A lexicon-based approach to detecting suicide-related messages on Twitter

Predicting patients' sentiments about medications using artificial intelligence techniques

The increasing development of technology has led to the increase of digital data in various fields, such as medication-related texts. Sentiment Analysis (SA) in medication is essential to give clinicians insights into patients' feedback about the treatment procedure. Therefore, this study intends to develop Artificial Intelligence (AI) models to predict patients' sentiments. This study used a large medication review dataset to perform a SA of medications. Three scenarios were considered for classification, including two, three, and ten classes. The Word2Vec algorithm and pre-trained word embeddings, including the general and clinical domains, were utilized in model development. Seven Machine Learning (ML) and Deep Learning (DL) models were developed for various scenarios. The best hyperparameters for all models were fine-tuned. Moreover, two ensemble learning models were developed from the proposed ML and DL models. For the first time, a technique was implemented to interpret the results for explainability and interpretability. The results showed that the developed deep ensemble model (DL_ENS), using PubMed and PMC, as pre-trained word embedding representation, achieved the best results, with accuracy and F1-Score of 92.96% and 92.27% in two classes, 92.18% and 88.50 in three classes, and 90.31% and 67.07% in ten classes, respectively. Combining DL models and developing a DL_ENS with clinical domain pre-trained word embedding representation can accurately predict classes and scores of patients' sentiments about medications compared to previous studies on the same dataset. Due to the transparency in decision-making, our DL_ENS model can be used as an auxiliary tool to help clinicians prescribe medications.

Enhancing Suicide Attempt Risk Prediction Models with Temporal Clinical Note Features

OBJECTIVES

 The objective of this study was to investigate the impact of enhancing a structured-data-based suicide attempt risk prediction model with temporal Concept Unique Identifiers (CUIs) derived from clinical notes. We aimed to examine how different temporal schemes, model types, and prediction ranges influenced the model's predictive performance. This research sought to improve our understanding of how the integration of temporal information and clinical variable transformation could enhance model predictions.

METHODS

 We identified modeling targets using diagnostic codes for suicide attempts within 30, 90, or 365 days following a temporally grouped visit cluster. Structured data included medications, diagnoses, procedures, and demographics, whereas unstructured data consisted of terms extracted with regular expressions from clinical notes. We compared models trained only on structured data (controls) to hybrid models trained on both structured and unstructured data. We used two temporalization schemes for clinical notes: fixed 90-day windows and flexible epochs. We trained and assessed random forests and hybrid long short-term memory (LSTM) neural networks using area under the precision recall curve (AUPRC) and area under the receiver operating characteristic, with additional evaluation of sensitivity and positive predictive value at 95% specificity.

RESULTS

 The training set included 2,364,183 visit clusters with 2,009 30-day suicide attempts, and the testing set contained 471,936 visit clusters with 480 suicide attempts. Models trained with temporal CUIs outperformed those trained with only structured data. The window-temporalized LSTM model achieved the highest AUPRC (0.056 ± 0.013) for the 30-day prediction range. Hybrid models generally showed better performance compared with controls across most metrics.

CONCLUSION

 This study demonstrated that incorporating electronic health record-derived clinical note features enhanced suicide attempt risk prediction models, particularly with window-temporalized LSTM models. Our results underscored the critical value of unstructured data in suicidality prediction, aligning with previous findings. Future research should focus on integrating more sophisticated methods to continue improving prediction accuracy, which will enhance the effectiveness of future intervention.

Resilience and social support as protective factors against suicidal ideation among tertiary students during COVID-19: a cross-sectional study

BACKGROUND

Suicidal ideation is a depression symptom which represents a key (cognitive) component of suicidality and plays an important role in suicide risk detection, intervention, and prevention. Despite existing research showing the importance of certain factors of depression symptoms and suicidal ideation, less is known about the interaction between the various risk and protective factors. The aim of the study was to examine whether living conditions characteristics and personal circumstances during the COVID-19 pandemic predicted the presence of depression symptoms and suicidal ideation among tertiary students and whether resilience and social support can mitigate the detrimental effects of difficult life circumstances.

METHOD

A large online cross-sectional study was conducted in March 2021 among 4,645 Slovenian tertiary students. Hierarchical multiple regression and hierarchical logistic regression methods were used to assess and compare the effect of life circumstances variables, as opposed to resilience and social support, on depression symptoms and suicidal ideation.

