Wearables and Analytics: Powerful Prescription for Better Healthcare

As a Machine Learning expert, I teach computers to recognize patterns that can be difficult for humans to notice. My professional experience has taught me that without the right perspective (or information), it’s easy to miss an important connection between what may appear to be unrelated variables. Due to a recent personal experience, I am painfully aware how this information gap can have potentially devastating results when it comes to patient care in the healthcare industry.

Let me explain.

Let’s say you are a caregiver for an elderly relative the parent of a young child, and you notice signals that should trigger a visit to a specialist. But because you lack detailed medical knowledge, you are unaware these are symptoms of a major medical issue.  During a regular check-up appointment, the doctor, who certainly has the medical knowledge, only sees a “snap shot” of the total picture and makes a medical assessment based on the information provided. The doctor doesn’t have enough information to ask the questions that would get to the heart of the larger diagnosis. The patient leaves undiagnosed.

I experienced this myself and ended up self-diagnosing a family member’s critical medical condition when the subtle signs were missed by our doctor.

This got me thinking.

As wearable technology becomes more pervasive, we have the technology to easily gather real-time data, and we have the analytics that can analyze it — so why not put them together?

Fast forward a few months. We are now testing a system designed to offer real data to enable patients and doctors to have more comprehensive information so better decisions can be made earlier.

Together with an extra-ordinary team of scientists here at PARC, co-led by Jonathan Rubin and Rui Maranhao and myself, we are testing this concept with panic attack patients. Panic attacks strike, sometimes for seemingly no reason, crippling the person with anxiety or fear. Despite their sudden onset, these attacks don’t come out of nowhere — they are triggered by factors inside the body. Additionally, there are behavior modifications such as breathing exercises or other stress-reduction techniques that can lesson or help avoid an attack, if the patient is forewarned.

Working with medical professionals, a pilot is underway that takes data gathered from a wearable device that is able to sense the subtle internal signs that a panic attack could be coming. The system sends a notification to the patient’s mobile device suggesting they take preemptive measures. While the pilot is not complete, so far the data shows promise in that an attack can be detected with advance notice. We hope that attack durations become shorter, less severe or even avoided as a result of the recommended breathing intervention.

This shows the power and promise of combining wearable devices, sensors, and data analytics to collect not just vital signs and quality of sleep, but also patient-reported data that can be used to predict and prevent medical issues. In the case of our panic attack trial, by helping a patient predict when an episode may happen they can better manage their condition and live a fuller life.

Shane Ahern just presented our paper on this work at The 2015 ACM International Conference on Pervasive and Ubiquitous Computing (UbiComp 2015) in Osaka, Japan in September 2015. The presentation sparked further great ideas in the direction of this vision.

An additional benefit of the system is that medical specialists or primary care providers can now be armed with more detailed and up-to-date information rather than just a “snapshot” from an in-person physical examination or blood tests taken a week ago.  Had my family doctor had this type of information, I feel we would have been informed about the diagnosis sooner.

There are other future possibilities as well. Right now, most wearable devices have a user-centric approach — for example, only the owner can access results recorded by a Fitbit. Think of the possibilities of that data being accessed by researchers and other healthcare professionals. Multiple streams of data could be used to study groups of people with similar conditions, adding more analytics layers. Of course, the data would be anonymous, unless the patient gives consent.

We are also looking to apply similar technology to enrich work done in the office and on the road – for example to predict extreme stress and fatigue, and increase safety for pilots or drivers.

I never realized my machine learning and data science background could save a life – but perhaps one day it will.

Hoda Eldardiry, research scientist, system sciences lab at PARC, a Xerox Company.