Wearable sensors for the monitoring of digital biomarkers are one of the most exciting trends for the future of healthcare. Most of this excitement stems from the fact that clinical assessments of motor (primarily) and non-motor (to a lesser extent) functions are recorded using qualitative means and scales during clinical assessments. As a result, assessments can be subjective, inaccurate, lengthy and, hence, insufficient. Wearables are perceived as a means of providing granular and patient-centric clinic-pathophysiologic maps. Potentially they can be used to identify/confirm early symptoms of a disease, response to treatment and disease progression – and, consequently, lead to actionable pharmacological and non-pharmacological insights.
A lot of progress has been made lately in the field of wearable sensors, with academics and entrepreneurs engaged in a race to come up with solutions that revolutionise how complex diseases and phenotypes are better understood and monitored. Pharmaceutical companies are already engaged in the development of tools that could help them offer a better service to their patients and healthcare professionals. To name a few examples, Novartis is developing a technology for use in clinical trials for patients suffering from chronic lung disease and Biogen is developing wearable tools to improve patient outcome predictions for multiple sclerosis. Recently, Takeda announced a programme to support inflammatory bowel disease management.
All of the above tools involve the use of wearable sensors and, to an extent, rely on the use of sensors capable of gathering motor data. This sounds straightforward, however, things are not as simple as they seem.
The barrier of ecological validity
Firstly, let’s look at the term ‘ecological’ validity. Imagine the extreme scenario of two individuals suffering from the same complex disease – let’s say this is multiple sclerosis – and we need to assess, using wearable monitors, how the disease of these individuals is progressing. In a nutshell, the idea is to capture activity/mobility data as a means of defining progression against expected outcomes.
These individuals will be given a set of wearable sensors possibly comprising a gyroscope and an accelerometer (again trying to keep things simple). The sensors will be deployed in a similar body part, they will be switched on and the patients will be sent home where any activity will be continuously monitored. In theory, this data will highlight physical activity levels, assess any progress (improvement or deterioration) and, based on this, clinicians will gain a better picture of how these individuals respond to treatment and if their quality of life is improving or not. Using insight from this data, clinicians may recommend follow-up treatment steps.
This scenario may sound reasonable. However, there is one parameter we have not accounted for – the environment. It should come as no surprise that activity and mobility levels are very likely to be impacted by the environment these individuals live in. For example, their homes – whether they live in a flat or a two-story maisonette – and whether they have the luxury of using a lift to get in or out of their building will contribute to the effort required to carry out everyday activities. Proximity to amenities can impact activity too. Individuals who are in the habit of walking to the shops vs. individuals who have to drive for their shopping will register very different activity levels. So will patients who live in rural vs. urban areas. The impact one-off events could have on the quality of data is important too. For example, fluctuations in the weather can play a key role in impacting how mobile individuals are. Unless the wearable monitoring systems are able to account for all these factors the quality of the data gathered will be questionable and healthcare professionals will be less empowered to make evidence-based recommendations.
Wearable sensors have the potential to develop into an invaluable module of the healthcare digital ecosystem. However, there are many parameters which that can influence an individual’s daily routine and wearable sensors will capture that. For the data to make sense, we need to be able to identify whether what is captured is actually correlated with the disease. Overcoming the boundaries of ecological validity is not straightforward.
The agility of data analysis and data interpretation platforms will make or break how impactful these systems can be. This may involve systems that can be calibrated to reflect the weighted contribution of an individual’s ‘ecology’. Similarly, the wearable sensors may feature machine-learning capabilities and filter out any data generated by the individual’s environment. Such sensors will be in a position to calculate reasonable benchmarks against which each individual will be assessed.Gathering meaningful data for a medical assessment in a non-controlled environment is a significant challenge and, unless it is overcome, medical device developers are unlikely to come up with solutions which provide trustworthy and impactful insights.
How well the methods, materials, and settings of a study using wearable sensors approximate real-world conditions.