The myths and realities of AI in health care


Health care is facing an unprecedented need to reform, drive quality and cut costs. Growth in targeted, specific treatments and diagnostic technology, coupled with a rise in people with long-term and multiple chronic conditions, is creating unsustainable demand on the system. To thrive – or even merely survive – health care organizations must adapt and find ways to deliver better, more efficient care.

This need for greater efficiency and rising demand has sparked a discussion about how artificial intelligence (AI) will save the day, replace the physician and deliver precision medicine to all. Is the concept myth, reality or perhaps a mix of both?

Until recently, the health care industry has had little to no automation, very limited use of predictive analytics and a lack of complexity in the predictions clinicians do make. And yet we commonly see headlines implying that health care organizations can take tools like AI, machine learning and advanced algorithms and vastly change the clinical landscape overnight.

The result? Many health care organizations attempt to leap to the most advance uses of AI – automated diagnostics, complex imaging analysis and robotic surgery – as a first step out of the gate. When those attempts fall short of hyper-inflated expectations, reactions range from head scratching to complete disillusionment with the promise of advanced analytics and AI.

Closer to reality, the industry can use analytics to help doctors and other clinical staff in ways that drive value and remove the need to for time-consuming, low-value tasks before they take on the complex, big-ticket items.

AI learns from pattern matching and context through data searches. It’s a process, and it takes time to train a model. Keep in mind, too, that it’s also impossible to train the machine without bias, so first tackling low-prediction, low-complexity solutions help organizations learn how to train and build for greater complexity and predictive ability.

We don’t let a newly graduated medical student perform open heart surgery unaided as a resident, right? We likewise shouldn’t expect the world of AI to make that leap. Instead, some great places to start include:

Helping doctors and patients get the most out of face-to-face interactions

Primary care physicians have growing patient needs and limited time for each visit. This stress causes burnout and poor-quality interaction. Why not automatically prepare the patient’s records prior to a visit to highlight areas to focus on related to the appointment? Then, using AI, take the patient’s demographics and social determinates to present potential questions and outcomes to the doctor and structure the appointment.

Automating comparison of radiographic imaging

Radiologists spend a lot of time examining the relevant pathology across large sets of medical images. How might AI help clinicians make better diagnostic and treatment decisions? Maybe once the radiologist highlights a pathology on just one image – a CT scan of a liver tumor, for example – AI could automate a comparative exploration, finding every image where the liver captured on any modality and present the analysis immediately to the doctor. The physician would be able to compare the images more quickly and it would make it easier to study the pathological progression of disease.

Alleviating alert fatigue

Building automation is growing in hospitals and it's generating alarms -- from doors, medical devices, overhead announcements and more. The result is alarm fatigue, a problem of epidemic proportion that causes caregivers to miss legitimate warnings. Alarm fatigue is killing patients and causing other negative outcomes. But what if hospitals started training AI to determine when an alarm is potentially serious and to prioritize and escalate the call to action accordingly? A door being opened in the maternity unit and the movement of a baby requires immediate attention, but a door opening on the back of a staff area during break time on a sunny day does not. Alarms for patient critical medication require a nurse to attend, but for a hydration infusion the need isn’t so critical.

The true promise of AI in health care

AI will be a tool that supports physicians, not replace them. AI will help them use more data points to make better informed treatment decisions and spend more time having meaningful conversations with their patients. Just as the stethoscope didn’t replace the doctor’s ears, but instead augmented their skills, AI is the latest tool that will enhance the ability to deliver care.

The power and predictive abilities of data and analytics will change health and medicine in many ways, but that transformation needs to start somewhere. If we strive only for the moon and don’t appreciate the value in what’s immediately around us, the true promise and potential of AI will only be delayed further still.

If you would like to learn more, download this ebook: Emerging topics in health care, and if you have any questions, please connect with me on LinkedIn.


About Author

Greg Horne

Global Principal, Health Care

Greg is the SAS Global Principal for Health and is based out of Toronto, Canada - he joined SAS in August 2012. In this role, Greg has the opportunity to work with healthcare strategy in a way that focuses on outcomes as well as the cost, quality and other challenges that any modern health system faces. He is considered a thought leader in the future of health care and the introduction of patient focused technology.

Leave A Reply

Back to Top