Today we live in two worlds. Our physical bodies occupy real space, physical interactions and face-to-face conversations— while our digital twins exist in the profiles, posts, steps, check-ins and likes that lay in their wake.
With social media infiltrating every aspect of our lives— we may have a new answer to an age old question:
If a tree falls in a forest and no one is around to hear it, does it make a sound?
Yes— because the tree probably posted a ‘check out me falling’ selfie to their Instagram story.
In medicine we’ve relied on physical check-ups to help diagnose patients—disease specific screenings, blood tests and patient self-reporting of symptoms.
While innovation in and scaling up of these physical check-ups have drastically improved care for many— not all of the conditions we face are easily diagnosable using these tools. This is specifically true of mental illness, where self-reporting of symptoms can often be insufficient for proper diagnosis.
Take for example depression. An article in Current Psychiatry quotes studies that found that “26% to 45% of patients referred for ‘depression’ did not meet diagnostic criteria for a depressive illness.”
Another analysis from 2009 published in The Lancet found that general practitioners correctly identified depression in only 47% of cases.
To help physicians better diagnose depression, researchers are designing algorithms that use Instagram photos to reveal predictive markers of depression.
New research in predictive diagnosis— especially for mental illness— is increasingly leveraging the clues our digital twins leave behind.
Diagnosing Using Digital Footprints
The idea of using social media for diagnosing depression isn’t new. In 2015 studies came out that linked smartphone usage data with depression. The research found a link between depression and the amount of time a person spent using a smartphone, as well as the number of locations the user visited in a given day.
But as sharing becomes more visual; algorithms need to become more advanced to analyze the new media we’re sharing. New questions arise:
- Do depressed individuals use specific filters on Instagram? Do they take pictures of different things?
- Can an analysis of a person’s previous post be a strong indicator for depression?
The answer: yes. A new study built an algorithm that analyzed Instagram posts of depressed users and found that photos in their posts:
- Were likely to be bluer, grayer, darker and receive fewer likes
- Had a tendency to filter out color
- Used fewer artificial lightening filters compared to non-depressed controls
- Were more likely to have faces, but they tended to post fewer faces per photo
The resulting models outperformed general practitioners’ average unassisted diagnostic success rate for depression. In an interview with CBS news, Andrew Reece, a lead researcher in the study shed light on its implications:
“It's clear that depression isn't easy to diagnose, and the computational approach we've taken here may end up assisting, rather than competing with, health care professionals as they seek to make accurate mental health assessments. Humans just aren't very good at keeping track of information over many thousands of data points.” – Andrew Reece, Researcher at Harvard University Department of Psychology.
Why This Matters
The trail of data our digital twins are leaving behind could be critical in the future of front-line diagnosis. Training algorithms using machine learning could help us detect symptoms earlier, intervene sooner and greatly reduce the cost of care.
Imagine a future where during a yearly physical, doctors check on your physical self and digital twin in order to better understand a complete picture of you overall health.
First, during check-in and setting up the appointment, a pre-appointment survey has you consent to sharing your social media accounts to detect for signs of mental illness.
Next, the pre-appointment survey accesses information from your Fitbit and Smart Scale to determine your recent activity and weight fluctuations.
The tool then asks you to share your Google Search data, to mine for specific symptoms you’ve searched for since your last visit.
While this may seem over the top, there is a probability that systems like this will exist in the not so distant future.
Obviously privacy is a major concern. So are false-positives associated with behaviors that fall outside the scope of the data algorithms were trained on.
But an ecosystem of tools— both those that create new streams of data our digital twins leave behind and those that leverage this data to algorithmically diagnose patients— is on the horizon. In the face of these realities, brands need to ask themselves:
- What role will we play in building these tools? What kinds of tools are relevant to the patients and physicians that use our products?
- What assets do we have or will we generate during the clinical trial process? What data are we generating through our patient support programs? Can we use this information to train algorithms?
- Do we build data-creating apps/sensors that generate data or algorithms to make sense of the data currently being generated?
- How comfortable are patients in receiving these types of ‘algorithmic’ diagnosis? How comfortable will physicians be in using them?
Although still in its infancy, this research gives us a brief snapshot into the future of diagnosis— a future where a ‘physical’ may also include a ‘digital’, too.