Is There an ROI on AI?

By David Harlow, JD MPH, Principal, The Harlow Group LLC
Twitter: @healthblawg
HostHarlow on Healthcare

A survey of C-level healthcare execs conducted by OptumIQ finds that the overwhelming majority of those surveyed believe that there will be a positive ROI on artificial intelligence investments in healthcare. Even allowing for the fact that OptumIQ sells AI solutions to those same healthcare organizations, the survey is worth mining for insight into the expectations that healthcare leaders have when it comes to AI.

  • Employers and health plans expect their AI investments to begin yielding a positive ROI within three years, and hospital execs expect it to take four or five years.
  • Employers are furthest along in deploying AI solutions in healthcare, though three-quarters of all those surveyed say they are either planning or implementing an AI strategy.
  • The key benefits respondents expect to see form AI are more accurate diagnosis and increased efficiency.

Of those health organizations that are already investing in and implementing AI:

  • 43 percent are automating business processes, such as administrative operations or customer service
  • 36 percent are using AI to detect patterns in health care fraud, waste and abuse
  • 31 percent are using AI to monitor users with Internet of Things (IoT) devices, such as a wearable technology

It is worth noting that much of the existing use of AI is in non-clinical areas of healthcare administration. The hope, of course, is that the clinical uses can prove themselves out and become more widely adopted. (Though, of course, at a certain point, an AI tool used in healthcare becomes subject to regulation by the FDA as a medical device; folks need to be aware of those lines, and of the shifting regulatory landscape.) Optum is looking at trying to predict atrial fibrillation using AI in order to permit clinicians to intervene earlier with preventive measures, thereby reducing the burden of disease (including the cost of treatment). Research indicates that 13% of atrial fibrillation is undiagnosed, so identifying patients in need of preventive services could affect a significant patient population. Since half of the 700,000 undiagnosed patients are at moderate to high risk of stroke, prevention could make a difference.

Another interesting survey released by PwC on Digital IQ, found that their respondents broke down into modernizers, efficiency seekers, industry explorers and redefiners. The first three groups seek to use digital tools to tweak their businesses, but the fourth group, the redefiners, comprising about a quarter of the respondents, are using these tools to transform their businesses at a more fundamental level, not by simply adopting digital tools here and there.

Similarly, if AI tools are going to transform healthcare organizations, they will need to be applied stem-to-stern, and not just on a piecemeal basis.

AI tools in healthcare (however we may define them) are still in their early stages of development, and they will need to be applied successfully to a broader range of populations and diseases, as well as business processes, in order to prove their value and truly deliver a compelling ROI. But the early returns are promising.

This article was originally published on HealthBlawg and is republished here with permission.