We observed changes in health risk behaviors of school children in Taiwan from 2001 to 2003. In terms of single behaviors, behaviors that increased with time were swearing, staying up late, and playing video games for prolonged periods, and behaviors that decreased with time were suppressing urination, drinking alcohol, chewing betel nut, hitting others and eating fast food. As a whole, the 13 health risk behaviors in elementary school students could be aggregated into three latent factors: unhealthy habits, aggressive behaviors, and substance use behaviors. The results are similar to findings in previous studies [8–13]. After controlling for residential area and gender, we found that unhealthy habits and aggressive behaviors increased from 2001 to 2003, but that substance use behaviors decreased slightly.
The two leading health risk behaviors for students since fourth grade were staying up late and eating snacks at night (Table 1). The least prevalent health risk behavior was chewing betel nut. The prevalences of these behaviors were relatively stable over the 3 years. However, the aggressive behavior of swearing jumped from being the fifth to being the third most common behavior from 2001 to 2003. The result of time trend analysis also showed that the frequency of swearing had the most significant increase with time. Compared to the rate of seeing other 13–15 year olds swearing, insulting or making nasty comments at least once or twice a month reported by a research group in the US , the frequency of swearing in this study sample (59.08%~71.22%) was higher than the US sample (14%~18%). Even though the participants in the US were older and reported what they observed rather than their own situation, the rate in our sample was almost three-fold that in the US. Therefore, grade 5 could be the appropriate time for preventive education targeting aggressive behaviors, especially verbal aggression in Northern Taiwan.
It is important not to ignore substance use behaviors, as once they have developed they can be difficult to break. When looking at individual behavioral items, two of the three substance use behaviors decreased from 2001 to 2003 (Table 5). The total score of aggregated substance use behaviors also decreased slightly. We used stricter definitions (student took one puff, took one sip of alcohol, and chewed one piece of betel nut) for substance use behaviors. We found that more children had drunk alcohol than smoked or chewed betel nut (Table 1). One possible reason is that during celebrations or feasts in Taiwan, parents often let children have a taste of alcohol. Although our study did not find high rates of substance use behaviors, the results still indicate that a small proportion of children have tried using these substances at a young age. Furthermore, the decrease in each substance use behavior was only found among boys, and girls did not show the same trends (Table 2). Research indicates that gender differences exist in health-related beliefs and health behavior [35–37]. We found that boys and girls had different substance use behavior patterns. Although boys engaged in substance use more frequently than girls, girls did not show the same pattern of decline as boys. Hence, attention still needs to be given to girls' substance using behaviors.
The three types of health risk behaviors showed different time trends. Unhealthy behavior and aggressive behavior increased with time, whereas substance use behavior decreased with time (Table 5). However, comparing the results of time trend analysis for the aggregated behaviors to those for single behaviors (Table 2), we can see that not all of the single behaviors followed the general time pattern. Staying up late and playing video games for prolonged periods showed the greatest contribution to the increasing pattern of unhealthy behavior. Likewise, swearing played a major role in the increase in aggressive behavior. For substance use behavior, there was more consistency between changes in single behaviors and the total score. From these results we can see that although a health risk behavior type may have a general time pattern as a whole, each single behavior may still have its own unique time pattern. For example, for the six behaviors that were grouped as unhealthy habits, the first behavior to increase could be staying up late. For aggressive behavior, swearing may be the first behavior in this category. Therefore, as health educators, although we can develop strategies targeting groups of related behaviors, we still need to pay attention to the different development time courses of each behavior in the group. As a result, two different strategies would be necessary for health education, namely those focusing on single behaviors as well as those focused on groups of behaviors.
The mean scores of aggressive behaviors and substance use behaviors suggest that these types of behaviors were quite infrequent and that only a small proportion of students experiment with substances at this young age (Table 3). This finding is probably due to the fact that the participants came from the general population. However, Millstein et al.  indicated that an extremely important and currently neglected area in professional education is knowledge about the general adolescent population. Our study is one of the few studies to fill this gap.
Furthermore, we unexpectedly found that some behaviors decreased in the 5th grade, as well as the mean scores of the three behavioral groups. This is probably because students in Taiwan are rearranged into new classes in the 5th grade and therefore, most of the students will be with new classmates at this time. They are too unfamiliar with each other to undertake risk behaviors together such as playing video games, throwing things when angry, and hitting others. They may also feel too stressed to relax on their own by doing things such as watching TV, eating snacks late at night and fast food. However, we have no evidence to support this hypothesis and further research is needed to solve this puzzle.
Laaksonen et al.  showed that having three or four health risk behaviors concurrently is related to age: having such behaviors was most common in those aged 20–34, less common in those aged 35–49, and least common in those aged 50–64. According to our results, health risk behaviors are engaged in as early as elementary school. Millstein et al.  suggested that we should stop viewing young adolescents as naive children and begin to view them as participants in a changing social environment. As a result, they should be educated about healthy behaviors and encouraged to develop healthy behaviors in childhood.
As shown in Table 5, substance use behaviors slightly decreased over time. The decrease in substance use behaviors is possibly due to implementation of anti-smoking legislation and education with an emphasis on drugs and smoking. Such legislation and education have strengthened the formation of negative social norms and restrictive attitudes about these kinds of behaviors. One possible reason for the increase in unhealthy habits and aggressive behaviors over time is maturity. As students move from grade 4 to grade 6, not only are their bodies developing and maturing, but they are also undergoing psychological development. As these students become more independent and develop a stronger sense of self, they are less willing to accept the restrictions placed on them by parents and schools. In addition, they are influenced by the media, television programs, and computer and video games. The increase in staying up late is likely due to greater amounts of homework, and the reason for the increase in swearing could be peer pressure or the increasing acceptability of swearwords. This project will continue to collect information on the students to learn more about the time trends of health risk behaviors and their effects.
The participants in this study represent only students of public schools in Taipei City and Hsinchu County in northern Taiwan. In addition, the study focused on health risk behaviors over only 3 years. These limitations restrict the generalizability of our findings. We suggest that future studies involve different populations and monitor behaviors for longer durations to more closely examine the issue of health risk behaviors in childhood.