Experts say the hype got ahead of the engineering but that doesn't mean algorithms can't make healthcare smarter and reduce costs too.
Success with artificial intelligence in healthcare depends on what you want the technology to do. A recent survey about AI in healthcare found that hospitals and insurance companies rate improving the payment process as the top priority for AI. Drug companies want to speed up the process of developing new medications. Employers want algorithms to do a better job monitoring IoT data.
Startups are using AI to do everything from analyzing heart rhythms to stopping the theft of drugs from hospitals. This doesn't even take into account what patients might want from AI, such as more accurate home monitoring or personalized healthcare that is available to everyone, not just wealthy patients.
This wide variety of use cases might be a clue about why IBM is thinking about selling Watson Health. The recent rumors may reveal more about IBM's approach to the challenge than what AI can do for healthcare.
Jeff Cribbs, a research vice president at Gartner, said that the rumors of IBM's decision to sell Watson Health is the familiar story of the hype cycle, with this version starring "the world's only genuine artificial intelligence celebrity (Watson) and an industry both very large and in particularly urgent need of transformation."
"There were genuine AI innovation triggers at Watson Health – in natural language processing and generation, knowledge extraction and management, and similarity analytics," he said. "The hype got ahead of the engineering, as the hype cycle says it almost always will, and some of those struggles became apparent."
SEE: Natural language processing: A cheat sheet (TechRepublic)
Cribbs said that some of the initial use cases for AI in healthcare that IBM was testing such as prior authorizations in insurance and oncology clinical decision making were more complex than the use cases that are showing early results.
"Where we have seen value from AI in healthcare in the last five years, it has maintained a narrow focus on a perceptual task – like diagnostic imaging, medical chart abstraction, remote patient monitoring or medical event prediction," he said.
In addition to tracking AI priorities, the Optum survey asked industry leaders how soon they expected AI investments to pay off. Pharma and biotech were the most optimistic with 74% of those executives expecting to see returns within the next three years. Only 56% of insurance companies and 49% of hospitals see returns coming that soon. Also, 57% of survey respondents in the late stages of AI deployment expect to see savings in two years. Based on this track record, IBM wasn't in for the long haul with Watson Health.
Adam Saltman, chief medical officer of Eko, said he wasn't surprised to hear the rumors of the sale because it seemed that IBM was trying to build a knowledge management system instead of a decision support tool for doctors.
"If you ask a doctor what they really need to make their practice better, an expert librarian system would not be the first thing they want," he said.
Eko makes algorithms and monitoring devices that can spot irregular heart rhythms. The medical device digitizes the heart sounds collected from a stethoscope or monitor device and sends the data to a smartphone or tablet. The algorithms look for crackles or wheezes that can indicate illness.
Saltman said that healthcare poses a unique problem for AI because many use cases aren't relevant to other industries. One example is improving care for people with heart failure. This complex condition often sends people back to the hospital and there isn't a one-size-fits-all healthcare plan for the illness.
"AI has to mimic the expert who has been practicing for 20 years," he said. "Image recognition is easy but patient observation isn't, and that's what the clinician does — makes an assessment of a person and then recommends a plan."
Healthcare organizations that do have a vested interest in helping doctors or patients might be better owners for Watson, he said. This could include a company like Optum, which is part of United Healthcare or Kaiser Permanente, which uses a business model that combines health insurance and care into one coordinated system.
"It doesn't have to be an insurer, just a company that recognizes what the real clinical driver is and then sets out to meet that," Saltman said.
Mark Scott, chief marketing officer at Apixio, said that AI's future ultimately relies on the continued digital transformation within healthcare systems to better measure clinical activities, guide care and make life-changing discoveries.
Apixio builds AI algorithms to do natural language processing for healthcare clients. Scott said the platform can surface insights from unstructured text and EHRs that were previously difficult to generate and use at scale.
Matching AI with human needs in healthcare
Kevin MacDonald is the CEO of Kit Check, a medication intelligence solution provider that uses AI to identify and stop drug theft in hospitals and other healthcare settings. He said the key to success with AI in healthcare is matching the technology with the expertise and needs of the humans who work in healthcare to make sure the tools fit the current demands.
"Partnering with providers has been imperative to ensure these AI solutions meet their needs and are practical in varied clinical applications," he said. "Working one-on-one to honor the provider must come first to keep any technology solution relevant."
Challenges unique to healthcare tech
Cribbs of Gartner said that another challenge is the fact that integrating acquired companies is even more difficult in healthcare than in other industries. This is often the case even when the acquired companies are innovative and well-aligned to the market.
"It is important that healthcare technology companies that acquire products do not oversell that synergy story or underestimate how difficult it will be to deliver that synergy value to the market," he said.
Another challenge is understanding and complying with the highly regulated healthcare supply chain.
"You can see this with Amazon who applied for pharmaceutical wholesaler licenses in their path to launch into the commercial drug space," MacDonald said.
- IT leader's guide to deep learning (TechRepublic Premium)
- Building the bionic brain (free PDF) (TechRepublic)
- Hiring Kit: Autonomous Systems Engineer (TechRepublic Premium)
- What is AI? Everything you need to know about Artificial Intelligence (ZDNet)
- Artificial Intelligence: More must-read coverage (TechRepublic on Flipboard)