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.

Feasibility and preliminary efficacy of brief tele-psychotherapy for COVID-19 patients and their first-degree relatives

BACKGROUND

The SARS-CoV-2 pandemic compromised the mental health of COVID-19 patients and their family members. Due to social distancing and lockdown measures, a remote, tele-psychotherapy program for former or current COVID-19 patients and their relatives was implemented.

OBJECTIVE

The primary goal of this project was to evaluate intervention feasibility. The secondary aim was to assess the impact of the intervention by means of pre-post psychological changes.

METHODS

After a phone-based eligibility screening and remote neuropsychological testing, participants completed online self-reports assessing baseline COVID-related psychopathology. Next, participants attended eight tele-psychotherapy sessions. After treatment, the online self-reports were completed again.

RESULTS

Of 104 enrolled participants, 88 completed the intervention (84.6 % completion rate). Significant pre-post improvements were observed for generalized anxiety (d = 0.38), depression (d = 0.37), insomnia (d = 0.43), post-traumatic psychopathology (d = 0.54), and general malaise (d = 0.31). Baseline cluster analysis revealed a subgroup of 41 subjects (47.6 %) with no psychopathology, and a second subgroup of 45 subject (52.3 %) with moderate severity. Thirty-three percent of the second group reached full symptom remission, while 66 % remained symptomatic after treatment.

CONCLUSIONS

Remote brief tele-psychotherapy for COVID-19 patients and their first-degree relatives is feasible and preliminary efficacious at reducing COVID-related psychopathology in a subgroup of patients. Further research is needed to investigate distinct profiles of treatment response.

A machine learning approach on whole blood immunomarkers to identify an inflammation-associated psychosis onset subgroup

Psychosis onset is a transdiagnostic event that leads to a range of psychiatric disorders, which are currently diagnosed through clinical observation. The integration of multimodal biological data could reveal different subtypes of psychosis onset to target for the personalization of care. In this study, we tested the existence of subgroups of patients affected by first-episode psychosis (FEP) with a possible immunopathogenic basis. To do this, we designed a data-driven unsupervised machine learning model to cluster a sample of 127 FEP patients and 117 healthy controls (HC), based on the peripheral blood expression levels of 12 psychosis-related immune gene transcripts. To validate the model, we applied a resampling strategy based on the half-splitting of the total sample with random allocation of the cases. Further, we performed a post-hoc univariate analysis to verify the clinical, cognitive, and structural brain correlates of the subgroups identified. The model identified and validated two distinct clusters: 1) a FEP cluster characterized by the high expression of inflammatory and immune-activating genes (IL1B, CCR7, IL12A and CXCR3); 2) a cluster consisting of an equal number of FEP and HC subjects, which did not show a relative over or under expression of any immune marker (balanced subgroup). None of the subgroups was related to specific symptoms dimensions or longitudinal diagnosis of affective vs non-affective psychosis. FEP patients included in the balanced immune subgroup showed a thinning of the left supramarginal and superiorfrontal cortex (FDR-adjusted p-values < 0.05). Our results demonstrated the existence of a FEP patients' subgroup identified by a multivariate pattern of immunomarkers involved in inflammatory activation. This evidence may pave the way to sample stratification in clinical studies aiming to develop diagnostic tools and therapies targeting specific immunopathogenic pathways of psychosis.

Language production impairments in patients with a first episode of psychosis

Language production has often been described as impaired in psychiatric diseases such as in psychosis. Nevertheless, little is known about the characteristics of linguistic difficulties and their relation with other cognitive domains in patients with a first episode of psychosis (FEP), either affective or non-affective. To deepen our comprehension of linguistic profile in FEP, 133 patients with FEP (95 non-affective, FEP-NA; 38 affective, FEP-A) and 133 healthy controls (HC) were assessed with a narrative discourse task. Speech samples were systematically analyzed with a well-established multilevel procedure investigating both micro- (lexicon, morphology, syntax) and macro-linguistic (discourse coherence, pragmatics) levels of linguistic processing. Executive functioning and IQ were also evaluated. Both linguistic and neuropsychological measures were secondarily implemented with a machine learning approach in order to explore their predictive accuracy in classifying participants as FEP or HC. Compared to HC, FEP patients showed language production difficulty at both micro- and macro-linguistic levels. As for the former, FEP produced shorter and simpler sentences and fewer words per minute, along with a reduced number of lexical fillers, compared to HC. At the macro-linguistic level, FEP performance was impaired in local coherence, which was paired with a higher percentage of utterances with semantic errors. Linguistic measures were not correlated with any neuropsychological variables. No significant differences emerged between FEP-NA and FEP-A (p≥0.02, after Bonferroni correction). Machine learning analysis showed an accuracy of group prediction of 76.36% using language features only, with semantic variables being the most impactful. Such a percentage was enhanced when paired with clinical and neuropsychological variables. Results confirm the presence of language production deficits already at the first episode of the illness, being such impairment not related to other cognitive domains. The high accuracy obtained by the linguistic set of features in classifying groups support the use of machine learning methods in neuroscience investigations.

Factors Associated With Severity of Delirium Complicating COVID-19 in Intensive Care Units

The clinical outcome of the disease provoked by the SARS-CoV-2 infection, COVID-19, is largely due to the development of interstitial pneumonia accompanied by an Acute Respiratory Distress Syndrome (ARDS), often requiring ventilatory support therapy in Intensive Care Units (ICUs). Current epidemiologic evidence is demonstrating that the COVID-19 prognosis is significantly influenced by its acute complications. Among these, delirium figures as one of the most frequent and severe, especially in the emergency setting, where it shows a significantly negative prognostic impact. In this regard, the aim of our study is to identify clinical severity factors of delirium complicating COVID-19 related-ARDS. We performed a comparative and correlation analysis using demographics, comorbidities, multisystemic and delirium severity scores and anti-delirium therapy in two cohorts of ARDS patients with delirium, respectively, due to COVID-19 (n = 40) or other medical conditions (n = 39). Our results indicate that delirium in COVID-19-related ARDS is more severe since its onset despite a relatively less severe systemic condition at the point of ICU admission and required higher dosages of antipsychotic and non-benzodiazepinic sedative therapy respect to non-COVID patients. Finally, the correlation analysis showed a direct association between the male gender and maximum dosage of anti-delirium medications needed within the COVID-19 group, which was taken as a surrogate of delirium severity. Overall, our results seem to indicate that pathogenetic factors specifically associated to severe COVID-19 are responsible for the high severity of delirium, paving the way for future research focused on the mechanisms of the cognitive alterations associated with COVID-19.

Classification of Psychoses Based on Immunological Features: A Machine Learning Study in a Large Cohort of First-Episode and Chronic Patients

For several years, the role of immune system in the pathophysiology of psychosis has been well-recognized, showing differences from the onset to chronic phases. Our study aims to implement a biomarker-based classification model suitable for the clinical management of psychotic patients. A machine learning algorithm was used to classify a cohort of 362 subjects, including 160 first-episode psychosis patients (FEP), 70 patients affected by chronic psychiatric disorders (schizophrenia, bipolar disorder, and major depressive disorder) with psychosis (CRO) and 132 health controls (HC), based on mRNA transcript levels of 56 immune genes. Models distinguished between FEP, CRO, and HC and between the subgroup of drug-free FEP and HC with a mean accuracy of 80.8% and 90.4%, respectively. Interestingly, by using the feature importance method, we identified some immune gene transcripts that contribute most to the classification accuracy, possibly giving new insights on the immunopathogenesis of psychosis. Therefore, our results suggest that our classification model has a high translational potential, which may pave the way for a personalized management of psychosis.