RESULTS

Female gender, single relationship status, living alone, a higher degree of household conflict, having a history of mental illness and chronic disease diagnosis were significant predictors of depression scores. All but gender were also predictors of suicidal ideation. Household conflict and a history of mental illness were the factors showing the strongest effect in both cases. On the other hand, social support and, in particular, resilience proved to be strong protective factors against depression symptoms and suicidal ideation. After accounting for one's resilience and social support, the explained variance in depression scores was more than doubled, while the harmful effect of household conflict and history of mental illness significantly decreased.

CONCLUSIONS

The findings stress the importance of one's resilience and social support and explain why some people manage to maintain mental well-being despite finding themselves in difficult life circumstances, which was the case for many tertiary students during the COVID-19 pandemic. These insights may inform preventive efforts against developing suicidal ideation and may be used as support for the design and implementation of interventions for improving resilience and social support from childhood onward.

Predicting Suicidal Ideation, Planning, and Attempts among the Adolescent Population of the United States

Suicide is the second leading cause of death among individuals aged 5 to 24 in the United States (US). However, the precursors to suicide often do not surface, making suicide prevention challenging. This study aims to develop a machine learning model for predicting suicide ideation (SI), suicide planning (SP), and suicide attempts (SA) among adolescents in the US during the coronavirus pandemic. We used the 2021 Adolescent Behaviors and Experiences Survey Data. Class imbalance was addressed using the proposed data augmentation method tailored for binary variables, Modified Synthetic Minority Over-Sampling Technique. Five different ML models were trained and compared. SHapley Additive exPlanations analysis was conducted for explainability. The Logistic Regression model, identified as the most effective, showed superior performance across all targets, achieving high scores in recall: 0.82, accuracy: 0.80, and area under the Receiver Operating Characteristic curve: 0.88. Variables such as sad feelings, hopelessness, sexual behavior, and being overweight were noted as the most important predictors. Our model holds promise in helping health policymakers design effective public health interventions. By identifying vulnerable sub-groups within regions, our model can guide the implementation of tailored interventions that facilitate early identification and referral to medical treatment.

Identifying X (Formerly Twitter) Posts Relevant to Dementia and COVID-19: Machine Learning Approach

BACKGROUND

During the pandemic, patients with dementia were identified as a vulnerable population. X (formerly Twitter) became an important source of information for people seeking updates on COVID-19, and, therefore, identifying posts (formerly tweets) relevant to dementia can be an important support for patients with dementia and their caregivers. However, mining and coding relevant posts can be daunting due to the sheer volume and high percentage of irrelevant posts.

OBJECTIVE

The objective of this study was to automate the identification of posts relevant to dementia and COVID-19 using natural language processing and machine learning (ML) algorithms.

METHODS

We used a combination of natural language processing and ML algorithms with manually annotated posts to identify posts relevant to dementia and COVID-19. We used 3 data sets containing more than 100,000 posts and assessed the capability of various algorithms in correctly identifying relevant posts.

RESULTS

Our results showed that (pretrained) transfer learning algorithms outperformed traditional ML algorithms in identifying posts relevant to dementia and COVID-19. Among the algorithms tested, the transfer learning algorithm A Lite Bidirectional Encoder Representations from Transformers (ALBERT) achieved an accuracy of 82.92% and an area under the curve of 83.53%. ALBERT substantially outperformed the other algorithms tested, further emphasizing the superior performance of transfer learning algorithms in the classification of posts.

CONCLUSIONS

Transfer learning algorithms such as ALBERT are highly effective in identifying topic-specific posts, even when trained with limited or adjacent data, highlighting their superiority over other ML algorithms and applicability to other studies involving analysis of social media posts. Such an automated approach reduces the workload of manual coding of posts and facilitates their analysis for researchers and policy makers to support patients with dementia and their caregivers and other vulnerable populations.

A machine learning approach to detect potentially harmful and protective suicide-related content in broadcast media

Suicide-related media content has preventive or harmful effects depending on the specific content. Proactive media screening for suicide prevention is hampered by the scarcity of machine learning approaches to detect specific characteristics in news reports. This study applied machine learning to label large quantities of broadcast (TV and radio) media data according to media recommendations reporting suicide. We manually labeled 2519 English transcripts from 44 broadcast sources in Oregon and Washington, USA, published between April 2019 and March 2020. We conducted a content analysis of media reports regarding content characteristics. We trained a benchmark of machine learning models including a majority classifier, approaches based on word frequency (TF-IDF with a linear SVM) and a deep learning model (BERT). We applied these models to a selection of more simple (e.g., focus on a suicide death), and subsequently to putatively more complex tasks (e.g., determining the main focus of a text from 14 categories). Tf-idf with SVM and BERT were clearly better than the naive majority classifier for all characteristics. In a test dataset not used during model training, F1-scores (i.e., the harmonic mean of precision and recall) ranged from 0.90 for celebrity suicide down to 0.58 for the identification of the main focus of the media item. Model performance depended strongly on the number of training samples available, and much less on assumed difficulty of the classification task. This study demonstrates that machine learning models can achieve very satisfactory results for classifying suicide-related broadcast media content, including multi-class characteristics, as long as enough training samples are available. The developed models enable future large-scale screening and investigations of broadcast media.

