The Brookings Institution released a report opining that AI adoption in health care is lagging. Is this true?

Their report entitled, "Why is AI adoption in health care lagging?" is part of the series "The Economics and Regulation of Artificial Intelligence and Emerging Technologies”, from the Brookings Center on Regulation and Markets. The series focuses on analyzing how AI and other emerging technologies impact the economy, markets, and society, and how they can be regulated most effectively.

The authors premise that adoption of AI in health care is based on their review of U.S. job advertisements - by industry - that require AI-related skills. Ranking various industry types, the information industry's job postings required the most AI-related skills. Professional services and finance also rank relatively high in job notices requiring AI skills, followed by manufacturing, mining, and agriculture. At the bottom of that list, just above construction, was health care. 

I am skeptical that the data support these conclusions. Citing the famous prediction by Turing Award winner Geoff Hinton that AI will replace radiologists, the authors glibly observe, "Yet there are still plenty of radiologists." They contend that Hinton’s prediction has not yet come to pass because of a lack of trust in the algorithms, challenges in data collection, regulatory barriers, and a "misalignment of incentives." In their study analyzing AI adoption through job postings, they observe that AI skills are less likely to be listed in job notices for clinical roles than in administrative or research roles. Hospitals with an integrated salary model, the authors believe, have a higher rate of adoption of AI for administrative and clinical roles than for research roles, as compared to hospitals managed by doctors. Teaching hospitals are no different from other hospitals in their adoption rate. 

So is it true that health care is slow to adopt AI? Not from my perspective. I like to tell others that everything I know is anecdotal. From that perspective, I am finding that radiologists are embracing these new tools. One needs only to review the new articles monthly in JACR, AJR, and Radiology to discern how much focus is being brought to AI in radiology. Far from fearing obsolescence, as they add AI algorithms to their diagnostic arsenals, today's radiologists appear to be in agreement with the now-famous quote from Stanford radiologist Curtis Langlotz, "AI won’t replace radiologists, but radiologists who use AI will replace radiologists who don’t."