Prediction of smoking behavior in Japanese young adults

Eighty-eight second grade students of a senior high school in Saitama prefecture in Japan participated in a prospective study to predict cigarette smoking behavior 3.5 years later. Predictor variables include sex, knowledge, beliefs and attitudes toward smoking, previous smoking behavior, and smoking behavior of their families. Stepwise discriminant analyses revealed that 90% of the smokers and 65% of the non-smokers were correctly classified. In this model, previous smoking behavior proved to be the best predictor. Attitude toward adult male's smoking, sex and smoking behavior of subjects' families were also related. These four variables explained 35% of the variance in smoking behavior. As for stepwise discriminant analyses among those who had not smoked at baseline, 78% of the smokers and 76% of the non-smokers were correctly classified. Attitude toward adult male's smoking, sex, knowledge about long-term effects of cigarette smoking and smoking behavior of their families entered the model in this order. These four variables explained 37% of the variance. Implications of this study for smoking prevention programs in Japan are discussed.