Dataset Information

Deep neural networks detect suicide risk from textual facebook posts.

ABSTRACT: Detection of suicide risk is a highly prioritized, yet complicated task. Five decades of research have produced predictions slightly better than chance (AUCs?=?0.56-0.58). In this study, Artificial Neural Network (ANN) models were constructed to predict suicide risk from everyday language of social media users. The dataset included 83,292 postings authored by 1002 authenticated Facebook users, alongside valid psychosocial information about the users. Using Deep Contextualized Word Embeddings for text representation, two models were constructed: A Single Task Model (STM), to predict suicide risk from Facebook postings directly (Facebook texts???suicide) and a Multi-Task Model (MTM), which included hierarchical, multilayered sets of theory-driven risk factors (Facebook texts???personality traits???psychosocial risks???psychiatric disorders???suicide). Compared with the STM predictions (0.621???AUC???0.629), the MTM produced significantly improved prediction accuracy (0.697???AUC???0.746), with substantially larger effect sizes (0.729???d???0.936). Subsequent content analyses suggested that predictions did not rely on explicit suicide-related themes, but on a range of text features. The findings suggest that machine learning based analyses of everyday social media activity can improve suicide risk predictions and contribute to the development of practical detection tools.

SUBMITTER: Ophir Y

PROVIDER: S-EPMC7542168 | biostudies-literature | 2020 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Deep neural networks detect suicide risk from textual facebook posts.

Ophir Yaakov Y Tikochinski Refael R Asterhan Christa S C CSC Sisso Itay I Reichart Roi R

Scientific reports 20201007 1

Detection of suicide risk is a highly prioritized, yet complicated task. Five decades of research have produced predictions slightly better than chance (AUCs = 0.56-0.58). In this study, Artificial Neural Network (ANN) models were constructed to predict suicide risk from everyday language of social media users. The dataset included 83,292 postings authored by 1002 authenticated Facebook users, alongside valid psychosocial information about the users. Using Deep Contextualized Word Embeddings for ...[more]

PMID: 33028921

Similar Datasets

Project description:BackgroundEach year, approximately 800,000 people die by suicide worldwide, accounting for 1-2 in every 100 deaths. It is always a tragic event with a huge impact on family, friends, the community and health professionals. Unfortunately, suicide prevention and the development of risk assessment tools have been hindered by the complexity of the underlying mechanisms and the dynamic nature of a person's motivation and intent. Many of those who die by suicide had contact with health services in the preceding year but identifying those most at risk remains a challenge.ObjectiveTo explore the feasibility of using artificial neural networks with routinely collected electronic health records to support the identification of those at high risk of suicide when in contact with health services.MethodsUsing the Secure Anonymised Information Linkage Databank UK, we extracted the data of those who died by suicide between 2001 and 2015 and paired controls. Looking at primary (general practice) and secondary (hospital admissions) electronic health records, we built a binary feature vector coding the presence of risk factors at different times prior to death. Risk factors included: general practice contact and hospital admission; diagnosis of mental health issues; injury and poisoning; substance misuse; maltreatment; sleep disorders; and the prescription of opiates and psychotropics. Basic artificial neural networks were trained to differentiate between the suicide cases and paired controls. We interpreted the output score as the estimated suicide risk. System performance was assessed with 10x10-fold repeated cross-validation, and its behavior was studied by representing the distribution of estimated risk across the cases and controls, and the distribution of factors across estimated risks.ResultsWe extracted a total of 2604 suicide cases and 20 paired controls per case. Our best system attained a mean error rate of 26.78% (SD 1.46; 64.57% of sensitivity and 81.86% of specificity). While the distribution of controls was concentrated around estimated risks < 0.5, cases were almost uniformly distributed between 0 and 1. Prescription of psychotropics, depression and anxiety, and self-harm increased the estimated risk by ~0.4. At least 95% of those presenting these factors were identified as suicide cases.ConclusionsDespite the simplicity of the implemented system, the proposed methodology obtained an accuracy like other published methods based on specialized questionnaire generated data. Most of the errors came from the heterogeneity of patterns shown by suicide cases, some of which were identical to those of the paired controls. Prescription of psychotropics, depression and anxiety, and self-harm were strongly linked with higher estimated risk scores, followed by hospital admission and long-term drug and alcohol misuse. Other risk factors like sleep disorders and maltreatment had more complex effects.

Project description:BackgroundSubstance use among young pregnant women is a common and significant public health concern associated with a number of adverse outcomes for both mothers and infants. Social media posts by young women can provide valuable, real-world insight into their perceptions of substance use immediately before and during pregnancy.ObjectiveThe aim of this study was to characterize the frequency and content of posts regarding substance use in the year before pregnancy and during pregnancy among young mothers.MethodsFacebook posts were mined from young pregnant women (age, 16-24 years) who consented from 2 Midwest primary care clinics that serve a predominantly low-income community. Natural language processing was used to identify posts related to substance use by keyword searching (eg, drunk, drugs, pot, and meth). Using mixed-methods techniques, 2 investigators iteratively coded and identified major themes around substance use from these mined Facebook posts. Outcome measures include the frequency of posts and major themes expressed regarding substance use before and during pregnancy.ResultsWomen in our sample (N=43) had a mean age of 21 (SD 2.3) years, and the largest subgroup (21/43, 49%) identified as non-Hispanic black; 26% (11/43) identified as non-Hispanic white; 16% (7/43) as Hispanic; and 9% (4/43) as non-Hispanic mixed race, Native American, or other. The largest subgroup (20/43, 47%) graduated high school without further education, while 30% (13/43) completed only some high school and 23% (10/43) completed at least some postsecondary education. Young women discussed substance use on social media before and during pregnancy, although compared with the year before pregnancy, the average frequency of substance-related posts during pregnancy decreased. Themes identified included craving alcohol or marijuana, social use of alcohol or marijuana, reasons for abstaining from substance use, and intoxication.ConclusionsFacebook posts reveal that young pregnant women discuss the use of substances, predominantly alcohol and marijuana. Future work can explore clinical opportunities to prevent and treat substance use before and during pregnancy among young, at-risk mothers.

