New predictive computer program could help detect people at high risk for depression

A team of scientists from Nanyang Technological University, Singapore (NTU Singapore) have developed a predictive computer program that could be used to detect people at increased risk of depression.

In trials using data from depressed and healthy groups of participants, the program achieved 80% accuracy in detecting those at high risk for depression and those at no risk.

Powered by machine learning, the program, named the Ycogni model, screens for depression risk by analyzing an individual’s physical activity, sleep patterns and circadian rhythms derived from data from wearable devices that measure their steps. , heart rate, energy expenditure, and sleep data.

Depression affects 264 million people worldwide and goes undiagnosed or untreated in half of cases, according to the World Health Organization. In Singapore, the COVID-19 pandemic has raised concerns about mental well-being. A new study from the Singapore Institute of Mental Health has highlighted a likely increase in mental health problems, including pandemic-related depression.

Activity trackers are estimated to be worn by almost a billion people, up from 722 million in 2019.

To develop the Ycogni model, scientists conducted a study involving 290 working adults in Singapore. Participants wore Fitbit Charge 2 devices for 14 consecutive days and completed two health surveys, which screened for depressive symptoms, at the start and end of the study.

The average age of participants was 33, with the sample closely reflecting Singapore’s ethnic population. Participants were instructed to wear trackers at all times and only remove them when taking a shower or when the device needs to be charged.

Professor Josip Car, Director of the Center for Population Health Sciences at NTU’s Lee Kong Chian School of Medicine (LKCMedicine), who co-led the study, said: “Our study successfully showed that we could leverage sensor data from wearable devices to help detect the risk of individuals developing depression. machine learning, along with the growing popularity of wearable devices, it could one day be used for rapid and unobtrusive screening for depression.”

This is a study that we hope can lay the groundwork for using wearable technology to help individuals, researchers, mental health practitioners and policy makers improve well-being. mental. But on a more generic and futuristic application, we believe that such signals could be integrated into Smart Buildings or even Smart Cities initiatives: imagine a hospital or a military unit that could use these signals to identify people at risk..”

Georgios Christopoulos, Study Co-Leader and Associate Professor, Nanyang Business School, Nanyang Technological University

The results of the study have been published in the peer-reviewed scientific journal JMIR mHealth and uHealth in November.

Vital signs linked to depressive symptoms

In addition to being able to accurately determine whether individuals had a higher risk of developing depression, the researchers were able to associate certain behaviors of the participants with depressive symptoms, including feelings of helplessness and hopelessness, loss of interest in daily activities and changes in appetite or weight.

Analyzing their results, the scientists found that those who had more varied heart rhythms between 2 a.m. and 4 a.m. and between 4 a.m. and 6 a.m. tended to be prone to more severe depressive symptoms. This observation confirms the results of previous studies, which had stated that changes in heart rate during sleep could be a valid physiological marker of depression.

The study also associated less regular sleep patterns, such as variable wake-up times and bedtimes, with a higher tendency to have depressive symptoms.

The scientists explained that although the rhythms of the week are mainly determined by the work routine, the ability to follow this routine better differentiates depressed individuals from healthy individuals, where healthy people have demonstrated greater regularity. in waking and falling asleep times.

Professor Car added: “We look forward to extending our research to include other vital signs in detecting the risk of depression, such as skin temperature. Adjusting our program could help facilitate a detection early, discrete, continuous and profitable. of depression in the general population.”

Prof Assoc Christopoulos added: “Our team will also be working on extending it to other types of psychological states, such as mental fatigue, which seems to be an alarming problem these days. ‘a feedback system that could help therapists better assess the psychological state of their patients – for example, improved sleep quality.’


Journal reference:

Rykov, Y. et al. (2021) Digital biomarkers for depression screening with wearable devices: a cross-sectional study with machine learning modelling. JMIR Mhealth Uhealth.

Gordon K. Morehouse