Can Google Search Help Us Understand Mental Health?

Research suggests internet searches could predict relapses in schizophrenia.

Posted Sep 04, 2020

Matthew Libassi/The Feinstein Institutes for Medical Research, used with permission
Google search bar.
Source: Matthew Libassi/The Feinstein Institutes for Medical Research, used with permission

It’s a very personal relationship; our computers may know more about us than our closest friends and family—especially Google.

From medical ailments to simple questions, our search histories say a lot about what’s keeping us up at night. What if we can use that data to inform clinical care for mental health issues?

I am proud to say that colleagues and I have been able to do just that in a promising new research study recently published in the journal JMIR Mental Health.

In contrast to other medical areas, psychiatry is still nearly entirely reliant on subjective family and self-reporting symptoms. It's a bit antiquated, and diagnosis and treatment delays are significant hurdles for many of those with mental health conditions, like schizophrenia spectrum disorders (SSD).

Just like a doctor may order a patient to receive blood work or go for an X-ray to diagnose an ailment or help develop a care plan, in the field of psychiatry, we are looking at new tools, and tech, to help us do the same. We want to figure out how to incorporate online search history and harness technology to collect objective digital data to help complement self-reports and facilitate more informed treatment decisions. We already have some preliminary research, with promising results looking at social media, specifically Facebook activity, to predict hospitalizations and episodes of psychosis.

Our current Google study aimed to develop computational algorithms designed to accurately identify individuals with SSD and predict psychotic relapse based on their search activity.

Both healthy participants and those with SSD consented to the study. We developed machine learning algorithms to examine 32,733 time-stamped search queries across 123 search histories to identify the differences in search activity timing, frequency, and the language used.

The results showed that those with SSD did fewer searches, and their searches consisted of fewer words. During a relapse period, participants with SSD were more likely to use words related to hearing (heard, listen, sound), bio (eat, blood, pain), perception (see, touch, listen), and anger. They were less likely to use words related to health.

The reduced search activity may suggest declining interests and engagement with the environment as symptoms (hearing voices, increased delusional preoccupations) escalate, and individuals with SSD may become less invested in their everyday lives.

Psychiatry is a personal field, and the results show that online search activity can be a valuable window into our overall health. As with any research, we need to repeat the study with more people and include participants with other psychiatric diagnoses. There is a host of fundamental ethical questions and concerns, acceptability and implementation questions, and privacy issues. We must critically evaluate before anything like this is used in the mainstream, but results like these are promising.

Although search data alone is not sufficient to make a diagnosis or to predict relapse, the integration of this type of data with information collected through traditional clinical means could potentially be useful in the near future—and help bring psychiatry into the modern, digital age.