Advances in mobile health technology have transformed the ability of physician groups, employers, nursing home facilities, and pharmaceutical companies to capture data in the healthcare space. Watches, trackers, and other devices can seamlessly provide real-time data streams to applications and third parties. Called wearables, these devices are worn or carried on your body and can collect many different types of data. Gone are the days of paper trackers for monitoring blood sugar and single-use exercise trackers like pedometers.
The data collected from wearables can be used for clinical research, patient monitoring, and wellness tracking, among other uses. Each data point collected can add complexity to your broader data set. Because of the amount and complexity of data, turning to machine learning (ML) can help organizations leverage their data to identify patterns and make data-driven decisions.
By applying machine learning techniques to wearable device data, we can now surface patterns in big data and make predictions about behavior. Machine learning enables healthcare-related industries to leverage wearable device data and identify trends, improve recommendations, and define research outcomes.
Popularity of wearable devices
Wearables are popular, and their adoption continues to grow. Globally, the wearable technology market is expected to grow from $69 billion in 2020 to $81.5 billion in 2021, an 18.1% increase, according to the latest forecast from Gartner. What’s fueling the growth? Demand for smart devices in the healthcare sector is rising, as is demand for Internet of Things (IoT) devices.
Many wearables are not fitness-specific, featuring notification of text messages, push notifications for mobile apps, and the ability to pay for items by scanning a QR code with Google Pay or Apple Pay. As such, they have broad appeal.
“As a result of the pandemic, we have seen wearable devices become much more than just activity trackers for sports enthusiasts. These devices are now capable of providing accurate measurements of your health vitals in real time. Improved measurement accuracy coupled with the latest advancements in ML make it possible to detect abnormalities before they lead to a major health event.”– Alex Matsukevich, Fusion Alliance Director of Mobile Solutions
Types of data collected by wearable devices
There are a variety of brands and categories of wearable devices, from mass-market consumer versions to highly specialized types created for niche uses. Apple, Fitbit, Google, Samsung, Garmin, LG, Sony, and Microsoft dominate the market. Though the concept of “wearables” includes a focus on wristwatches, while exercise equipment, glasses, and textile sensors are also becoming more common. Wearable devices can measure:
- Sleeping patterns
- Heart rate
- Irregular heart rhythms
- Location/route during exercise
- Pace, stride, and distance while moving
- Blood oxygen levels
Limitations of wearable device accuracy
Wearables do have limitations, and accuracy is a concern. Healthcare decisions made using erroneous data could have outcomes detrimental to a patient’s overall health.
A study from the University of Michigan reviewed 158 publications examining nine different commercial device brands. In laboratory-based settings, Fitbit, Apple Watch, and Samsung appeared to measure steps accurately. Heart rate measurement was more variable, with Apple Watch and Garmin being the most accurate, and Fitbit tending toward underestimation. But for energy expenditure (calories burned), no brand was deemed accurate. This does not mean that the results are invalid, but that there is a significant difference between results from wearables and clinical results in a lab setting.
Wearable devices are constantly upgraded and redesigned as technology improves. And data collected by wearables does not provide a clinical diagnosis. As such, this data is just part of the larger picture of health and can be used only in conjunction with other factors to evaluate your overall wellbeing.
Overcoming the biggest challenge of wearable device data analysis
Healthcare professionals are already using ML to analyze data for patients. Research published in the International Journal of Research and Analytical Reviews confirms that ML techniques are successful in predicting health conditions such as heart disease, diabetes, breast cancer, and thyroid cancer.
The biggest hurdle to incorporating wearable device data into broader data sets is the addition of new inputs, such as hours of sleep or total steps walked per day. Traditional data points such as total cholesterol or blood pressure readings are less frequent, so there is a smaller amount of data overall.
The challenge in finding value from wearable device data is how to best incorporate it into other data sets. It will take collaboration among ML researchers and healthcare professionals to find value in wearable device data.
The future of wearable device data and machine learning
We can glimpse into the future of wearable device data and machine learning with Microsoft’s recent patent filing. Their potential product aims to provide wellness recommendations based on biometric data, such as blood pressure and heart rate, pertaining to work events.
To do this, Microsoft requests access to applications used by employees. Microsoft then tracks data points such as:
- Duration of time spent writing emails
- Number of times a user refreshes their inbox
- Time spent reading emails
- Number of corrections made when writing emails
- Recipient list for emails
- Number of meetings in a day
- Tone of language in emails
By combining this information with biometric data (from a secondary device such as a Fitbit or Apple Watch) and machine learning, Microsoft could begin to understand what work events trigger a response.
For example, suppose an employee received an email from their manager. Microsoft might observe that the employee spent a higher-than-average amount of time reading the email and that the employee’s heart rate was also elevated during this time.
Based on these insights, Microsoft could propose recommendations for helping employees manage stress levels, highlighting events that trigger anxiety.
With a broad user base using both Office and Teams already, Microsoft has a deep understanding of work-related events. As Facebook built their business making sense of our social lives, Microsoft has the potential to optimize our work lives.
“Wearables combined machine learning will become the new standard in personalized consumer electronics, rapidly increasing in popularity and scale every year until then. An integrated device of the future will be able to get a baseline of your health and will alert you to any abnormalities present. We already see this happening with the new Apple Watch, and it will be very soon that this technology becomes commonplace.”– Michael Vieck, Fusion Alliance Software Developer
Wearable devices will transform healthcare experiences
Data is the key to predicting, understanding, and improving health outcomes. IBM Research anticipates that the average person will generate more than 1 million gigabytes of health-related data in their lifetime, equivalent to 300 million books. The sheer volume of data means that machine learning will be vital in making sense of it.
Paired together, wearable devices and machine learning have the potential to transform healthcare experiences. Today’s applications and uses are only the start.