Can cognition help predict suicide risk in patients with major depressive disorder? A machine learning study

Abstract

Background Previous studies suggest that deficits in cognition may increase the risk of suicide. Our study aims to develop a machine learning (ML) algorithm-based suicide risk prediction model using cognition in patients with major depressive disorder (MDD). Methods Participants comprised 52 depressed suicide attempters (DSA) and 61 depressed non-suicide attempters (DNS), and 98 healthy controls (HC). All participants were required to complete a series of questionnaires, the Suicide Stroop Task (SST) and the Iowa Gambling Task (IGT). The performance in IGT was analyzed using repeated measures ANOVA. ML with extreme gradient boosting (XGBoost) classification algorithm and locally explanatory techniques assessed performance and relative importance of characteristics for predicting suicide attempts. Prediction performances were compared with the area under the curve (AUC), decision curve analysis (DCA), and net reclassification improvement (NRI). Results DSA and DNS preferred to select the card from disadvantageous decks (decks "A" + "B") under risky situation (p = 0.023) and showed a significantly poorer learning effect during the IGT (F = 2.331, p = 0.019) compared with HC. Performance of XGBoost model based on demographic and clinical characteristics was compared with that of the model created after adding cognition data (AUC, 0.779 vs. 0.819, p > 0.05). The net benefit of model was improved and cognition resulted in continuous reclassification improvement with NRI of 5.3%. Several clinical dimensions were significant predictors in the XGBoost classification algorithm. Limitations A limited sample size and failure to include sufficient suicide risk factors in the predictive model. Conclusion This study demonstrate that cognitive deficits may serve as an important risk factor to predict suicide attempts in patients with MDD. Combined with other demographic characteristics and attributes drawn from clinical questionnaires, cognitive function can improve the predictive effectiveness of the ML model. Additionally, explanatory ML models can help clinicians detect specific risk factors for each suicide attempter within MDD patients. These findings may be helpful for clinicians to detect those at high risk of suicide attempts quickly and accurately, and help them make proactive treatment decisions.

Role of machine learning algorithms in suicide risk prediction: a systematic review-meta analysis of clinical studies

“…The AUC values in the range of 0.97 for XGBoost model suggest that the ML models had a high degree of accuracy in classifying individuals with respect to their risk of suicidal attempts. The findings of this study are consistent with previous research conducted by [ 36 ] which confirmed acceptable performance of XGBoost algorithm in cognition of patients with major depressive disorder. This result may be due to the fact that XGBoost is an ensemble model that constructs various models to reduce classification errors on each iteration.…”

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

“…74 MDD suicide ideation vs NS Acc: 0.75 AUC: 0.78 Sens: 0.78 Spec: 0.74MDD attempts vs ideation Acc: 0.68 AUC: 0.74 Sens: 0.68 Spec: 0.71 The features that contributed to stratifying MDD patients with diverse suicide risk levels mainly involved the visual-related and DMN-related inter-network connectivity within the weakly connected state. Zheng et al, [ 108 ] 52 MDD with suicidal attemps (40/12); 61 MDD without suicidal attempts (36/25); 98 HC (49/49) MDD Not specified XBoost Sociodemographic, clinical and cognitive features (total: 20 features) Suicide attempts Acc: 0.71 AUC: 0.82 Sens: 0.6 Spec: 0.79 PPV: 0.69 NPV: 0.71 Adding cognitive information significantly increased model prediction; the most important feature was HAMD-24 score Shin et al, [ 70 ] 83 MDD (64/19); 83 HC (69/14) MDD Not specified Naive Bayes classifier (5-folds CV) Sociodemographic and text-based High vs low-risk suicide (based on the MINI interview) Acc: 0.75 AUC: 0.80 Sens: 0.82 Spec: 0.65 When predicting suicide, only the ensemble analyses (namely, sociodemographic + text) resulted in significant prediction. Demographic alone: AUC 0.5 Text alone: AUC 0.64 Miranda et al, [ 36 ] 38807 PTSD patients PTSD Not specified RNN EMRs, including sociodemographic, clinical and lab features (>100 features) Suicide-related events within 3 months AUC:0.92 Lab tests (i.e., glucose, glucose urine, chloride, hemoglobin (HGB), hematocrit, mean corpuscular volume, white blood cell, neutrophils, potassium, INR, calcium, mean platelet volume) combined with medications and diagnoses can enhance the prediction of suicide in PTSD patients.…”

Testing the bipolar assumption of Singer-Loomis Type Deployment Inventory for Korean adults using classification and multidimensional scaling

“…Several kinds of frameworks can predict individual differences in traits and behavior based on neuroimaging data, such as the multivariate-prediction method using support vector regression ( Harris et al, 1996 ). Along with the intersubject network similarity approach ( Liu et al, 2019 ), many studies have been suggesting the use of the multivariate technique in identifying the participants in terms of not only predicting behaviors from brain connectivity patterns but also in pure behavioral ( Kim et al, 2022 , Jang, in press), psychosocial profiles ( Ding, 2006 ), or clinical data, such as PTSD ( Rosellini et al, 2018 ; Shim et al, 2019 ; Wshah et al, 2019 ), anxiety ( Carpenter et al, 2016 ; Anugraha and Vineetha, 2018 ; Muhammad et al, 2020 ; Xing et al, 2020 ), depression ( Priya et al, 2020 ; Asare et al, 2021 ; Su et al, 2021 ; Yang et al, 2022 ), and suicide ( Walsh et al, 2017 ; Littlefield et al, 2021 ; Schafer et al, 2021 ; Zheng et al, 2022 ). Usually, individual differences in scales are represented by a single score, the average performance, or with total scores of subscales.…”