Pediatric Cancer Communication on Twitter: Natural Language Processing and Qualitative Content Analysis

BACKGROUND

During the COVID-19 pandemic, Twitter (recently rebranded as "X") was the most widely used social media platform with over 2 million cancer-related tweets. The increasing use of social media among patients and family members, providers, and organizations has allowed for novel methods of studying cancer communication.

OBJECTIVE

This study aimed to examine pediatric cancer-related tweets to capture the experiences of patients and survivors of cancer, their caregivers, medical providers, and other stakeholders. We assessed the public sentiment and content of tweets related to pediatric cancer over a time period representative of the COVID-19 pandemic.

METHODS

All English-language tweets related to pediatric cancer posted from December 11, 2019, to May 7, 2022, globally, were obtained using the Twitter application programming interface. Sentiment analyses were computed based on Bing, AFINN, and NRC lexicons. We conducted a supplemental nonlexicon-based sentiment analysis with ChatGPT (version 3.0) to validate our findings with a random subset of 150 tweets. We conducted a qualitative content analysis to manually code the content of a random subset of 800 tweets.

RESULTS

A total of 161,135 unique tweets related to pediatric cancer were identified. Sentiment analyses showed that there were more positive words than negative words. Via the Bing lexicon, the most common positive words were support, love, amazing, heaven, and happy, and the most common negative words were grief, risk, hard, abuse, and miss. Via the NRC lexicon, most tweets were categorized under sentiment types of positive, trust, and joy. Overall positive sentiment was consistent across lexicons and confirmed with supplemental ChatGPT (version 3.0) analysis. Percent agreement between raters for qualitative coding was 91%, and the top 10 codes were awareness, personal experiences, research, caregiver experiences, patient experiences, policy and the law, treatment, end of life, pharmaceuticals and drugs, and survivorship. Qualitative content analysis showed that Twitter users commonly used the social media platform to promote public awareness of pediatric cancer and to share personal experiences with pediatric cancer from the perspective of patients or survivors and their caregivers. Twitter was frequently used for health knowledge dissemination of research findings and federal policies that support treatment and affordable medical care.

CONCLUSIONS

Twitter may serve as an effective means for researchers to examine pediatric cancer communication and public sentiment around the globe. Despite the public mental health crisis during the COVID-19 pandemic, overall sentiments of pediatric cancer-related tweets were positive. Content of pediatric cancer tweets focused on health and treatment information, social support, and raising awareness of pediatric cancer.

Lexicon-Based Sentiment Analysis in Behavioral Research

A complete science of human behavior requires a comprehensive account of the verbal behavior those humans exhibit. Existing behavioral theories of such verbal behavior have produced compelling insight into language's underlying function, but the expansive program of research those theories deserve has unfortunately been slow to develop. We argue that the status quo's manually implemented and study-specific coding systems are too resource intensive to be worthwhile for most behavior analysts. These high input costs in turn discourage research on verbal behavior overall. We propose lexicon-based sentiment analysis as a more modern and efficient approach to the study of human verbal products, especially naturally occurring ones (e.g., psychotherapy transcripts, social media posts). In the present discussion, we introduce the reader to principles of sentiment analysis, highlighting its usefulness as a behavior analytic tool for the study of verbal behavior. We conclude with an outline of approaches for handling some of the more complex forms of speech, like negation, sarcasm, and speculation. The appendix also provides a worked example of how sentiment analysis could be applied to existing questions in behavior analysis, complete with code that readers can incorporate into their own work.

The hidden depths of suicidal discourse: Network analysis and natural language processing unmask uncensored expression

Background

The socially unattractive and stigmatizing nature of suicidal thought and behavior (STB) makes it especially susceptible to censorship across most modern digital communication platforms. The ubiquitous integration of technology with day-to-day life has presented an invaluable opportunity to leverage unprecedented amounts of data to study STB, yet the complex etiologies and consequences of censorship for research within mainstream online communities render an incomplete picture of STB manifestation. Analyses targeting online written content of suicidal users in environments where fear of reproach is mitigated may provide novel insight into modern trends and signals of STB expression.

