Abstract
Machine learning analysis of social media data represents a promising way to capture longitudinal environmental influences contributing to individual risk for suicidal thoughts and behaviors. Our objective was to generate an algorithm termed “Suicide Artificial Intelligence Prediction Heuristic (SAIPH)” capable of predicting future risk to suicidal thought by analyzing publicly available Twitter data. We trained a series of neural networks on Twitter data queried against suicide associated psychological constructs including burden, stress, loneliness, hopelessness, insomnia, depression, and anxiety. Using 512,526 tweets from N = 283 suicidal ideation (SI) cases and 3,518,494 tweets from 2655 controls, we then trained a random forest model using neural network outputs to predict binary SI status. The model predicted N = 830 SI events derived from an independent set of 277 suicidal ideators relative to N = 3159 control events in all non-SI individuals with an AUC of 0.88 (95% CI 0.86–0.90). Using an alternative approach, our model generates temporal prediction of risk such that peak occurrences above an individual specific threshold denote a ~7 fold increased risk for SI within the following 10 days (OR = 6.7 ± 1.1, P = 9 × 10−71). We validated our model using regionally obtained Twitter data and observed significant associations of algorithm SI scores with county-wide suicide death rates across 16 days in August and in October, 2019, most significantly in younger individuals. Algorithmic approaches like SAIPH have the potential to identify individual future SI risk and could be easily adapted as clinical decision tools aiding suicide screening and risk monitoring using available technologies.
Highlights
Suicide has been an intractable public health problem despite advances in the diagnosis and treatment of major mental disorders[1]
Support vector machines (SVMs) trained on the same bag of words as the neural networks and evaluated in the same manner appeared to be more specific in certain instances such as in evaluating sleep and loneliness (Supplementary Fig. 1); AUC levels generated by support vector machines (SVMs) were significantly lower than that of neural networks (Student’s T, Mean NN: 0.68 ± 0.11, Mean SVM = 0.63 ± 0.12, p = 0.014) (Supplementary Fig. 2)
Using a large data set of social media content collected over two years, we generated a series of neural networks to first generate proxy indicators of psychological state based on text that subsequently feed into random forest models to classify an individual as at risk for suicidal ideation (SI) and those with a past history of suicide attempt (SA) or suicide plan from ideators
Summary
Suicide has been an intractable public health problem despite advances in the diagnosis and treatment of major mental disorders[1]. Previous studies have shown that youth are likely to disclose suicidal thoughts and suicidal risk factors online and on social media. A study examining emergency room assessments of suicidality found that adolescents were likely to report suicidal ideations verbally, and via electronic means, which included posts on social networking sites, blog posts, instant messages, text messages, and emails[3]. The authors reported an increase in the number of electronic communications of suicidality over the four year study-period, suggesting that this mode for expression of distress may become more common. Online expression of distress and suicidality may not be disclosed to physicians[4,5]. We develop a machine learning approach based on Twitter data that predicts individual level future suicidal risk based on online social media data prior to any mention of suicidal thought. We expand to a population level for validation and demonstrate that regional suicide rate data can be modeled by algorithmic scoring of randomly sampled Twitter data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
