As artificial intelligence continues to transform the area of health and wellness, companies are looking for solutions to bring down the incredible costs of healthcare down. Robert Haslinger of Bose (an ODSC East 2020 partner and a brand new player in the world of health devices) talks about the future of data science for health and wellness in his talk for Ai x at ODSC 2019. Let’s take a look.
What’s Happening in Data Science for Health and Wellness?
The most visible method of using AI for healthcare is for straightforward problems, such as what Haslinger calls “Find the Tumor.” These are straightforward problems that require simple data in and data out approaches with machine learning. Get approved by the FDA, and your solution is good to go.
This type of diagnosis is relatively straightforward, and it’s important. It has the potential to expand healthcare to underserved communities and other areas. Following this method is a second common usage, patient monitoring. Whether it’s in a hospital setting or at home, these avenues are vital to aiding doctors with patient care.
A third type is far less visible. Chronic disease is taking up a massive chunk of the healthcare pie—conditions such as heart disease, tooth decay, depression, and others. These don’t fit into the standard model of care because the diagnosis and treatment aren’t straightforward.
The World Health Organization defines health as “A state of complete physical, mental, and social wellbeing and not merely the absence of disease or infirmity.” This comprehensive view of health is the next frontier of AI research, according to Haslinger. Six out of ten adults have a chronic illness, and each one is caused and exacerbated by a variety of factors, including the decisions we make every day.
Changing behavior is a vital part of the fight against chronic diseases, but Haslinger notes that this is also one of the most challenging aspects of wellness. So how does AI fit into this piece of the healthcare puzzle?
Building AI-Driven Health Apps
The application of data science is a vast topic, but how do we make this concrete in the healthcare realm? There are three types of data that we might use to build real-world examples of products that can influence that critical aspect of health and wellness, patient behavior.
Making an application for behavioral change, you must first learn from the experts. Data looks very different in ideal conditions than it does out in the real world. You must find the expert and formalize the knowledge they have into an automated decision-making process.
Expert knowledge helps ensure that apps work out of the box. Healthcare apps must be actionable from the very beginning; you don’t have time to learn everything from data at a later date. User data may help you learn or adapt over time, but it must start by working well from the outset. Otherwise, the user won’t keep it.
The loop may look like a performance metric that adjusts the advice over time, plus a real-time control loop. First, however, you must gain expert knowledge just to get started.
Data About Physical State – Sensors
Sensors inform us virtually about the physical state. Haslinger predicts six types of sensors that will provide valuable feedback for AI-driven healthcare tech.
- movement—acceleration, GPS
- heart rate—ECG, PPG
- electrical—EEG, EOG
- respiratory—blood O2, breath rate
- dermal—temperature, GSR
- environment—microphone, light meter
These examples provide a way to analyze changes over time. These time-series analyses offer healthcare professionals and users the chance to find small details, aberrations in the normal, and offer insight into states.
Reducing the noise of these sensors is part of the next generation of healthcare apps, providing a real problem to work on for data scientists working specifically in the healthcare field.
Sensors also give an incomplete picture of state. You may measure healthcare data like heart rate, but you can’t measure how the user felt. Why was the workout less effective? Was it that she felt fatigued already from working late into the evening? This leads to the final piece of the puzzle.
Direct User Input and Feedback
User input and feedback not only gives the user some control over the app, a critical function in discouraging users from simply turning the app off, but it can provide the final piece of information healthcare providers, and users need for a complete picture of wellness.
Types of user input include things like targeted questions, ratings and scales, and even interactive chatbots. This personalizes the app and allows the user to tailor things to preferences, further encouraging them to interact with the site.
Information is vital for feedback, but that’s not the end. Giving actionable advice for more significant benefit helps people pay attention and gives them ways to continue on their health journey. Otherwise, they may not continue with the change.
Not Just AI, Design
AI and design is a loop. Nowhere is this more critical than in healthcare apps. The design is the experience itself. If the design isn’t quite right, it won’t offer a good experience, and it won’t be effective. You must have something that AI can support, and that support comes from logical design.
These designs have the potential to go clinical, as well. The field is moving away from managing chronic disease with pills and towards lifestyle management as a way to treat these conditions. The field is moving towards addressing the underlying cause, so support from AI-driven design may provide an extra layer of patient-centered, holistic treatments.
Future apps will be regulated under something called Software as a Medical Device. Once an app gets through the 510k, a regulation from the FDA, it’s considered usable, safe, effective, and actionable.
The Future of Data Science for Health and Wellness with AI-Driven Devices
Currently, the FDA is getting feedback for introducing more efficient regulations for AI. The field is moving towards augmenting what the physician can do by being present all the time when physicians can’t and providing a clearer, broader picture of a patient’s health. Getting the balance between AI and design correct could offer enormous opportunities for future engineers looking to break into the specific field of data science for health and wellness.