The Study On The Early Warning BP Neural Network And Predictor Of The Attempted Suicide For Rural Residents In Shandong Province

BackgroundAttempted Suicide is that the person with suicidal ideation taken suicidal action for various reasons and results in disability, but failed to die. Chinese classification and diagnostic criteria for defining mental disorders defined attempted suicide, which are the person with suicide motive who took suicide behavior that can lead to death, but did not cause death. Along with the economic and social development, China’s suicide rate presents downtrend in the past 20 years, the total national suicide rate dropped from 17.65 per ten thousand person in 1987 to 6.60 per ten thousand person in 2008[1] By integrating the official data, Zhang jie calculated that the urban and rural suicide rates are 3.59/10 and 8.25/10 per ten thousand person in 2008 respectively and men and women’s suicide rate in rural China are 8.62 and 7.87 per ten thousand people respectively[1].Although the suicide rate decreased in China, but it is still the main cause of death and can cause serious burden to the family and society. The causes of suicide are still a common concern for domestic and foreign experts.Previous studies have shown that attempted suicide is an important predictor of suicide success, about 30% to 60% suicide completed has the history of attempted suicide. Attempted suicide is the clear features of suicidal behavior[2]. So it is very important significant to find the risk factors of attempted suicide, establish a set of early warning index system, and put forward feasible targeted preventive measures for suicide before suicide completed in our country.By studying the related literature of attempted suicide at home and abroad, we can find that there are not many foreign papers on the Chinese study of attempted suicide. Some scholar has studied the characteristic of the Chinese attempted suicide basing on emergency cases[3]. Some expert has discussed the predictors on attempted suicide on the low and high suicide intention of Chinese rural people[4]. There are some special suicide literatures on Chinese patients with schizophrenia[5,6], on the attempted suicide basing on hospital data in Shenyang province, on the suicide studied of Chinese teenagers and college students. So far, there are so many domestic literates on Chinese attempted suicide. Jia has researched the epidemiological characteristics of attempted suicide in China. Some scholars have reviewed the epidemic characteristics and risk factors on Chinese attempted suicide, discussed the psychological and social factors of suicide attempts. There have been so many previous reports on the risk factors of attempted suicide in China. Phillips etc. have studied the dangerous factors of attempted suicide in rural Chinese people and in Shandong province.Suicide prevention research center of Shandong university has still researched the risk factor of suicidal intention and attempted suicide in early time, and carried out the molecular epidemiological studies and discussed the environmental factors and the genetic factors of attempted suicide. These early literatures provide good basis and research experience for exploring the risk factors of attempted suicide furtherly.Although there are so many literatures on the attempted suicide, suicide phenomenon is a comprehensive result of multiple factors. The influence factor may content the individual factors, biological factors, psychological factors, social factors, environmental factors, cultural factors and so on. Although previous studies have explored the influence factor, research contains were not comprehensive enough. Most of the early literature on Shandong province was basing on the data collected form one area or a hospital and the sample size was not abundant. Therefore, it is very necessary to explore the main influence factors of attempted suicide from a broader perspective, use reasonable statistical methods to screen the warning index and put forward its prevention interventions in Shandong province.Objectives1. To explore and discusses the possibility main influencing factors of attempted suicide from a broader perspective view, according to the extensive literatures.2. To measure and mine the major risk factors of the rural residents in Shandong province by reasonable statistical methods.3. Build the multivariate Logistic warning model for attempted suicide and forecasting neural network basing on the BP neural network and test their validation.4. Build a set of early warning index for attempted suicide, and put forward the crisis management strategies and measures.MethodsThis research collected the case data coming from the attempted suicide information of hospital emergency and hospitalization registration. The 659 cases were selected by the random sampling method from 13 counties in Shandong province such as Longkou, Lijin, Ningyang, Gaotang, Shen county etc. Meanwhile, the control respondent of each attempted suicide case was chosen according to the same gender and similar age to carry out the case-control study.The residents’health questionnaire was used in this study, which contains various aspects of information of investigation, including personal information, economic, cultural, social, spiritual and psychological characteristics and so on. Home page of the residents’ health questionnaire including some of the relevant survey information, such as the number of questionnaire, category of target, category of interview people, gender, age, the interview time, interview place ect. The main contents of the residents health questionnaire includes 19 questionnaires/scales (questionnaire 1-19) and SCID-I/P axis I diagnosis clinical interview questionnaire.The x± SD was used to descript the quantitative data if the data according to the normal distribution and median and quartile range were used if the data not according to the normal distribution; rate and ratio were used to descript the qualitative data. The t test and t test and nonparametric