Characteristics of a suicide attempt predict who makes another attempt after hospital discharge: A decision-tree investigation

Diagnosis of mental disorders using machine learning: Literature review and bibliometric mapping from 2012 to 2023

Background

Mental disorders (MDs) are becoming a leading burden in non-communicable diseases (NCDs). As per the World Health Organization's 2022 assessment report, there was a steep increase of 25 % in MDs during the COVID-19 pandemic. Early diagnosis of MDs can significantly improve treatment outcome and save disability-adjusted life years (DALYs). In recent times, the application of machine learning (ML) and deep learning (DL)) has shown promising results in the diagnosis of MDs, and the field has witnessed a huge research output in the form of research publications. Therefore, a bibliometric mapping along with a review of recent advancements is required.

Methods

This study presents a bibliometric analysis and review of the research, published over the last 10 years. Literature searches were conducted in the Scopus database for the period from January 1, 2012, to June 9, 2023. The data was filtered and screened to include only relevant and reliable publications. A total of 2811 journal articles were found. The data was exported to a comma-separated value (CSV) format for further analysis. Furthermore, a review of 40 selected studies was performed.

Results

The popularity of ML techniques in diagnosing MDs has been growing, with an annual research growth rate of 17.05 %. The Journal of Affective Disorders published the most documents (n = 97), while Wang Y. (n = 64) has published the most articles. Lotka's law is observed, with a minority of authors contributing the majority of publications. The top affiliating institutes are the West China Hospital of Sichuan University followed by the University of California, with China and the US dominating the top 10 institutes. While China has more publications, papers affiliated with the US receive more citations. Depression and schizophrenia are the primary focuses of ML and deep learning (DL) in mental disease detection. Co-occurrence network analysis reveals that ML is associated with depression, schizophrenia, autism, anxiety, ADHD, obsessive-compulsive disorder, and PTSD. Popular algorithms include support vector machine (SVM) classifier, decision tree classifier, and random forest classifier. Furthermore, DL is linked to neuroimaging techniques such as MRI, fMRI, and EEG, as well as bipolar disorder. Current research trends encompass DL, LSTM, generalized anxiety disorder, feature fusion, and convolutional neural networks.

Machine learning and the prediction of suicide in psychiatric populations: a systematic review

Machine learning (ML) has emerged as a promising tool to enhance suicidal prediction. However, as many large-sample studies mixed psychiatric and non-psychiatric populations, a formal psychiatric diagnosis emerged as a strong predictor of suicidal risk, overshadowing more subtle risk factors specific to distinct populations. To overcome this limitation, we conducted a systematic review of ML studies evaluating suicidal behaviors exclusively in psychiatric clinical populations. A systematic literature search was performed from inception through November 17, 2022 on PubMed, EMBASE, and Scopus following the PRISMA guidelines. Original research using ML techniques to assess the risk of suicide or predict suicide attempts in the psychiatric population were included. An assessment for bias risk was performed using the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guidelines. About 1032 studies were retrieved, and 81 satisfied the inclusion criteria and were included for qualitative synthesis. Clinical and demographic features were the most frequently employed and random forest, support vector machine, and convolutional neural network performed better in terms of accuracy than other algorithms when directly compared. Despite heterogeneity in procedures, most studies reported an accuracy of 70% or greater based on features such as previous attempts, severity of the disorder, and pharmacological treatments. Although the evidence reported is promising, ML algorithms for suicidal prediction still present limitations, including the lack of neurobiological and imaging data and the lack of external validation samples. Overcoming these issues may lead to the development of models to adopt in clinical practice. Further research is warranted to boost a field that holds the potential to critically impact suicide mortality.

Free-Water Diffusion Magnetic Resonance Imaging Differentiates Suicidal Ideation From Suicide Attempt in Treatment-Resistant Depression

BACKGROUND

Suicide attempt is highly prevalent in treatment-resistant depression (TRD); however, the neurobiological profile of suicidal ideation versus suicide attempt is unclear. Neuroimaging methods including diffusion magnetic resonance imaging-based free-water imaging may identify neural correlates underlying suicidal ideation and attempts in individuals with TRD.