Project description:BackgroundEffective interventions aimed at correcting COVID-19 vaccine misinformation, known as fact-checking messages, are needed to combat the mounting antivaccine infodemic and alleviate vaccine hesitancy.ObjectiveThis work investigates (1) the changes in the public's attitude toward COVID-19 vaccines over time, (2) the effectiveness of COVID-19 vaccine fact-checking information on social media engagement and attitude change, and (3) the emotional and linguistic features of the COVID-19 vaccine fact-checking information ecosystem.MethodsWe collected a data set of 12,553 COVID-19 vaccine fact-checking Facebook posts and their associated comments (N=122,362) from January 2020 to March 2022 and conducted a series of natural language processing and statistical analyses to investigate trends in public attitude toward the vaccine in COVID-19 vaccine fact-checking posts and comments, and emotional and linguistic features of the COVID-19 fact-checking information ecosystem.ResultsThe percentage of fact-checking posts relative to all COVID-19 vaccine posts peaked in May 2020 and then steadily decreased as the pandemic progressed (r=-0.92, df=21, t=-10.94, 95% CI -0.97 to -0.82, P<.001). The salience of COVID-19 vaccine entities was significantly lower in comments (mean 0.03, SD 0.03, t=39.28, P<.001) than in posts (mean 0.09, SD 0.11). Third-party fact checkers have been playing a more important role in more fact-checking over time (r=0.63, df=25, t=4.06, 95% CI 0.33-0.82, P<.001). COVID-19 vaccine fact-checking posts continued to be more analytical (r=0.81, df=25, t=6.88, 95% CI 0.62-0.91, P<.001) and more confident (r=0.59, df=25, t=3.68, 95% CI 0.27-0.79, P=.001) over time. Although comments did not exhibit a significant increase in confidence over time, tentativeness in comments significantly decreased (r=-0.62, df=25, t=-3.94, 95% CI -0.81 to -0.31, P=.001). In addition, although hospitals receive less engagement than other information sources, the comments expressed more positive attitudinal valence in comments compared to other information sources (b=0.06, 95% CI 0.00-0.12, t=2.03, P=.04).ConclusionsThe percentage of fact-checking posts relative to all posts about the vaccine steadily decreased after May 2020. As the pandemic progressed, third-party fact checkers played a larger role in posting fact-checking COVID-19 vaccine posts. COVID-19 vaccine fact-checking posts continued to be more analytical and more confident over time, reflecting increased confidence in posts. Similarly, tentativeness in comments decreased; this likewise suggests that public uncertainty diminished over time. COVID-19 fact-checking vaccine posts from hospitals yielded more positive attitudes toward vaccination than other information sources. At the same time, hospitals received less engagement than other information sources. This suggests that hospitals should invest more in generating engaging public health campaigns on social media.

Project description:PurposeTo determine whether convolutional neural networks can detect morphological differences between images of microbiologically positive and negative corneal ulcers.MethodsA cross-sectional comparison of prospectively collected data consisting of bacterial and fungal cultures and smears from eyes with acute infectious keratitis at Aravind Eye Hospital. Two convolutional neural network architectures (DenseNet and MobileNet) were trained using images obtained from handheld cameras collected from culture-positive and negative images and smear-positive and -negative images. Each architecture was trained on two image sets: (1) one with labels assigned using only culture results and (2) one using culture and smear results. The outcome measure was area under the receiver operating characteristic curve for predicting whether an ulcer would be microbiologically positive or negative.ResultsThere were 1970 images from 886 patients were included. None of the models were better than random chance at predicting positive microbiologic results (area under the receiver operating characteristic curve ranged from 0.49 to 0.56; all confidence intervals included 0.5).ConclusionsThese two state-of-the-art deep convolutional neural network architectures could not reliably predict whether a corneal ulcer would be microbiologically positive or negative based on clinical photographs. This absence of detectable morphological differences informs the future development of computer vision models trained to predict the causative agent in infectious keratitis using corneal photography.Translational relevanceThese deep learning models were not able to identify morphological differences between microbiologically positive and negative corneal ulcers. This finding suggests that similar artificial intelligence models trained to identify the causative pathogen using only microbiologically positive cases may have potential to generalize well, including to cases with falsely negative microbiologic testing.

Dataset Information

Deep neural networks detect suicide risk from textual facebook posts.

Publications

Deep neural networks detect suicide risk from textual facebook posts.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets