AI set to improve healthcare in the next 12-24 months

February 27, 2019
Fred Rahmanian, Chief Analytics & Technology Officer


AI is poised to improve healthcare

At long last, artificial intelligence (AI) is poised to improve healthcare. And I’m not the only one who agrees.

To name but one example, two-thirds of the attendees polled at a recent innovation summit by The Economist agreed on one thing: healthcare is the sector that will benefit most from AI.

AI has potential to address low-hanging fruit such as identifying which patients may have difficulty paying their bill as well as more complex challenges including determining which patients respond best to a particular mode of engagement or care management, mitigating physician burnout and improving access to emergency care in rural areas through AI-enhanced virtual care.

That’s why I hosted an AI Fireside Chat at HIMSS 2019. Three esteemed colleagues joined me:

  • Josh Newman, MD, MSHS, Chief Medical Officer, Salesforce
  • Nassar Nizami, Senior Vice President & Chief Information Officer, Thomas Jefferson University Hospitals
  • Bharat Rao, Partner & Leader Data Analytics for Healthcare & Life Sciences, KPMG

In the coming weeks, we’ll be posting videos from our session, but in the meantime, I want to share some of the themes of our discussion with you.

AI adoption in healthcare lags other industries

We started our conversation by discussing why AI is lagging in healthcare. Undoubtedly, siloed data and the absence of true interoperability help to explain why. We have reason to believe that U.S. Secretary of Health and Human Services Alex M. Azar and Centers for Medicare & Medicaid Services (CMS) Administrator Seema Verma agree.

During HIMSS, Secretary Azar’s office proposed new rules to improve interoperability of electronic health information, saying,

"These proposed rules strive to bring the nation’s healthcare system one step closer to a point where patients and clinicians have the access they need to all of a patient’s health information, helping them in making better choices about care and treatment…These steps forward for health IT are essential to building a healthcare system that pays for value rather than procedures, especially through empowering patients as consumers."

Among the headlines at HIMSS 2019 were comments by Verma,

"I applaud the Secretary’s [Secretary Azar] message today that made clear we must work together to build an interoperable, patient-focused health IT system, and to ensure that patients are at the forefront….

Let me be clear…the idea that patient data belongs to providers or vendors, is an epic misunderstanding. Patient data belongs to patients, period!

Information blocking is a thing of the past."

AI in medical imaging

Despite the impact the absence of interoperability has had on AI in healthcare, our panelists agreed AI is widely used in medical imaging. Take pap smears, for example. It is thought that more than 85 percent of pap smears are read using an AI algorithm. By extension, the acute care setting is ripe for AI due to the preponderance of available patient data.

Liberating the data

We had a robust discussion about the willingness of patients to share their data. An emerging consensus was sick patients as a group would be willing to liberate their data. Similarly, our panelists believe millennials as a population will freely share their data as they do today with Amazon, Facebook and Google. Confident millennials would reap the benefits of complete and integrated data, one panelist wondered aloud whether our healthcare system would be interoperable in time for him and his generation to benefit.

‘Explainability’

I’ve spoken often of the need for ‘explainability’ of analytics insights. At Geneia, we typically don’t allow algorithms to go into production unless we can explain the output of our models to clinicians.

Explainability key to machine learning's future in healthcare

Say one can predict the onset of type 2 diabetes, for example. It’s one thing to say we think there’s a propensity but, typically the next question the provider asks is “Why?” Most algorithms don’t, but we make sure that if we provide a prediction, we can also answer those types of questions.

Our AI Fireside Chat panelists, one of whom is a physician, agreed that more explainability would improve provider adoption and, in turn, patient satisfaction. In the words of one of our panelists, “When a patient asks why, it’s not enough to say ‘the computer said it.’’

Stay tuned. We’ll be posting videos from the AI Fireside Chat in the coming weeks.


Related Blogs