or installing AI scribes to help ease the administrative burden for clinicians and streamline the revenue cycle process. But adoption doesn’t necessarily mean immediate results, and the hidden costs could run deep.

“The actual ROI depends on adoption, ongoing maintenance and integration costs,” she said. “The tool must be measured by real world impact, not just theoretical savings. The metrics for measuring ROI should include clinician adoption rates, reducing cognitive load, and improved documentation quality. This is what I call ‘soft ROI.’ It’s not like a hard cost ROI, however, there is enough social science behind [the idea that] if you do all this, if you reduce cognitive load, if you increase the retainability of a clinician when they are less burnt out, if you improve documentation quality, that improved billing, improved engagement with the system, being able to see more patients, that hard ROI will come. But it’s not there today. Having that conversation and presenting it in that manner is very important.”

Advertisement

Next Up in Artificial Intelligence

Advertisement

Comments are closed.