test (Wilcoxon rank test/the Mann-Whitney U) were used to compare the difference of quantitative data between case and control groups and x2 test, adjust x2 test, Fisher’s exact probability method were used to compare the difference of qualitative data between two groups. Logistic regression model was used to construct the early warning model of attempted suicide and the warning neural network was built based on BP neural network. Statistical analysis software such as SAS9.2, SPSS17.0 and Matlab R2010b ware applied in current study.Results1. There are no statistical differences between the attempted suicide group and the control group in the variable such as gender, age, residence and nationality, which indicated that the main confounding factors of demographic variable were balanced and controlled well. The Cronbach’s coefficients of the scales used in current study above 0.67, which showed that the internal consistency reliabilities were good. Those ensure that the results of current research are accurate and reliable.2. The sample statistical analysis results showed that demographic characteristics variables such as "less the education time", "lower of the highest degree" were the risk factor, and "married or cohabiting" was the protection factor for suicide attempts. The variable such as "be religious", "be superstitious", "be poorer health", "has chronic disease" and "has mental health" were the risk factors for attempted suicide in the individual aspect, "poorer relations between husband and wife "," less population", "poorer relations with parents ", "lower status in the family", "suicide history of family members " were the major risk factors for attempted suicide in the family environment aspect. When it comes to the economic aspect, "lower annual average family income", "lower proportion of personal property" and "poorer family economic condition" were the major risk factor for attempted suicide. In terms of social aspect, "live in dormitories", "poor social relationship", "can not get help when in trouble", "lack of social support", "negative life events" were the major risk factors of attempted suicide and "higher WHO community scale score", "closer relation with villagers ", "be optimism on the future of the village" were the protection factors. Four aspects of the psychological strain theory, namely, "values strain", "desire strain", "relative deprivation strain" and "strain of lack crisis response capacity" were the major risk factors for attempted suicide. Personality psychological characteristics, such as "despair", "impulse", "anxiety", "depression" and "higher suicide attitude and cognition" were risk factors for attempted suicide.3. The multivariate Logistic regression results showed that there were 12 statistically significant variables in the prediction model."low highest degree", "family history of suicide", "be poor on overall health", "has mental health", "desire not be realize", "despair", "impulse", "depression", "negative life events" were the risk factors for attempted suicide. And "social support", "crisis coping skills", "higher WHO community scale score" were the protection factors for attempted suicide.4. The current study built a three layers neural network including with one hidden layer, which including 9 hidden layer nodes. The results showed that the prediction accuracy of the built neural network reached 84.6%, whose discriminate efficiency was superior to the traditional multivariate analysis, which lay the foundation for build the expert diagnosis system for the attempted suicide in the future.Conclusion1. It is feasible to use the traditional statistical methods combined with neural network to explore influence factors and build the warning neural network for attempted suicide; basing on the result of this study, the expert auxiliary diagnosis system of attempted suicide can be built in the future.2. Attempted suicide is the comprehensive result of many factors, individual factors, family environment, economic factors, social factors, psychological strains and psychological characteristics of individual were the main influence factors of suicide attempts.3. The probability of attempted suicide can be calculated according to the multivariate Logistic prediction model and the neural networks, which can be used to screen the high-risk groups and take comprehensive interventions to prevent the attempted suicide. The targeted intervention measures still can be carry out according to the seven influence aspects of the attempted suicide to preventing the occurrence of suicidal behavior effectively.Significances and the innovation1. Basing on more spacious viewer, this research explored and analyzed the major influence factor of attempted suicide, which included the individual factors, family factors, social factors, economic factors, cultural factors, social factors, psychological strains, etc. and the current research added new contents such as value and psychological strains which were not involved in the previous study.2. What is more, the current study reorganized and grouped the influence factors of attempted suicide into sevev categories by professional knowledge, embodying the holistic concept and top-level design concept and provided a good theoretical reference for building the index system of preventing of attempted suicide.3. In terms of methodology aspects, the traditional multivariate Logistic regression analysis and artificial neural network were combined applied to build the early warning model of attempted suicide. The combined application of traditional statistical methods and artificial intelligence neural network play the advantages of each method. The combined application ensure the built forecasting model is more approach to the actual nonlinear mapping, ensure the results of discriminant has higher accuracy. The intelligent expert auxiliary diagnosis system of attempted suicide can be constructed by this theoretical basis. This combined application has not been used in previous research on attempted suicide research.