Yanmei Shen1, Wenyu Zhang2, Bella Siu Man Chan3, Yaru Zhang4, Fanchao Meng4, Elizabeth A Kennon5, Hanjing Emily Wu5, Xuerong Luo6, Xiangyang Zhang7. 1. Department of Psychiatry, The Second Xiangya Hospital, Central South University, Changsha, Hunan 410011, China; Mental Health Institute of Central South University, China National Clinical Research Center on Mental Disorders (Xiangya), China National Technology Institute on Mental Disorders, Hunan Technology Institute of Psychiatry, Hunan Key Laboratory of Psychiatry and Mental Health, Changsha, Hunan 410011, China; The Department of Educational and Counselling Psychology, and Special Education, The University of British Columbia, Vancouver, Canada. 2. School of Electronic and Information Engineering, Key Laboratory of Communication and Information Systems, Beijing Municipal Commission of Education, Beijing Jiaotong University, Beijing, 100044, China. 3. The Department of Educational and Counselling Psychology, and Special Education, The University of British Columbia, Vancouver, Canada. 4. Department of Psychiatry, The Second Xiangya Hospital, Central South University, Changsha, Hunan 410011, China; Mental Health Institute of Central South University, China National Clinical Research Center on Mental Disorders (Xiangya), China National Technology Institute on Mental Disorders, Hunan Technology Institute of Psychiatry, Hunan Key Laboratory of Psychiatry and Mental Health, Changsha, Hunan 410011, China. 5. Department of Psychiatry and Behavioral Sciences, The University of Texas Health Science Center at Houston, Houston, TX, USA. 6. Department of Psychiatry, The Second Xiangya Hospital, Central South University, Changsha, Hunan 410011, China; Mental Health Institute of Central South University, China National Clinical Research Center on Mental Disorders (Xiangya), China National Technology Institute on Mental Disorders, Hunan Technology Institute of Psychiatry, Hunan Key Laboratory of Psychiatry and Mental Health, Changsha, Hunan 410011, China. Electronic address: luoxuerong@csu.edu.cn. 7. Department of Psychiatry and Behavioral Sciences, The University of Texas Health Science Center at Houston, Houston, TX, USA. Electronic address: zhangxy@psych.ac.cn.
Abstract
BACKGROUND: Suicide has become one of the most prominent concerns for public health and wellness; however, detecting suicide risk factors among individuals remains a big challenge. The aim of this study was to develop a machine learning algorithm that could effectively and accurately identify the probability of suicide attempts in medical college students. METHODS: A total of 4,882 medical students were enrolled in this cross-sectional study. Self-report data on socio-demographic and clinical characteristics were collected online via website or through the widely used social media app, WeChat. 5-fold cross validation was used to build a random forest model with 37 suicide attempt predictors. Model performance was measured for sensitivity, specificity, area under the curve (AUC), and accuracy. All analyses were conducted in MATLAB. RESULTS: The random forest model achieved good performance [area under the curve (AUC) = 0.9255] in predicting suicide attempts with an accuracy of 90.1% (SD = 0.67%), sensitivity of 73.51% (SD = 2.33%) and specificity of 91.68% (SD = 0.82%). LIMITATION: The participants are primarily females and medical students. CONCLUSIONS: This study demonstrates that the random forest model has the potential to predict suicide attempts among medical college students with high accuracy. Our findings suggest that application of the machine learning model may assist in improving the efficiency of suicide prevention.
BACKGROUND: Suicide has become one of the most prominent concerns for public health and wellness; however, detecting suicide risk factors among individuals remains a big challenge. The aim of this study was to develop a machine learning algorithm that could effectively and accurately identify the probability of suicide attempts in medical college students. METHODS: A total of 4,882 medical students were enrolled in this cross-sectional study. Self-report data on socio-demographic and clinical characteristics were collected online via website or through the widely used social media app, WeChat. 5-fold cross validation was used to build a random forest model with 37 suicide attempt predictors. Model performance was measured for sensitivity, specificity, area under the curve (AUC), and accuracy. All analyses were conducted in MATLAB. RESULTS: The random forest model achieved good performance [area under the curve (AUC) = 0.9255] in predicting suicide attempts with an accuracy of 90.1% (SD = 0.67%), sensitivity of 73.51% (SD = 2.33%) and specificity of 91.68% (SD = 0.82%). LIMITATION: The participants are primarily females and medical students. CONCLUSIONS: This study demonstrates that the random forest model has the potential to predict suicide attempts among medical college students with high accuracy. Our findings suggest that application of the machine learning model may assist in improving the efficiency of suicide prevention.
Authors: Sönke Johann Peters; Mario Schmitz-Buhl; Olaf Karasch; Jürgen Zielasek; Euphrosyne Gouzoulis-Mayfrank Journal: BMC Psychiatry Date: 2022-07-14 Impact factor: 4.144