Multilevel Models Analysis On The Influence Of Smoking Among Middle School Students And University Students In Tianjin

Objective Smoking is the serious problem to public health now. It can arose a great deal of disease and it is likely to induce some disease. But smoking rate in China has risen these years especially adolescent. Therefore, discussing the influential factors of smoking among middle school students and university students on individual level and class level, confirming high risk group can provide verity proof that may establish pertinent prevention measures which can decrease smoking behavior of middle school students.Methods The self report data surveies on knowledge, attitude and behavior about smoking comes from a random cluster sample of 25 classes in two middle schools and 24 classes in 12 universities' students in Tianjin. First, we describe the characters of smoking about middle school students and university students. And then use factor analyse to extract the relative content. At last, we constitute two multilevel models for middle school students smoking status and university students smoking status respectively, discover the influence of smoking.Results Through analyzing 1061 questionnaires of middle school students, we can learn the trying smoking rate of middle school students is 17.6%. When finding the content of questionnaire is not independent and there is statistical significance, we use factor analysis to extract three factors knowledge, attitude and environment. Finally, taking 'if having tried smoking' as depended value and age, gender, common factors and so on as independed value to constitute a binomial two-level model. Comparing with general logistic regression analysis,the standard errors analyzed by multilevel models are small. The risk factors of middle school students' trying smoking are age (OR= 1.421), attitude to smoke (OR=l.2l3) and smoking behavior in environment (OR=4A72) on individual level. Inculcation (OR=\.\13) on class level can enhance the contact between smoking behavior in environment and middle school students' trying smoking. The protection factor is gender (O/?=0.324) on individual level. Among the influence factors, age, attitude and gender are not difference among classes. Environments are difference among classes.Through analyzing 743 questionnaires of university students, we can learn the occasional smoking rate is 10.0%, the regular smoking rate is 7.4%, the having given up smoking rate is 5.9%. Similarly, when finding the content of questionnaire is not independent and there is statistical significance, we use factor analysis to extract three factors knowledge, attitude and environment. Finally, taking 'smoking status' as depended value and age, gender and so on as independed value to constitute a multinomial two-level model. The same as middle school students' analysis, comparing with general logistic regression analysis, the standard errors analyzed by multilevel models are small. For the students of smoking occasionally, the risk factor is smoking behavior in environment (Q/?=2.722), the protection factors are gender (O/?=0.243) and moods (Oi?=0.071). Both of environment and gender are not difference among classes. Moods are difference among classes. For the students of smoking regularity, the risk factor is smoking behavior in environment (OR= 19.075), the protection factors are gender (<9#=0.186) and moods (Q/?=0.078). All of them are not difference among classes. For the students who are having given up smoking, the risk factor are gender (##=0.180) and moods (##=0.221), the protection factors are age (OR=\.567) and smoking behavior in environment (OR-2.776).