OBJECTIVE: Online forums allow people to semi-anonymously discuss their struggles, often leading to greater honesty. This characteristic makes forums valuable for identifying users in need of immediate help from mental health professionals. Because it would be impractical to manually review every post on a forum to identify users in need of urgent help, there may be value to developing algorithms for automatically detecting posts reflecting a heightened risk of imminent plans to engage in disordered behaviors. METHOD: Five natural language processing techniques (tools to perform computational text analysis) were used on a data set of 4,812 posts obtained from six eating disorder-related subreddits. Two licensed clinical psychologists labeled 53 of these posts, deciding whether or not the content of the post indicated that its author needed immediate professional help. The remaining 4,759 posts were unlabeled. RESULTS: Each of the five techniques ranked the 50 posts most likely to be intervention-worthy (the "top-50"). The two most accurate detection techniques had an error rate of 4% for their respective top-50. DISCUSSION: This article demonstrates the feasibility of automatically detecting-with only a few dozen labeled examples-the posts of individuals in need of immediate mental health support for an eating disorder.
OBJECTIVE: Online forums allow people to semi-anonymously discuss their struggles, often leading to greater honesty. This characteristic makes forums valuable for identifying users in need of immediate help from mental health professionals. Because it would be impractical to manually review every post on a forum to identify users in need of urgent help, there may be value to developing algorithms for automatically detecting posts reflecting a heightened risk of imminent plans to engage in disordered behaviors. METHOD: Five natural language processing techniques (tools to perform computational text analysis) were used on a data set of 4,812 posts obtained from six eating disorder-related subreddits. Two licensed clinical psychologists labeled 53 of these posts, deciding whether or not the content of the post indicated that its author needed immediate professional help. The remaining 4,759 posts were unlabeled. RESULTS: Each of the five techniques ranked the 50 posts most likely to be intervention-worthy (the "top-50"). The two most accurate detection techniques had an error rate of 4% for their respective top-50. DISCUSSION: This article demonstrates the feasibility of automatically detecting-with only a few dozen labeled examples-the posts of individuals in need of immediate mental health support for an eating disorder.
Authors: Natasha L Burke; Guido K W Frank; Anja Hilbert; Thomas Hildebrandt; Kelly L Klump; Jennifer J Thomas; Tracey D Wade; B Timothy Walsh; Shirley B Wang; Ruth Striegel Weissman Journal: Int J Eat Disord Date: 2021-09-23 Impact factor: 5.791
Authors: Christophe Lemey; Aziliz Le Glaz; Yannis Haralambous; Deok-Hee Kim-Dufor; Philippe Lenca; Romain Billot; Taylor C Ryan; Jonathan Marsh; Jordan DeVylder; Michel Walter; Sofian Berrouiguet Journal: J Med Internet Res Date: 2021-05-04 Impact factor: 5.428