| PART Ⅰ THE EPIDEMIOLOGICAL CHARACTERISTICS AND RISK FACTORS FOR SUICIDAL AND SELF-INJURIOUS BEHAVIORS IN ADOLESCENTSObjective: To investigate the epidemiological characteristics and risk factors for suicidal and self-injurious behaviors in adolescents.Methods: The stratified cluster random sampling was used to investigate the mental health status of grade one and two students in senior high schools in Chongqing.22284 students aged 16-20 were selected from20 middle schools.The contents of the questionnaires included: the general demographic characteristics,personalities(Eysenck Personality Questionnaire,Barratt Impulsiveness Scale,Buss & Perry Aggression Questionnaire),current psychological status,the quality of life,health risk behaviors,and the suicidal and self-injurious ideation and behaviors.The descriptive statistical analysis,t-test,and chi-square test(if necessary,Fisher’s exact test)were used to analyze the data with SPSS 26.0 software.P<0.05 was regarded as statistically significant.Results: In the last 12 months,19.6%,12.9% and 7.2% of the adolescents sometimes or often felt that it was boring to live,it might be better to die,and thought about suicide or self-injury.The proportion of adolescents who sometimes or often thought about suicide and self-injury in the last month was 5.1%.The proportion of the adolescents who committed suicide or self-injury at least once was 2.7%.In the last month,the proportion of the adolescents with poor or very poor physical condition,mental and psychological status,economic status,and learning status were4.5%,7.2%,13.1% and 13.9%,respectively.The proportion of adolescents who had a poor or very poor relationship with family members or other people in the last month was 4.4% and 2.8%,respectively.Adolescents with female,urban,suicidal or self-injurious ideation in the last 12 or 1month,poor physical and mental health status,poor economic status,poor learning status,and poor interpersonal relationships in the last month were more likely to commit suicide or self-injury.In the last month,adolescents who had more days in insomnia,smoking,drinking,gambling,and the longer time their lives were affected by drinking,psychological or physical problems,were more likely to commit suicide or self-injury.Adolescents with high scores of P,N,E,impulse,aggression,total score of SCL-90 and mean values of its 10 sub-scales were more likely to commit suicide or self-injury.Conclusion: The suicidal and self-injurious behaviors in adolescents were closely related to gender,family location,personalities,current quality of life,recent mental health status,and recent health risk behaviors.In the future,we should learn more experience of prevention and intervention to suicidal and self-injurious behaviors in adolescents from mature foreign communities and school institutions,then combine them into the practice of our own country.These will give more supports to families in the prevention of suicidal and self-injurious in adolescents in China.PART Ⅱ EARLY WARNING MODEL OF SUICIDAL AND SELF-INJURIOUS BEHAVIORS IN ADOLESCENCE WITH MACHINE LEARNING APPROACHESObjective: To establish an early warning model of suicidal and self-injurious behaviors in adolescence with machine learning approaches.Methods: The stratified cluster random sampling was used to investigate the mental health status of grade one and two in senior high schools in Chongqing.Finally,22284 valid questionnaires were obtained.A total of 16 features including gender,family location,EPQ personality,impulsive personality,and aggressive personality were selected.SPSS 26.0software was used for logistic regression analysis,and the test level was0.05.We used the support vector machine,random forest,decision tree,extreme gradient enhancement,and the Bagging decision tree classifier to construct the early warning model of suicidal and self-injurious behaviors in adolescence.Sensitivity and specificity were used to evaluate the machine learning model.Results: Logistic regression showed that urban areas,girls,high neuroticism and psychoticism,high impulsive planning,high physical aggression,and self-aggression were the risk factors of suicidal and self-injurious behaviors in adolescence.The specificity and sensitivity of the five machine learning algorithms used for the early warning model of the suicidal and self-injurious behaviors in adolescence were 35.5%-84.7%and 39.2%-82.8%,respectively.Overall,the random forest and Bagging decision tree classifier performed better.The top five most important features of the early warning model of suicidal and self-injurious behaviors in adolescence are impulsive action,psychoticism,self-aggression,verbal aggression,and hostility.Conclusion: The machine learning approaches performed well in building