Becker's Hospital Review

45 CIO / HEALTH IT IBM's retreat from health business spotlights AI challenges in healthcare: 4 things to know By Jackie Drees W ith IBM's exploration to sell off its health busi- ness unit IBM Watson Health, e Wall Street Journal highlighted several issues with artificial intelligence in healthcare that can hinder tech companies' innovation efforts. Despite spending several billion dollars on acquisitions to scale Watson Health, IBM's health business currently isn't profitable and is looking to sell, according to the Journal. IBM declined to comment on the sale, but offered the following statement to the publication about its successes over the past decade: "is work began nearly 10 years ago, at the beginning of the AI revolution, and we explored groundbreaking space in helping physicians advance healthcare through AI. IBM is continuing to evolve the Watson Health business, based on our decade of experience, to meet the needs of patients and physicians." Here are four things to know about the challenges AI tech faces in healthcare, according to the report: 1. Attempting to apply AI to treating complex medical con- ditions is one of the main challenges the industry faces, with contributing factors such as human, financial and technological barriers as well as getting access to data that represents patient populations broadly, healthcare experts told the Journal. 2. Tech companies sometimes also lack the deep expertise and knowledge of how healthcare works, which can exacerbate struggles of implementing AI in patient settings, said omas Fuchs, dean of AI and human health at New York City-based Mount Sinai Health System. "You truly have to understand the clinical workflow in the trenches," he said. "You have to understand where you can in- sert AI and where it can be helpful." 3. Some tech industry AI projects such as IBM Watson Health may be overly ambitious about healthcare's growth market; they sometimes take on too broad, but complicated health questions. IBM Watson Health, for example, was mar- keted broadly at finding answers to all types of cancer, the healthcare experts said. 4. e lack of data collection standards can also contribute to AI challenges in healthcare because it makes taking an al- gorithm developed in one specific setting difficult to apply in other situations. n How UConn Health, Novant Health are fixing tech glitches with vaccine appointment sites By Hannah Mitchell U niversity of Connecticut Health and Novant Health fixed glitches with vaccine appointment websites after complaints about glitches crashed in. Here are five things they have done to mitigate large flows of online traffic, according to a March 9 article published in The Wall Street Journal: 1. UConn Health, based in Farmington, Conn., added three servers to handle increased traffic. 2. UConn Health's vaccine booking tool did not hold a spot until the final confirmation page. It was a race to get to the confirmation page the fastest. It changed this to be fairer to users less adept with technology. 3. UConn Health shortened the automatic logout time for idle accounts to prevent the system from getting overwhelmed by large amounts of online users keeping the website open to nab newly available appointments. 4. Winston-Salem, N.C.-based Novant Health is using a chat box to answer hundreds of COVID-19 queries daily. 5. A separate call center is able to schedule 650 vaccine ap- pointments each day at Novant Health. n Pittsburgh hospital employee inappropriately disclosed patient health information, UPMC says By Jackie Drees P ittsburgh-based UPMC St. Margaret began notifying pa- tients March 5 that their protected health information had been exposed by a former hospital employee who wrongfully shared medication administration data with an outside organization. UPMC said it became aware of the breach Aug. 8. The health system discovered that a former employee had sent a med- ication administration report to an outside organization without a business need. UPMC terminated the employee's access to its information systems and said the individual is no longer affiliated with the health system, according to UPMC's March 5 statement. Patient information exposed included names, internal UPMC identification numbers and medication administration data, which may include drug names, dosages and reasons for admin- istration. The health system said no Social Security numbers or medical records were inappropriately accessed or disclosed. n

• Cover