Using machine learning to track the mental health impact of the...

Dealing with a global pandemic has affected the mental health of millions of people. A team of researchers from MIT and Harvard University have shown that they can measure these effects by analyzing the language people use to express their fear online.

Using machine learning to analyze the text of more than 800,000 Reddit posts, the researchers were able to identify changes in the tone and content of the language used as the first wave of the Covid-19 pandemic from January to April 2020. Their analysis revealed several important changes in mental health conversations, including an overall increased discussion of anxiety and suicide.

“We found that there were these natural clusters that arose in connection with suicide and loneliness, and the number of jobs in these clusters more than doubled during the pandemic compared to the same months last year, which is a serious one Problem, ”says Daniel Low, a graduate student in the language and auditory life sciences and technology programs at Harvard University and MIT, and lead author of the study.

The analysis also found different effects on people already suffering from different types of mental illness. The results could help psychiatrists, or potential moderators of the Reddit forums studied, better identify and help people whose mental health is suffering, the researchers say.

“When the mental health needs of so many people in our society are inadequately met, even at the beginning of their studies, we wanted to draw attention to the ways many people suffer during this time, in order to reinforce and inform the allocation of resources, to support them. ”“, says Laurie Rumker, PhD student in the PhD program for bioinformatics and integrative genomics at Harvard and one of the authors of the study.

Satrajit Ghosh, a senior scientist at MIT’s McGovern Institute for Brain Research, is the lead author of the study, which was included in the Journal of Internet Medical Research. Other authors on the paper include Tanya Talkar, a PhD student in the Language and Hearing Biosciences and Technology Program at Harvard and MIT; John Torous, director of the digital psychiatry division at Beth Israel Deaconess Medical Center; and Guillermo Cecchi, a principal research fellow at the IBM Thomas J. Watson Research Center.

A wave of fear

The new study arose from the MIT class 6.897 / HST.956 (Machine Learning for Health Care) at the MIT Institute for Electrical Engineering and Computer Science. Low, Rumker, and Talkar, who all took the course last spring, had already done some research on the use of machine learning to identify mental disorders based on the way people speak and what they say. After the Covid-19 pandemic began, they decided to focus their class project on analyzing Reddit forums that deal with different types of mental illness.

“When Covid hit, we were all curious if it affects certain communities more than others,” says Low. “Reddit gives us the ability to see all of these subreddits, which are specialized support groups, in real time. It’s a truly unique opportunity to see how these different communities were affected differently during the wave. ”

The researchers analyzed contributions from 15 subreddit groups addressing a variety of mental illnesses, including schizophrenia, depression, and bipolar disorder. This also included a handful of groups working on issues not specifically related to mental health, such as: B. Personal finance, fitness, and parenting.

Using various types of natural language processing algorithms, the researchers measured the frequency of words associated with topics such as fear, death, isolation, and substance abuse, and grouped posts based on similarities in the language used. These approaches allowed researchers to identify similarities between the posts in each group after the pandemic broke out, as well as significant differences between groups.

The researchers found that while most members of the support groups started publishing Covid-19 in March, the group that addressed health anxiety started much earlier in January. However, as the pandemic progressed, the other mental health groups looked very similar to the health fear group in terms of the most common language used. At the same time, the personal finance group showed the most negative semantic change from January to April 2020, significantly increasing the use of words related to economic stress and negative mood.

They also discovered that the mental health groups most negatively affected at the beginning of the pandemic were those related to ADHD and eating disorders. The researchers hypothesized that people who suffer from these disorders will find it much harder to cope with their conditions due to lockdowns without their usual social support systems. In these groups, the researchers found posts about hyperfocusing in the news and relapses in anorexia-type behaviors as meals were not monitored by others due to quarantine.

Using a different algorithm, the researchers grouped posts into clusters such as loneliness or substance use, and then tracked how those groups changed over the course of the pandemic. Suicide-related contributions have more than doubled from pre-pandemic levels, and the groups that were significantly associated with the suicide cluster during the pandemic were the support groups for borderline personality disorders and post-traumatic stress disorder.

The researchers also found the introduction of new topics specifically aimed at helping with mental health or social interaction. “The topics within these subreddit support groups have shifted a bit as people tried to adapt to a new life and focus on how to get more help when needed,” says Talkar.

While the authors stress that they cannot imply the pandemic as the sole cause of the observed linguistic changes, they note that there were much larger changes in the period from January to April 2020 than in the same months of 2019 and 2018, indicating this Changes cannot be explained by normal annual trends.

Mental health resources

This type of analysis could help mental health care providers identify segments of the population most vulnerable to mental health declines not only from the Covid-19 pandemic, but also from other mental stressors such as controversial elections or natural disasters caused.

When this analysis is applied in real time to Reddit or other social media posts, it can be used to offer additional resources to users, such as: B. Instructions for another support group, information on how to find psychiatric treatment, or the number for a suicide hotline.

“Reddit is a very valuable source of support for many people suffering from mental health problems, many of whom may not have formal access to other types of mental health support that Reddit could provide,” says Rumker.

The researchers now plan to use this approach to investigate whether posts on Reddit and other social media sites can be used to identify mental disorders. A current project involves screening posts on a social media site for veterans for suicide risk and post-traumatic stress disorder.

Follow the latest news on the Coronavirus (COVID-19) outbreak

More information:
Daniel M. Low et al. Natural Language Processing Reveals Vulnerable Mental Health Support Groups And Increased Health Anxiety On Reddit During COVID-19: Observational Study, Journal of Medical Internet Research (2020). DOI: 10.2196 / 22635

Provided by the Massachusetts Institute of Technology