the early warning model of suicidal and self-injurious behaviors in adolescence.Personalities have the potential to provide more reliable and predictive information than psychological symptoms and mental disorder diagnosis.Comprehensive prevention strategies should be adopted for suicidal and self-injurious behaviors in adolescence.PART Ⅲ EARLY WARNING MODEL OF 12-MONTH SUICIDAL AND SELF-INJURIOUS BEHAVIORS IN ADOLESCENTS WITH MACHINE LEARNING APPROACHESObjective: To establish the early warning model of 12-month suicidal and self-injurious behaviors in adolescents with machine learning approaches.Methods: A total of 182836 first-and second-year students were selected from middle schools,vocational high schools,and universities through the stratified cluster random sampling.A cohort was established,and the same group of people were re-tested at the same time period in the second year.A total of 25767 students participated in the both two tests.Twenty-two features were selected including age,recent mental health status,quality of life,health risk behaviors,suicidal and self-injurious ideation.SPSS 26.0 software was used for logistic regression analysis,and the test level was 0.05.We use five machine learning algorithms including support vector machine,random forest,decision tree,extreme gradient boosting,and Bagging decision tree classifier to construct an early warning model of 12-month suicidal and self-injurious behaviors in adolescents.We evaluated the performance of the machine learning algorithms with sensitivity and specificity.Results: A total of 58 adolescents with 12-month suicidal and self-injurious behaviors were found in the second-year screening,with an incidence rate of 0.23%.There were significant differences between adolescents with and without 12-month suicidal and self-injurious behaviors in age,smoking,internet addiction,somatization,impulsive,sensitivity to interpersonal relationships,depression,anxiety,hostility,phobia,paranoia,psychotic,and others(appetite and sleep)(P<0.05).Logistic regression showed that more smoking in the last 30 days,hostility,younger in age,felt boring in life in the last 1 year,and suicidal and self-injurious ideation in the last month were risk factors to 12-month suicidal and self-injurious behaviors in adolescents.The specificity and sensitivity of the five machine learning algorithms is 32.9%-98.1%,and16.7%-66.7%,respectively.The decision tree and Bagging decision tree classifiers performed better.The top 10 most important features of12-month suicidal and self-injurious behaviors in adolescents were:interpersonal sensitivity,suicidal and self-injurious ideation in the past year,hostility,phobia,psychosis,internet addiction,poor economic status in the past month,poor relationship with family members in the past month,paranoia,and anxiety.Conclusion: The machine learning approaches,especially the Bagging decision tree classifier,showed good performance in the early warning models of suicidal and self-injurious behaviors in both 12-month and lifetime in adolescents.The early warning system requires the close cooperation of the schools,communities,hospitals,and families to effectively reduce the risk of 12-month suicidal and self-injurious behaviors in adolescents.PART Ⅳ EVALUATION OF THE EARLY WARNING MODEL FOR SUICIDAL AND SELF-INJURIOUS BEHAVIORS IN ADOLESCENCEObjective: To evaluate the early warning model of suicidal and self-injurious behaviors in adolescence among the vocational high school students.Methods: A total of 55172 students from 19 vocational high schools were selected through the stratified cluster random sampling.Gender,family location,EPQ personality,impulsive personality,and aggressive personality composited the total of 16 features in this part.Statistical analysis was performed with SPSS26.0 software.Descriptive statistical analysis was conducted to understand the profile of the data.The early warning model of suicidal and self-injurious behaviors in adolescence constructed before was evaluated by Bagging decision tree classifier.Results: The valid sample size was 55172,with 29741 males and24873 females.The mean age was 19.2±1.19 years.558 students committed suicidal and self-injurious behaviors,accounting for 1.01%.The early warning model(Bagging decision tree classifier)of suicidal and self-injurious behaviors in adolescence which was built in middle school students,was moved to the group of vocational high school students,and showed a good performance.The sensitivity was 60.0%.Conclusion: The early warning model of suicidal and self-injurious behaviors in adolescence showed a good transferability in the vocational high school students,which provides a good foundation for the practical application of this model. |
PDF Full Text Request