Dr. Sheyn on the potential of AI in FPMRS

In this interview, David Sheyn, MD, discusses anticipations for the field of female pelvic medicine and reconstructive surgery (FPMRS) in the coming years, including insight into the role that artificial intelligence (AI) can play in new developments. Sheyn was just named the Division Chief of Female Pelvic Medicine at University Hospitals (UH), and he is also an assistant professor of urology and reproductive biology at Case Western Reserve University in Cleveland, Ohio.

Video Transcript:

I think there's going to be a rapid adoption of artificial intelligence platforms in our field. We saw that at the most recent SUFU [Society of Urodynamics, Female Pelvic Medicine & Urogenital Reconstruction] meeting. A lot of the research is in the development of these kinds of models, but very little is directed at proving that they're useful in the clinical setting. What I would like to see and what we're doing ourselves is we built models, and now we're doing randomized controlled trials, implementing these models in clinical practice to see if they're actually beneficial.

Last year, we developed a model to predict opioid use after urogynecologic surgery. The model was fairly predictive of how much narcotic to prescribe, and now [UH has] a randomized controlled trial that's halfway done recruiting, where we're randomly assigning patients to either a surgeon deciding how much medication they get to go home with, or we're allowing the algorithm to decide. What we hope will happen is that A.) there will be no difference in pain control; that patients in either group will have low levels of pain, but B.) that this model can help substantially decrease the number of opioids prescribed.

One of the biggest issues with the opioid epidemic is not necessarily that the patients are developing dependence on these medications, it's that there's a lot of leftover medication and there's diversion of these pills and no good way of getting rid of them. So, if you prescribe 1 or 2 instead of 20, then we can maybe make a dent in this in the future. When we were actually building the model, even though we thought we were doing a good job and prescribing not very many, we found that we were overprescribing by 200% to 300% on average.

This transcription has been edited for clarity.