Methods

Complete written content of N = 192 users, including n = 48 identified as potential suicide completers/highest-risk users (HRUs), on the pro-choice suicide forum, Sanctioned Suicide, was modeled using a combination of lexicon-based topic modeling (EMPATH) and exploratory network analysis techniques to characterize and highlight prominent aspects of censorship-free suicidal discourse.

Results

Modeling of over 2 million tokens across 37,136 forum posts found higher frequency of positive emotion and optimism among HRUs, emphasis on methods seeking and sharing behaviors, prominence of previously undocumented jargon, and semantics related to loneliness and life adversity.

Conclusion

This natural language processing (NLP)- and network-driven exposé of online STB subculture uncovered trends that deserve further attention within suicidology as they may be able to bolster detection, intervention, and prevention of suicidal outcomes and exposures.

Identifying suicide ideation in mental health application posts: A random forest algorithm

The growing use of digitized mental health applications requires new reliable early screening tools to identify user suicide risk. We used a lexicon-based random forest machine learning algorithm to predict suicide ideation scores from 714 online community text posts from December 2019 to April 2020. We validated predicted scores against expert-rated suicide ideation scores. The algorithm-predicted scores offered high validity and a low error rate and correctly identified 95% of expert-rated high-risk suicide ideation posts. Our findings highlight a potential new method to detect suicidal ideation of digital mental health application users.

A non-invasive machine learning mechanism for early disease recognition on Twitter: The case of anemia

Social media sites, such as Twitter, provide the means for users to share their stories, feelings, and health conditions during the disease course. Anemia, the most common type of blood disorder, is recognized as a major public health problem all over the world. Yet very few studies have explored the potential of recognizing anemia from online posts. This study proposed a novel mechanism for recognizing anemia based on the associations between disease symptoms and patients' emotions posted on the Twitter platform. We used k-means and Latent Dirichlet Allocation (LDA) algorithms to group similar tweets and to identify hidden disease topics. Both disease emotions and symptoms were mapped using the Apriori algorithm. The proposed approach was evaluated using a number of classifiers. A higher prediction accuracy of 98.96 % was achieved using Sequential Minimal Optimization (SMO). The results revealed that fear and sadness emotions are dominant among anemic patients. The proposed mechanism is the first of its kind to diagnose anemia using textual information posted on social media sites. It can advance the development of intelligent health monitoring systems and clinical decision-support systems.

Mining voices from self-expressed messages on social-media: Diagnostics of mental distress during COVID-19

The COVID-19 pandemic has had a severe impact on mankind, causing physical suffering and deaths across the globe. Even those who have not contracted the virus have experienced its far-reaching impacts, particularly on their mental health. The increased incidences of psychological problems, anxiety associated with the infection, social restrictions, economic downturn, etc., are likely to aggravate with the virus spread and leave a longer impact on humankind. These reasons in aggregation have raised concerns on mental health and created a need to identify novel precursors of depression and suicidal tendencies during COVID-19. Identifying factors affecting mental health and causing suicidal ideation is of paramount importance for timely intervention and suicide prevention. This study, thus, bridges this gap by utilizing computational intelligence and Natural Language Processing (NLP) to unveil the factors underlying mental health issues. We observed that the pandemic and subsequent lockdown anxiety emerged as significant factors leading to poor mental health outcomes after the onset of COVID-19. Consistent with previous works, we found that psychological disorders have remained pre-eminent. Interestingly, financial burden was found to cause suicidal ideation before the pandemic, while it led to higher odds of depressive (non-suicidal) thoughts for individuals who lost their jobs. This study offers significant implications for health policy makers, governments, psychiatric practitioners, and psychologists.

Natural language processing applied to mental illness detection: a narrative review

Mental illness is highly prevalent nowadays, constituting a major cause of distress in people’s life with impact on society’s health and well-being. Mental illness is a complex multi-factorial disease associated with individual risk factors and a variety of socioeconomic, clinical associations. In order to capture these complex associations expressed in a wide variety of textual data, including social media posts, interviews, and clinical notes, natural language processing (NLP) methods demonstrate promising improvements to empower proactive mental healthcare and assist early diagnosis. We provide a narrative review of mental illness detection using NLP in the past decade, to understand methods, trends, challenges and future directions. A total of 399 studies from 10,467 records were included. The review reveals that there is an upward trend in mental illness detection NLP research. Deep learning methods receive more attention and perform better than traditional machine learning methods. We also provide some recommendations for future studies, including the development of novel detection methods, deep learning paradigms and interpretable models.