Application of the Health Action Process Approach Model for Improving Excessive Internet use Behaviors Among Rural Adolescents in China: A School-Based Intervention Pilot Study

Chengmeng Tang Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Hein Raat Department of Public Health, Erasmus MC University Medical Center Rotterdam, The Netherlands Mingxia Yan Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Qiang Zhang Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Kehan Li Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Min Jiang Department of Epidemiology and Health Statistics, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Wanjie Tang Centre for Educational and Health Psychology, Sichuan University, Chengdu, China Jiayi Chen Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Ying Zhao Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China Qiaolan Liu (  liuqiaol@scu.edu.cn ) Department of Health Behavior and Social Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, China


Introduction
The Internet increasingly has become one of the ways for adolescents to learn, as well as to entertain (1). According to data released by the China Internet Network Information Center (CNNIC) in 2020, the number of users below the age of 18 years has been growing steadily; the Internet coverage rate among children and adolescents in rural areas in 2019 was 90.3%, which is almost the same rate as in urban areas (93.9%) (2). The Internet may provide adolescents with more chances to acquire knowledge, and to communicate with others (1, 3), and even become an important medium for maintaining and consolidating friendship (4). However, excessive Internet use may have a negative impact on the physical and mental health of adolescents (5)(6)(7).
Excessive Internet use is a behavior that is more prone to addiction, mainly manifested as a loss of control over Internet use (8). Excessive use of Internet may be associated with visual fatigue and with the risk to develop myopia; it is also associated with sedentary behavior, which is a risk factor of obesity and cardiovascular diseases (7,9). Previous studies have shown that adolescents who spend relatively more time online tend to acquire more information about health risk behaviors such as smoking and drinking alcohol, but on the other hand tend to adopt such health risk behaviors through the Internet (4). Moreover, adolescents who use the Internet relatively more often and longer, were shown to have more often problems with sleep disorders, depression, anxiety, and suicide (7,10).
Exploring the in uencing factors of adolescents' Internet use behaviors is crucial for the development of targeted interventions. Previous studies have found that age (4,11,12), gender (4,11,13), academic pressure (4,14,15), and parental restrictions (6,16) are related to adolescents' Internet use behaviors. However, cognitive psychological factors of adolescents themselves were not studied well; so far, most studies paid more attention to sociodemographic characteristics and social factors about family and peers (17). As a consequence, few interventions to improve excessive Internet use behaviors of adolescents are based on psychological-behavioral theories. A meta-analysis identi ed that interventions based on such theories, like cognitive-behavioral therapy (CBT) and peer support services, have positive effects when applied to Internetaddicted adolescents (18). This suggests that to prevent addictive Internet use, it is also relevant to explore the psychological factors in uencing Internet use behaviors of normal adolescents.
Some evidence suggests that behavior interventions based on theoretical frameworks are more effective than interventions that lack a theoretical basis (19). Cognitive psychological factors were included in several behavioral change theories, one of which is the health action processes approach model (HAPA). The HAPA model combines the stages of change model with the continuous change model to explain the emergence and maintenance of behaviors, as well as related processes from the perspective of motivation and volition (20,21). According to the HAPA model, there are two phases before a behavior occurs: the motivational phase and the volitional phase. In the motivational phase, the behavior intention is determined by risk perception, outcome expectancy, and action self-e cacy. The volitional phase mainly includes planning and maintenance self-e cacy (20,21).
The HAPA model has been used to study the determinants of behavior, as well as for the development of behavior change interventions in many studies; examples are studies with regard to physical activities and eating behaviors (20). Studies show that the HAPA model can explain food safety behaviors very well (22).
And, for example, a study employing the HAPA model as a theoretical framework for an intervention to target sedentary behavior of college students was successful to reduce the frequency of sedentary behaviors (23). Pinidiyapathirage et al. illustrated that the HAPA model is useful to develop effective interventions to promote physical exercise of women with a history of gestational diabetes (24).
To date, only very few studies addressed the determinants of excessive Internet use by adolescents in rural areas (13,25,26); most studies focus on populations in urban areas (5,(27)(28)(29)(30)(31). However, since the Internet coverage in rural areas has reached similar levels as in urban areas there is an urgent need to study the determinants of excessive Internet use of adolescents in rural areas to inform the development of effective interventions to promote 'healthy' Internet use by adolescents in rural areas. The HAPA model is expected to be very relevant to study this topic. Therefore, the objective of this study is to examine the applicability and effectiveness of the HAPA model in the intervention of excessive Internet use behaviors in Chinese rural adolescents.

Participants
The participants were derived from the rst, third, and fth survey data of a 2-year longitudinal study for rural adolescents in Sichuan Province (2015-2017). Typical sampling and cluster sampling were used to select the study areas and participants in November 2015. In the rst stage, a typical sampling method was used to select Zizhong County as the study area, representing rural areas in Sichuan. In the second stage, cluster sampling was used to randomly select two middle schools in the study area, one as the control school and the other as the experimental school. All seventh and tenth graders in both schools were included. After the rst baseline survey, a follow-up survey was conducted every six months with a total of ve surveys. In the fth survey, the samples were in the nine-grade and twelve-grade. 1044, 973, and 874 samples were obtained in the experimental school in the rst, third, and fth survey, respectively, and 1399, 1777, 1583 samples respectively in the control school.
According to the baseline survey, participants from the experimental school who met one of the inclusion criteria were considered to have excessive use of Internet behavior and were involved in this study. The inclusion criteria were (1) average daily online time from Monday to Friday ≥ 2 h (hour); (2) average daily online time on weekends ≥ 4 h; (3) usually daily game time ≥ 2 h; (4) being online overnight at least once in the past 30 days. 327 eligible participants were enrolled in the current analysis.

