Detailed contribution information
| Contribution title | Imminent Suicide Risk Identification through Suicide Crisis Syndrome (SCS)-Based Computerized Adaptive Screening |
|---|---|
| Contribution code | D2.133 |
| Authors | |
| Form of presentation | Poster |
| Topic |
|
| Abstract |
Imminent Suicide Risk Identification through Suicide Crisis Syndrome (SCS)-Based Computerized Adaptive Screening Shira Barzilay1,2, Shiri Liber1,2, Alan Apter1,2 Department of Community Mental Health, University of Haifa Mental Health Department, Schneider Children’s Medical Center of Israel Aims: Develop a brief, computerized adaptive screening tool to assist clinicians assessing suicide risk. Methods: 287 patients aged 8–17 were recruited from Schneider Children’s Hospital after reporting suicidal thoughts or behaviours. Participants completed questionnaires at intake, one month, three months. The study compared predictive performance of two item sets: SCS combined with Columbia Suicide Severity Rating Scale (SCS+CSSRS), and an expanded pool of 209 items (EXPANDED), incorporating additional scales for psychopathological symptoms, sleep disturbances, interpersonal problems. Three adaptive algorithms—Decision Tree, Logistic Regression, and LightGBM—were evaluated to identify the most effective predictors of suicide attempts at one and three months. Results: The SCS+CSSRS set, using the LightGBM algorithm, generated the most effective short adaptive screening questionnaire. At one month, performance metrics were AUC 0.614, Accuracy 0.763, Sensitivity 0.853, Specificity 0.853, and PPV 0.217, requiring a maximum of six questions. At three months, the metrics were AUC 0.664, Accuracy 0.676, Sensitivity 0.556, Specificity 0.710, and PPV 0.350, still with a maximum of six questions. The EXPANDED set, also using the LightGBM algorithm, demonstrated slightly higher AUC values but required up to eight questions. At one month, metrics were AUC 0.684, Accuracy 0.728, Sensitivity 0.500, Specificity 0.767, and PPV 0.217. At three months, metrics were AUC 0.680, Accuracy 0.666, Sensitivity 0.540, Specificity 0.701, and PPV 0.337. Conclusions: A brief computerized adaptive screening tool based on the SCS and past suicidal behaviour may provide clinicians with a valuable resource for identifying short-term suicide risk in youth. |