For the past few years, Georgia Tech School of Interactive Computing Assistant Professor Munmun De Choudhury has pursued research that gathers insights about mental health through digital traces individuals leave behind on social media.
Under a new $2.7 million grant from the National Institutes of Mental Health (NIMH), she and a team of researchers at Northwell Health will apply that new information in a clinical setting in hopes of improving treatment.
“In our past research, we have gained a number of new insights, but I see an opportunity to influence real world people and outcomes,” De Choudhury said. “Going beyond just academic and empirical findings, how do you take that information and make a difference in people’s lives? What research challenges do such translations pose to the computing domain?”
This grant offers the researchers that opportunity. It will be one of the first in which computing researchers and leading experts in psychiatry research are coming together to influence how treatment can be delivered harnessing patient-contributed data. The grant is funded through a new NIMH program designed to inform and support delivery of high quality mental health services.
The idea is to build machine learning algorithms based on data that mental health patients voluntarily share with the research team, including both clinicians at Northwell Health and researchers in De Choudhury’s lab at Georgia Tech. With these algorithms, they hope to identify different risk markers and symptom changes that appear in social media posts to identify changes and trends in an individual over time.
By combing a number of different social media sources, primarily Facebook and Twitter, they will look at the use of words or patterns of words an individual uses. In mental illnesses like schizophrenia, the main population they will explore, that is important information to know.
“If they are feeling delusional or experiencing paranoia, what is it that they are saying,” De Choudhury said. “We can look at social interactions and see whether they might be feeling isolation, which can have a negative impact on mental health. Nuances of language styles, like the way people use articles or pronouns, can say a lot about their psychological state, as well, which has been shown in our and co-investigator (University of Texas Professor) Jamie Pennebaker’s prior work.”
The population they will focus on comprises younger individuals, largely teens and early 20s, who have had a first episode of schizophrenia. Most will have only recently been diagnosed and admitted to a specialized treatment facility directed by the collaborators on the project in New York. The goal is to use the information gathered in their digital traces to identify risk markers that signal a potential relapse.
“Schizophrenia is a challenging and debilitating illness,” De Choudhury said. “Even people under treatment have a high chance of relapse with negative outcomes on quality of life, productivity, and functioning. Symptoms often come back, and most mental illnesses are only managed, not cured.”
Better management means that the treatment is timely and highly adaptable to the patient’s needs, De Choudhury said. Unfortunately, that’s a challenge because, in clinical settings, there is very little knowledge about a patient’s day-to-day life. Unlike a disease such as cancer, which has an objective screening that can identify its presence and severity, mental illnesses are based on what is reported. These self-reports are often skewed, based on what a patient wants to tell or remembers.
“In some ways, the treatment paradigm right now is not very evidence based,” she said. “But to prevent relapse, it’s important that we try to be as precise and proactive as possible.”
The project will span four years and began on April 15.