METHODS

Diffusion magnetic resonance imaging data were obtained from 64 male and female participants (mean age 44.5 ± 14.2 years), including 39 patients with TRD (n = 21 and lifetime history of suicidal ideation but no attempts [SI group]; n = 18 with lifetime history of suicide attempt [SA group]), and 25 age- and sex-matched healthy control participants. Depression and suicidal ideation severity were examined using clinician-rated and self-report measures. Whole-brain neuroimaging analysis was conducted using tract-based spatial statistics via FSL to identify differences in white matter microstructure in the SI versus SA groups and in patients versus control participants.

RESULTS

Free-water imaging revealed elevated axial diffusivity and extracellular free water in fronto-thalamo-limbic white matter tracts of the SA group compared with the SI group. In a separate comparison, patients with TRD had widespread reductions in fractional anisotropy and axial diffusivity, as well as elevated radial diffusivity compared with control participants (thresholded p < .05, familywise error corrected).

CONCLUSIONS

A unique neural signature consisting of elevated axial diffusivity and free water was identified in patients with TRD and suicide attempt history. Findings of reduced fractional anisotropy, axial diffusivity, and elevated radial diffusivity in patients versus control participants are consistent with previously published studies. Multimodal and prospective investigations are recommended to better understand biological correlates of suicide attempt in TRD.

Comparison of three machine learning models to predict suicidal ideation and depression among Chinese adolescents: A cross-sectional study

BACKGROUND

Machine learning (ML) algorithms based on various clinicodemographic, psychometric, and biographic factors have been used to predict depression, suicidal ideation, and suicide attempt in adolescents, but there is still a need for more accurate and efficient models for screening the general adolescent population. In this study, we compared various ML methods to identify a model that most accurately predicts suicidal ideation and level of depression in a large cohort of school-aged adolescents.

METHODS

Ten psychological scale scores and 20 sociodemographic parameters were collected from 10,243 Chinese adolescents in the first or second year of middle school and high school. These variables were then included in a random forest (RF) model, support vector machine (SVM) model, and decision tree model for factor screening, dichotomous prediction of suicidal ideation (yes/no), and trichotomous prediction of depression (no depression, mild-moderate depression, or major depression).

RESULTS

The RF model demonstrated greater accuracy for predicting suicidal ideation (mean accuracy (ACC) = 87.3 %, SD = 3.2 %, area under curve (AUC) = 92.4 %) and depressive status (ACC = 84.0 %, SD = 2.8 %, AUC = 90.1 %) than SVM and decision tree models. We have also used the RF model to predict adolescents with both depression and suicidal ideation with satisfactory results. Significant differences were found in several sociodemographic parameters and scale scores among classification groups and differences in six factors between sexes.

CONCLUSIONS

This RF model may prove valuable for predicting suicidal ideation, depression, and non-suicidal self-injury among the general population of Chinese adolescents.

Suicide and Self-Harm Risk Assessment: A Systematic Review of Prospective Research

OBJECTIVE

Suicide and self-harm are widespread yet underreported. Risk assessment is key to effective self-harm and suicide prevention and management. There is contradicting evidence regarding the effectiveness of risk assessment tools in predicting self-harm and suicide risk. This systematic review examines the effect of risk assessment strategies on predicting suicide and self-harm outcomes among adult healthcare service users.

METHOD

Electronic and gray literature databases were searched for prospective research. Studies were screened and selected by independent reviewers. Quality and level of evidence assessments were conducted. Due to study heterogeneity, we present a narrative synthesis under three categories: (1) suicide- and self-harm-related outcomes; (2) clinician assessment of suicide and self-harm risk; and (3) healthcare utilization due to self-harm or suicide.

RESULTS

Twenty-one studies were included in this review. The SAD PERSONS Scale was the most used tool. It outperformed the Beck Scale for Suicide Ideation in predicting hospital admissions and stay following suicide and self-harm, yet it failed to predict repeat suicide and self-harm and was not recommended for routine use. There were mixed findings relating to clinician risk assessment, with some studies recommending clinician assessment over structured tools, whilst others found that clinician assessment failed to predict future attempts and deaths.

CONCLUSIONS

There is insufficient evidence to support the use of any one tool, inclusive of clinician assessment of risk, for self-harm and suicidality. The discourse around risk assessment needs to move toward a broader discussion on the safety of patients who are at risk for self-harm and/or suicide.HIGHLIGHTSThere is insufficient evidence to support using standalone risk assessment tools.There are mixed findings relating to clinician assessment of risk.Structured professional judgment is widely accepted for risk assessment.

