Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care
In this article, we assessed the performance of several predictive modeling algorithms of suicide attempt resulting in inpatient hospitalization or suicide among youths ages 9 to 18 (N = 34,528) after contact (6–12 months) with a mental-health specialist in Stockholm, Sweden, from 2006 to 2012. Usin...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
SAGE Publications
2024
|
| _version_ | 1824458891129257984 |
|---|---|
| author | O’Reilly, LM Fazel, S Rickert, ME Kuja-Halkola, R Cederlöf, M Hellner, C Larsson, H Lichtenstein, P D’Onofrio, BM |
| author_facet | O’Reilly, LM Fazel, S Rickert, ME Kuja-Halkola, R Cederlöf, M Hellner, C Larsson, H Lichtenstein, P D’Onofrio, BM |
| author_sort | O’Reilly, LM |
| collection | OXFORD |
| description | In this article, we assessed the performance of several predictive modeling algorithms of suicide attempt resulting in inpatient hospitalization or suicide among youths ages 9 to 18 (N = 34,528) after contact (6–12 months) with a mental-health specialist in Stockholm, Sweden, from 2006 to 2012. Using 209 predictors across domains (e.g., clinical, demographic, family, neighborhood, social) identified from national registers, we applied standard logistic regression, regularized logistic regression, and machine-learning algorithms (i.e., random forests, gradient boosting, support vector machines). Standard logistic regression (area under the receiver operating characteristic curve [AUC] = 0.77, 95% confidence interval [CI] = [0.72, 0.82]) and random-forest models (AUC = 0.80, 95% CI = [0.74, 0.86]) demonstrated the highest AUCs. Sensitivities ranged from 0.33 (support vector machines) to 0.91 (standard logistic regression). Although the study was underpowered to detect a difference between logistic regression and machine-learning algorithms (outcome prevalence = 0.7%), performance metrics were similar across models. Logistic regression is not clearly worse than machine-learning approaches. Ongoing research is needed to examine how prediction models can augment clinical decision-making. |
| first_indexed | 2025-02-19T04:33:05Z |
| format | Journal article |
| id | oxford-uuid:63c10ea2-ec47-437c-994d-48ad42830faf |
| institution | University of Oxford |
| language | English |
| last_indexed | 2025-02-19T04:33:05Z |
| publishDate | 2024 |
| publisher | SAGE Publications |
| record_format | dspace |
| spelling | oxford-uuid:63c10ea2-ec47-437c-994d-48ad42830faf2025-01-20T10:53:02ZEvaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty careJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:63c10ea2-ec47-437c-994d-48ad42830fafEnglishSymplectic ElementsSAGE Publications2024O’Reilly, LMFazel, SRickert, MEKuja-Halkola, RCederlöf, MHellner, CLarsson, HLichtenstein, PD’Onofrio, BMIn this article, we assessed the performance of several predictive modeling algorithms of suicide attempt resulting in inpatient hospitalization or suicide among youths ages 9 to 18 (N = 34,528) after contact (6–12 months) with a mental-health specialist in Stockholm, Sweden, from 2006 to 2012. Using 209 predictors across domains (e.g., clinical, demographic, family, neighborhood, social) identified from national registers, we applied standard logistic regression, regularized logistic regression, and machine-learning algorithms (i.e., random forests, gradient boosting, support vector machines). Standard logistic regression (area under the receiver operating characteristic curve [AUC] = 0.77, 95% confidence interval [CI] = [0.72, 0.82]) and random-forest models (AUC = 0.80, 95% CI = [0.74, 0.86]) demonstrated the highest AUCs. Sensitivities ranged from 0.33 (support vector machines) to 0.91 (standard logistic regression). Although the study was underpowered to detect a difference between logistic regression and machine-learning algorithms (outcome prevalence = 0.7%), performance metrics were similar across models. Logistic regression is not clearly worse than machine-learning approaches. Ongoing research is needed to examine how prediction models can augment clinical decision-making. |
| spellingShingle | O’Reilly, LM Fazel, S Rickert, ME Kuja-Halkola, R Cederlöf, M Hellner, C Larsson, H Lichtenstein, P D’Onofrio, BM Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care |
| title | Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care |
| title_full | Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care |
| title_fullStr | Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care |
| title_full_unstemmed | Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care |
| title_short | Evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental-health specialty care |
| title_sort | evaluating machine learning for predicting youth suicidal behavior up to 1 year after contact with mental health specialty care |
| work_keys_str_mv | AT oreillylm evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT fazels evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT rickertme evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT kujahalkolar evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT cederlofm evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT hellnerc evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT larssonh evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT lichtensteinp evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare AT donofriobm evaluatingmachinelearningforpredictingyouthsuicidalbehaviorupto1yearaftercontactwithmentalhealthspecialtycare |