‘Think outside the box’: UL doctoral graduate applies machine learning to mental health screening

A researcher and doctoral degree graduate at University of Limerick has applied machine learning to gain new insights into symptoms of depression.

Darragh Glavin, from Ennis in County Clare but now resident in the US, graduated this week with a PhD focusing on the area of machine learning in mental health from UL’s Faculty of Science and Engineering.

“I hope to provide rich and important insights and findings through data in the healthcare sphere that will have an impact and be beneficial for the people,” Darragh explained ahead of his graduation with a Doctorate in Machine Learning-based Optimisation of Ultra-brief Questionnaires for Mental Health Disorders. 

Now based in Los Angeles where he was thankfully unaffected by the recent fires, Darragh did travel back to Ennis a few days earlier than planned to escape the poor air quality ahead of his graduation, where he was eagerly awaited by his family.

Growing up, the UL graduate was always technically inclined and drawn to understanding how things work. He built his own computer and was interested in learning more. A keen swimmer, he took part in galas at the UL Sports Arena and was impressed by its facilities and the campus as a whole. A degree in Electronic and Computer Engineering at UL was the obvious choice for him when it was time to go to college. 

Darragh enjoyed the software coding and machine learning aspects of his undergraduate degree. He decided to pursue a PhD, and when it came to choosing an area of research, his personal experience with mental health and his interest in machine learning and its potential applications, prompted him to explore how machine learning can contribute to advancements in the field of digital mental health.

“My supervisor’s ongoing research was in the mental health sphere. I knew it was a research area that I would be passionate about working in. I could apply my technical skills to tackle challenges in this sphere from a different perspective than the majority researching in the area, such as psychologists and psychiatrists,” Darragh explained.

Machine learning is a branch of artificial intelligence that enables computer systems to analyse data, identify patterns and generate insights or predictions based on that data. 

In the area of mental health research, machine learning can analyse large sets of data to help understand and predict who in the population is likely to experience mental health difficulties.

Darragh secured a place on a PhD programme run by Science Foundation Ireland’s Centre for Research Training in Foundations of Data Science, based at UL, where his research focused on the use of short questionnaires for screening of mental health disorders. 

“Questionnaires are commonly used in clinical settings to initially screen for common mental disorders because they are significantly cheaper and quicker than a full clinical diagnostic interview and have been proven to be quite effective and accurate at identifying the disorders,” he explained.

Screening for depression offers a gauge to see if an individual is suffering from depressive symptoms before they complete a full clinical interview for diagnosis. These questionnaires typically contain a list of nine questions asking the respondent to choose from a list of symptoms of depression (eg. lack of interest in doing things you once enjoyed, concentration difficulties, agitation, sleep difficulties).

The aim of Darragh’s research was to use machine learning to optimise questionnaires that are commonly used for the detection of mental disorders. 

“Ultra-brief versions of these questionnaires (two questions or less) are often used in their place to reduce respondent burden and save time. Most of these ultra-brief questionnaires were created based on the consensus of experts based on their clinical experience and knowledge of the disorder symptoms,” he said. 

Using machine learning techniques that ‘come without bias or predisposition’, Darragh wanted to create more predictive questionnaires that could identify more people with the disorders, expand existing knowledge and consider other combinations of symptoms. 

“I used a data-driven machine learning-based approach to develop new ultra-brief questionnaires that were more predictive of depression than the existing gold-standard questionnaires. I saw how the questions interact as a pairing plays an important role in maximising their predictiveness.” 

For his research, Darragh used data sets gathered by his wider research group, with sample sizes ranging from 800 to a staggering 98,000 people.

As part of his PhD, Darragh had the opportunity to coordinate a new study into the importance of question order through a research internship at Vrije Universiteit, Amsterdam. Here he developed new questionnaires for depression, anxiety, insomnia, stress, procrastination, and general mental health status screening in a university student population. These new ultra-brief questionnaires outperformed the existing gold-standard questionnaires in screening for the disorders in this population.

Darragh’s research has successfully demonstrated the application of machine learning in the field of mental health, with his findings giving new insights into symptoms of depression and other mental health disorders that may have been overlooked by the current consensus-based approach.

Darragh recently moved to LA to join Irish start-up Akara Robotics as Director of Analytics, where he has seen how data science and machine learning is now a key driver in how businesses and organisations process information to improve efficiencies. 

“My role focuses on precisely measuring and optimising the utilisation of surgical centre operating rooms to maximise productivity and revenue, and more accurately predicting surgery durations to reduce downtime.”

Darragh feels that his PhD prepared him well for industry. 

“The Centre for Research Training in Foundations of Data Science gave us a lot of exposure to different companies and people in industry.

“The experience of developing data pipelines for data pre-processing, and machine learning model training and evaluation have proven key to hitting the ground running in my job”, he said. 

Having moved to the US, Darragh has seen trends emerging in the area of data science and health care.

“I sense that US companies are happier to deploy AI based tools quicker in the healthcare sphere. The rise of ChatGPT has shown the potential power of AI to a wider audience than any AI tool before and now everyone wants to hop on the AI train. A lot want their own large language model, trained to their specific needs, so that users and stakeholders can access insights from their data using free text.”

Darragh has sage advice for future leaders in the field of data science and machine learning.

“Get stuck into projects to gain hands-on experience with data and coding but don’t ignore learning foundational knowledge on the statistics and machine learning algorithms that underpin the work. Be adaptive because not everything will work as you hoped. Think outside of the box and try any ideas and solutions you think may solve your problem. Be rigorous and adhere to good data handling and machine learning practices - you’ll have more robust models as a result.”