Electronic media and electronic case records offer many new ways to conduct affordable and timely research in mental health. The two studies summarised here are examples of an emerging trend.
Rashmi Patel, Nishamali Jayatilleke, Richard Jackson, Robert Stewart and Philip McGuire used the electronic case records of 7678 adults with schizophrenia receiving care in 2011 at Maudsley. A text mining tool was developed using machine learning software to collect information on negative symptoms. Association of negative symptomatology with age, sex, relationship status, impairment of activities of daily living, and (for inpatients) length of hospital stay was studied. 56% had at least one negative symptom. Younger males were particularly at risk. Negative symptoms were associated with more admissions. Emotional withdrawal and apathy were more associated with longer hospital stay.The data suggest that negative symptoms are evident in most patients with schizophrenia and are associated with poor clinical outcomes. Most important conclusion is that large scale electronic data mining is possible with machine learning softwares.
Do tweets containing suicide-related content correlate with actual rates of suicide in an area?
Jared Jashinsky et al analysed the twitter feeds using specific criteria for filtering and identifying potential suicide-related tweets. From these ‘at risk tweets’, ones with location/state were selected. Exclusion techniques were then used to exclude sarcastic/jokes/non pertinent tweets. The proportions of at-risk tweeters per state with respect to the total number of at-risk tweeters were computed. A baseline based on a random sample was also generated for comparison. Between May 2012, to August 2012, a total 1,659,274 tweets from 1,208,809 unique users throughout the world were studied.
Analysis showed an association between rates of tweets by users determined to be at risk for suicide and actual suicide rates. States in the midwestern and western US region and Alaska were observed to have a higher proportion of suicide-related tweeters than expected. These states also have the highest actual rates of suicide.
Text mining/ social media mining offer an affordable and rapid research tool. It can aid in rapid detection of trends in suicidal thinking/suicidal acts. It may also help in detecting individuals with suicidal intentions. This may subsequently provide a platform to improve suicide prevention strategies through timely intervention.
Summary of articles
Tracking suicide risk factors through Twitter in the US. Jashinsky J, Burton SH, Hanson CL, West J, Giraud-Carrier C, Barnes MD, Argyle T. Crisis. 2014;35(1):51-9.
Investigation of negative symptoms in schizophrenia with a machine learning text-mining approach, Rashmi Patel, Nishamali Jayatilleke, Richard Jackson, Robert Stewart, Philip McGuire . The Lancet, Volume 383, Page S16, 26 February 2014