In last week’s blog post, I wrote about the value of being able to make personalized care recommendations based on the evidence base. The challenge of Randomized Control Trials (RCTs) is it generally tells us a lot about the average population but not the individual implications. The promise of new data sources and new analytic pathways suggest we can get more granular for individuals.
One area that is growing in both significance and specificity is digital phenotyping- an ability to track your daily cadence of life via smartphone and sensor derived data. With continuous data monitoring, the potential to develop new insights on preventing and treating disease is dramatic. A new paper by Kit Huckvale from the University of New South Wales in Sydney and colleagues in Nature Digital Medicine brings an ethical dimension to the opportunity and purpose of clinical digital phenotyping.
What if we could early warning signals when someone who has bipolar disorder starts to decompensate? What if we could tell by sensor data when someone with Parkinson’s needs more assistance, and their symptoms have worsened. The potential use cases for clinical digital phenotypes are already in testing – the application of these new methods seems manifold. Apart from the practical aspects of interoperability, have we also thought enough about the safeguards we need to put in place, so we have appropriate data use and transparency?
Huckvale and team propose several areas for consideration in evolving digital phenotypes, and they are listed below.
Development in this space is currently in the health tech realm- few if any health systems are collecting digital signals from their patients and fewer still have mechanisms to import data into their Electronic Medical Record (EMR). The potential to develop new models of care to better predict someone’s care trajectory seems worthy of exploration. All too often health systems are reactive to health issues, for example, a new bipolar episode – what if care teams could reliably predict that episode and thus reduce the intensity and duration so that person would have more symptom-free days and a better quality of life?
Use of texting cadence, geolocation, phone call frequency, and duration offer us a window into new metrics for prediction modeling in mental health. Huckvale suggests data strategies need to evolve well beyond the current state to be clinically valid and accepted by care teams and patients. Digital phenotyping potentially allows us to move toward a deeper understanding of mechanisms that underly mental health conditions (changes in communication patterns, etc.), supporting a focus on preventing an episode, not just the prediction of outcomes.
The digital health era generates a lot of data. Huckvale cautions that first use cases for digital phenotyping must have clinical relevance and not just focus on issues that might grab headlines but won’t solve for a current pain point in health care. Concentrating on what can be modified seems prudent.
The ability to personalize care recommendations is exciting and also comes with ethical obligations. Considerations on data use, transparency, and privacy all loom large in digital phenotyping. Often disparate data sources (e.g., purchasing behavior patterns and geolocation) may form the phenotype, how do we ensure data will be used ethically? What permissions will need to be in place to ensure this? How much of the phenotype is monetized and how transparent is that process? Will we have interoperability between all data sources and how will digital interventions emerge from all this data?
This article is provocative and insightful about creating a path forward where key aspects like data exchange and interoperability and built-in from the outset. Huckvale calls for the creation of a shared data repositories so extensive research can occur, not unlike the UK Biobank which opened in 2012 and enables research access to data to further our understanding of prevention and treatment opportunities.
Developments in this space are exciting and fast-paced, it will be interesting to see if the genesis of opportunities comes purely from tech companies or if partnerships with academia, research, government agencies, and health systems are forged to realize the potential of clinical digital phenotyping fully.
Thanks for reading – Trina
(Opinions are my own)
Toward clinical digital phenotyping: a timely opportunity to consider purpose, quality, and safety.