Could ‘machine learning’ hold the key to detecting eating disorder risk?
Eating disorders are serious, potentially life-threatening mental illnesses, which affect around 1.2 million people in the UK.
Anorexia has the highest death rate of any psychiatric disorder, and other eating disorders such as bulimia can lead to severe medical complications.
Despite their devastating impact on the lives of young people, our understanding of how eating disorders develop is limited.
My initial findings show that personality profiles may be useful in predicting the development of different sets of disordered eating symptoms.Dr Zuo Zhang
King's College London
Dr Zuo Zhang, a postdoctoral researcher at King’s College London’s Institute of Psychiatry, Psychology & Neuroscience, is harnessing the power of ‘machine learning’ to better understand the risk factors, causes and interconnections between different types of eating disorders, including anorexia nervosa, bulimia nervosa and binge eating disorder.
Machine learning uses algorithms to look at lots of different factors at once and pick out the ones that best predict a given behaviour or disorder – in this instance, eating disorders.
Dr Zuo Zhang is interrogating large sets of data, from more than 2,000 adolescents, to identify risk factors and common characteristics of eating disorders, including measures of the brain, personalities, environment and genetics. He then examines how accurately the risk factors can predict future symptoms, as well as investigating how these risk factors interact with symptoms.
So far, Dr Zhang has looked at 200 potential risk factors, with around 20 of these strongly associated with eating disorders. For instance, his early findings around personality traits suggest that high neuroticism (characterised by unstable and negative emotions) is a shared trait of anorexia and bulimia.
Other traits appear to be unique to different types of eating disorders. For example, Dr Zhang’s research suggests that feelings of hopelessness are more common in young people with anorexia, compared to those with bulimia. “It indicates that building hope and addressing negative thinking is particularly important when thinking about potential treatments for anorexia,” says Dr Zhang.
Dr Zhang hopes his research will reveal some of the main causes of eating disorders, which could enable earlier and more accurate detection of high-risk groups, and ultimately, improve treatment options.
“My initial findings show that personality profiles may be useful in predicting the development of different sets of disordered eating symptoms,” says Dr Zhang. “Now we want to validate these models across larger sets of data, before looking at whether risk scales could be useful in a clinical setting.
“We’re just at the beginning of understanding how computer algorithms could be used to detect eating disorder risk. But the potential of machine learning models to offer accurate, accessible and cost-effective information, for predicting, diagnosing and treating eating disorders, is hugely exciting.”