LAS VEGAS—Artificial intelligence and machine learning were everywhere here at the giant HIMSS (Healthcare Information and Management Systems Society) 2018 Conference. It’s still early days for AI, but despite the industry’s history of struggles adopting new technologies, AI is getting into real products and starting to deliver tangible results.
Health care providers and insurance companies are adding machine learning to improve internal processes and doctor productivity, aid persons with disabilities and to find new treatments for disease.
For instance, at the Beth Israel Deaconess Medical Center in Boston, CIO Dr. John Halamka has several projects in the works, including using machine learning to automatically input text-based forms into an electronic medical record (EMR), to schedule operating rooms more efficiently and to predict “patient throughput,” which is when a patient will be discharged or if a patient will have to be re-admitted.
“There are several use cases,” he said in an interview. “How do I recognize pieces of paper and to file those pieces of paper to the right meta data, schedule an OR slot based on the procedure, doctor and patient?
Halamka noted that performing effective throughput forecasting is important because “we’re dinged [by payers] for 30-day re-admission rates. If I can say every time we’ve had a patient like this and they have always been re-admitted, then that tells us we should be doing something different.”
Halamka, whose staff is using Amazon and Google clouds for their machine learning efforts, is also using the cloud as a means to enable interoperability between patient records, an issue that has plagued health care organizations since 2009, when EMR mandates were enacted with the HITECH Act.
“You could take all the data from all the providers of care in a region and then figure out a patient’s longitudinal record,” he said. “You’ve got a lot of data and you have primary care, specialists and urgent care, so the cloud brings together all this stuff and you run your care management against that database.”
Assisting patients
Other solutions seek to have a more direct effect on patients’ lives and access to health care. Zocdoc, a service that helps people find doctors and schedule appointments, has a new service called Insurance Checker that can “read” insurance cards by snapping a photo of the card with a smartphone app. This can help patients who may not be able to decipher a card determine whether a doctor will accept their health plan.
Insurance Checker was announced last fall, and since then it has resulted in a 30 percent drop in no-shows and cancellations, according to Zocdoc CIO Serkan Kutan, who spoke at HIMSS.
AT&T and Aira announced they are collaborating on a prescription medication reader to help people who are blind or have low vision. The solution includes Aira’s Horizon glasses (see photo) with a camera and a smart phone that is tied to the “Chloe” digital assistant who can help the wearer. A live person is also available.
“No joke, it’s life-changing,” said Greg Stilson, who is Aira’s Director of Product Management, who also is blind. Tools such as Aira can help a blind person navigate complex territory, such as an airport, he said. “As a blind person, we are very A to B oriented. A sighted person can see along the way. But using Aira for the past 18 months I have not needed assistance. I can just get off a plane and go.”
Voice-enabled assistants are also starting to make an impact. Nuance, the makers of the ubiquitous speech recognition software, was demonstrating an Alexa-type system that can help doctors navigate EMRs and assist them in fleshing out a diagnosis or ordering tests or medicines. The service is being embedded in EMRs such as EPIC, the market leader.
Likewise, cloud-based EMR vendor athenahealth is working on mobile AI-based virtual assistants. The assistant can highlight terms and contextualize information. It can also memorize the screens that are needed for certain conditions. “Clinical inbox automation is where we want to go,” said Girish Venkatachaliah, VP of Engineering Analytics at athenahealth.
Business analytics vendor HealthCatalyst announced a benchmarking product that uses AI to find patterns in data across a provider’s enterprise, from clinical data to billings. Touchstone brings together patient data from the HealthCatalyst customers who opt in, compiling up to 125 million records since the product’s launch last month. Those numbers give the Touchstone AI a good sample of data from which to compare processes and outcomes that can then be acted upon.
Somatix combines a smartwatch app that monitors several vectors of the wearer’s behavior, including body location and movement as part of its anti-smoking product, SmokeBeat. The app can send messages of encouragement to smokers or tell them they are over their daily limit. Doctors can monitor progress and guide patient care on a real-time basis. Early results show that use of SmokeBeat resulted in a 15-20 percent decrease in smoking volume, said Somatix CEO Eran Ofir.
Nothing is certain here, as the health care industry has a poor history of integrating new information technology and making it work. But one thing is certain: Since the 2009 HITECH Act was passed, over 90 percent of hospitals are using EMRs, according to HIMSS. That adds up to a lot of available data with which to perform analytics and machine learning. It will be interesting to watch in the next year how far AI can go in health care.
Scot Petersen is a technology analyst at Ziff Brothers Investments, a private investment firm. He has an extensive background in the technology field. Prior to joining Ziff Brothers, Scot was the editorial director, Business Applications & Architecture, at TechTarget. Before that, he was the director, Editorial Operations, at Ziff Davis Enterprise. While at Ziff Davis Media, he was a writer and editor at eWEEK. No investment advice is offered in his blog. All duties are disclaimed. Scot works for a private investment firm, which may at any time invest in companies whose products are discussed in this blog, and no disclosure of securities transactions will be made.