This morning I was listening to a really good (though now a bit old) episode of the awesome podcast Partially Derivative. The episode featured Glen Coppersmith, the founder and CEO of Qntfy (pronounced “Quantify”). Qntfy is working to tap into non-healthcare data to gain insights into mental health processes and outcomes.
The episode does a great job discussing the barriers to communication between fields as disparate as statistics/data science and clinical mental health. They talk about the ways that different fields often operate with the same methods, but frequently use different terms, which leads to translation issues. For instance, clinicians and health practitioners are often interested in specificity and sensitivity, while statisticians seek to optimize model fit. While these concepts get at similar properties of the model, they need to be discussed using common language in order to have understanding between communities.
Pointing to CDC research on suicide, the hosts highlight the rising trend in suicide rates over the past 15-20 years. Moreover, Coppersmith highlights the fact that while the CDC data is invaluable, it is consistently published with about a two year lag. While this delay may be acceptable in some policy areas, it should be considered as a prime obstacle to confronting what has been called the “suicide epidemic”.
With the challenges of data availability and the costs of data collection in mind, Qntfy has provided the technical implementation for a really cool data donation project called OurDataHelps. OurDataHelps allows people to donate access to their social media, fitness tracker, and/or wearable sensor data for the purposes of furthering mental health research. It began as a way to use the data of suicide victims, donated by their loved ones, to contribute to this project. This kind of access should not be done lightly, and Qntfy seems to have done a really good job in ensuring the protection and anonymity of the data they are entrusted with. They thoroughly vet the organizations that seek access to this data, and only grant it with the approval of their advisory board.
Coppersmith pointed out the need for data not only from those suffering from mental illness, but also healthy individuals. There will rarely ever be a perfectly balanced, unbiased dataset in the social sciences or public health, but having data on reasonably well matched controls helps tremendously. So while I do not want to urge everyone to blindly grant access to their data, I would encourage people to look into OurDataHelps, and consider donating access.
And listen to Partially Derivative if you are into data stuff and also like beer.