South Korean researchers developed a prediction algorithm they believe could significantly influence the treatment of depression by more accurately predicting mood changes using data from smart bands and phones.

The Korea University College of Medicine staff say their algorithm can predict mood swings to an accuracy rate of 98 per cent using data relating to sleep and time spent awake.

In a statement, the team noted wearable devices hold promise in helping to detect disruption to regular sleep cycles, which they explain can hold serious sway over a person’s mental wellbeing.

Wearables enable “passive collection of circadian rhythm indicators” spanning “sleep, heart rate and step-count data”, which can then be used to predict changes to a patient’s mood. The university team argue the wider range of data required by current machine learning approaches thwarts their practicality.

The algorithm was developed in work involving 168 patients, with mathematical modelling helping to identify 36 relevant sleep and circadian rhythm elements which effectively predicted the likelihood of a manic or hypomanic episode occurring within a 24-hour period when used in conjunction with historical information.

Lead professor Heon-Jeong Lee said the research could ultimately lead to the development of digital therapeutic methods for people prone to “mood episodes”.