Machine Learning-Assisted Evaluation of Circulating DNA Quantitative Analysis for Cancer Screening

While the utility of circulating cell-free DNA (cfDNA) in cancer screening and early detection have recently been investigated by testing genetic and epigenetic alterations, here, an original approach by examining cfDNA quantitative and structural features is developed. First, the potential of cfDNA quantitative and structural parameters is independently demonstrated in cell culture, murine, and human plasma models. Subsequently, these variables are evaluated in a large retrospective cohort of 289 healthy individuals and 983 patients with various cancer types; after age resampling, this evaluation is done independently and the variables are combined using a machine learning approach. Implementation of a decision tree prediction model for the detection and classification of healthy and cancer patients shows unprecedented performance for 0, I, and II colorectal cancer stages (specificity, 0.89 and sensitivity, 0.72). Consequently, the methodological proof of concept of using both quantitative and structural biomarkers, and classification with a machine learning method are highlighted, as an efficient strategy for cancer screening. It is foreseen that the classification rate may even be improved by the addition of such biomarkers to fragmentomics, methylation, or the detection of genetic alterations. The optimization of such a multianalyte strategy with this machine learning method is therefore warranted.

Artificial Intelligence and Suicide Prevention: A Systematic Review of Machine Learning Investigations

Suicide is a leading cause of death that defies prediction and challenges prevention efforts worldwide. Artificial intelligence (AI) and machine learning (ML) have emerged as a means of investigating large datasets to enhance risk detection. A systematic review of ML investigations evaluating suicidal behaviors was conducted using PubMed/MEDLINE, PsychInfo, Web-of-Science, and EMBASE, employing search strings and MeSH terms relevant to suicide and AI. Databases were supplemented by hand-search techniques and Google Scholar. Inclusion criteria: (1) journal article, available in English, (2) original investigation, (3) employment of AI/ML, (4) evaluation of a suicide risk outcome. N = 594 records were identified based on abstract search, and 25 hand-searched reports. N = 461 reports remained after duplicates were removed, n = 316 were excluded after abstract screening. Of n = 149 full-text articles assessed for eligibility, n = 87 were included for quantitative synthesis, grouped according to suicide behavior outcome. Reports varied widely in methodology and outcomes. Results suggest high levels of risk classification accuracy (>90%) and Area Under the Curve (AUC) in the prediction of suicidal behaviors. We report key findings and central limitations in the use of AI/ML frameworks to guide additional research, which hold the potential to impact suicide on broad scale.

Predictors of relapse in high lethality suicide attempters: a six-month prospective study

BACKGROUND

This study is aimed to investigate the association between clinical, metabolic, inflammatory and environmental (photoperiod defined as daily sunlight exposure) parameters and suicide re-attempts after the index suicide attempt. Possible predictors of suicide re-attempts were also explored.

METHODS

Overall, 432 subjects with suicide attempts, of which 79 relapsed within the following six months were included in this prospective study. We adopted the Joiner's definition of suicide lethality, as "the acquired ability to enact lethal self-injury". The Cox regression was used to test the association between the mentioned variables and Kaplan-Meier plots showed the trend of suicide re-attempts.

RESULTS

Among participants, 30.8% committed a high-lethality suicide attempt. Cox regression confirmed the association between lifetime suicide attempts and number of suicide attempts in the study time-frame and suicide-reattempts. The longer photoperiod (Spring/Summer) was associated with suicide re-attempts, particularly patients with admission in June/July for the index event. Total cholesterol (TC), low-density lipo-protein cholesterol and c-reactive protein serum levels were significantly associated with suicide re-attempts but Cox regression confirmed only the association between lower TC serum levels and suicide re-attempt.

LIMITATIONS

Patients' seasonal environment, psychological factors, presence of acute life-events fostering the suicidal crisis and detailed medical history have been not investigated. Findings were derived from a single psychiatric unit.

CONCLUSIONS

Lifetime suicide attempts, higher number of previous suicide attempts, lower total cholesterol levels, and suicide attempt during longer photoperiod were significant predictors of suicide re-attempts. Further studies are needed in order to better characterize single- vs. multiple suicide attempter's profiles.