Walk, talk, think, see and feel: harnessing the power of digital biomarkers in healthcare
The advent of digital measurements has opened up new avenues for comprehensively assessing health and wellness, enabling the detection of subtle changes that may aid in diagnosis and prognosis. However, to fully realise these benefits, there are a number of difficult challenges to navigate.
Ensuring diverse and representative datasets
While digital biomarkers offer much promise, those developing these approaches need to be mindful not to widen existing health inequalities. One challenge within existing data is the significant reliance on homogenous data sets used to train and benchmark new products or methods of analysis. Indeed despite considerable validation work within consumer grade wearables, it has been highlighted only a small handful of studies that examine the impact of skin tone on accuracy rates directly7,8. This has raised concern that there may be significant inaccuracies within data, which may be limiting accurate health-related information for individuals with darker skin tones- exacerbating existing structural health inequalities. There is also emerging evidence that pulse oximeters may have increased error rates based on skin tone differences by devices in response to changes in activity9.
Overall, a renewed focus on increasing the diversity of data will yield much richer data sets and improve the health equity of digital biomarker.
Navigating the ‘drawbacks of dimensionality’
While biomarkers provide clear opportunity and increased dimensionality of data sources, the increased array of sensors and, thus, data capture needs to be carefully considered against the broader healthcare aims, including sustainability and societal heuristics or behaviours displayed the public. AI is likely to aid in the processing and understanding of complex data, however increasingly there is broader ecosystem challenges that need to be addressed in parallel. Just because ‘we can, doesn’t mean we should’ is an essential reminder of frugality in the face of innovation. Increased ‘sensing’ and data capture have potential unintended negative impacts of data junk, storage and ultimately environmental impacts10.
Authenticity, regulation and security
The pace and scale of change to generative Artificial Intelligence (AI) provides unique opportunities (integration of multimodal datasets) and challenges (verification, validation, explainability and regulation). Likewise, healthcare is increasingly delivered within community and home settings, which offers many potential social, economic and environmental benefits. The negatives of this decentralised and consumer-first approach include exposure and increased vulnerability to cyber threats or compromise of personal data. These regulatory and security challenges require closer examination from both digital biomarkers and digital health ecosystem lenses10,11,12.
link
