Year: 2010 Source: Biomedical Informatics Insights, v.3, (2010), p.19-28 SIEC No: 20100830

This paper presents the authors’ second attempt to determine the role of computational algorithms in understanding a suicidal patient’s thoughts as represented by suicide notes. It was hypothesized that machine learning algorithms can categorize suicide notes as well as mental health professionals & psychiatric trainees do. 33 notes from suicide victims were matched to 33 notes elicited from healthy controls. 11 mental health professionals & 31 psychiatric trainees were asked to decide if a note was genuine or elicited. Their decisions were compared to 9 machine-learning algorithms. Results indicated trainees accurately classified notes 49% of the time, mental health professionals 63% of the time, & the best machine learning algorithm, 78% of the time. (39 refs.) JA