Intervention measures
After the baseline, four waves of intervention were conducted on the participants in the experimental school and each wave intervention was conducted before each follow-up survey. The purpose of the rst intervention was to cultivate the awareness and willingness of participants about Internet use behaviors. The main contents embraced publicizing the harms caused by excessive Internet use and the bene ts of reducing Internet use time, with the objective of promoting the action self-e cacy about Internet use behaviors. Health education courses and intervention manuals were the primary methods. The second intervention aimed to consolidate the effect of the rst intervention and to promote the formation of the intention of Internet use behavior. The professional guidance on plan formulation was added to the third intervention while strengthening the previous intervention contents. The main purpose of the fourth intervention was to strengthen the maintenance self-e cacy to improve Internet use behaviors. The principal method, health education courses with speci c cases, aimed to build up con dence via ameliorating Internet use behaviors. Moreover, freely available table tennis rackets and other sports equipment were provided to encourage participants to take part in outdoor activities, thereby prompting them to cut down the online time.

Sociodemographic characteristics
Sociodemographic characteristics included age, gender, grade, and left-behind status. The left-behind status was referred to adolescents aged 17 years or younger who have been left in their rural communities by one or both parents migrating in search of work in cities.

Variables of the HAPA model
The items of the HAPA model variables were measured and modi ed in accordance with the compilation principles of Schwarzer Ralf, and related behaviors (32,33).
Risk perception was assessed by three items (for example, "If you compare your current Internet use behaviors to that of individuals with the same gender and age, how likely do you think your academic performance will decline?"). Items were scored with 5 Likert scales (1 = very low, 5 = very high). Cronbach's alpha was 0.809.
Outcome expectancy was examined by ve items (for example, "Reducing Internet time is bene cial for maintaining vision."). Items were scored with 6 Likert scales (1 = completely disagree, 6 = completely agree). Cronbach's alpha was 0.864.
Action self-e cacy was estimated by ve items (for example, "I am con dent that I would rarely go online in my free time."). Items were scored with 5 Likert scales (1 = completely disagree, 5 = completely agree).
Intention was assessed by three items (for example, "I plan to reduce my Internet use time in the next month."). The item was scored with 6 Likert scales (1 = completely disagree, 6 = completely agree). Cronbach's alpha was 0.896.
Planning was determined by three items (for example, "I have made a plan for controlling the time spent online every week."). Items were scored with 6 Likert scales (1 = completely disagree, 5 = completely agree).
Maintenance self-e cacy was assessed by two items (for example, "After insisting on controlling the internet use time for one month, I can keep controlling even if one day exceeds the prescribed time.") Items were scored with 6 Likert scales (1 = completely disagree, 6 = completely agree). Cronbach's alpha was 0.833.
Sociodemographic characteristics and excessive Internet use behaviors were derived from the baseline survey.
When tting the HAPA model, we used risk perception, outcome expectancy, action self-e cacy, and intention from the third survey, and obtained planning, maintenance self-e cacy, and Internet use behaviors from the fth survey.

Data collection and analysis
We used uni ed questionnaires and survey process in the investigations. Before the interventions and surveys, we conducted uniform training for investigators to standardize the intervention and investigation process. Selfadministered questionnaires were used to obtain the data of participants. After each investigation, interactive inspections were executed in investigators to nd and rectify mistakes such as mis led, omissions, and logic errors. To ensure the accuracy of the data, double entry was used for questionnaire data entry.
Page 7/17 Epidata 3.1 was used to establish the database and data entry. SPSS 24.0 was used to sort and analyze the data. AMOS 21.0 was used to t the SEM. The Maximum Likelihood (ML) method was used for parameter estimation, and the Bootstrap method was used synchronously when tting the model. P-value < 0.05 was considered statistically signi cant.
In the descriptive analysis, the mean ± standard deviation (M ± SD) was used to describe quantitative variables and the frequency (%) was used to describe qualitative variables. The Chi-square test and the t-test were used to compare the differences of HAPA variables and Internet use behaviors before and after the interventions.
The bivariate analysis was adopted to analyze the correlations between the variables of the HAPA model. The

Sociodemographic characteristics and Internet use behaviors
In the baseline survey, the average age of the participants was (15.37 ± 1.31) years old, 168 (51.4%) were boys, 279 (85.3%) were in tenth grade, and 217 (66.4%) were left-behind adolescents. The results in Table 1 displayed that outcome expectancy, intention, and maintenance self-e cacy were signi cantly improved. The rate of excessive Internet use was declined from 100-70%, meanwhile, all Internet use behaviors have been declined to some extent, and there were statistically signi cant differences in daily Internet time on weekends and daily game time. Scores of the HAPA model variables Table 2 illustrated the comparison results of the HAPA model variables of participants with different sociodemographic characteristics after the interventions. Action self-e cacy of females was higher than that of males (P 0.05). Compared to low grade, senior students had a better performance in planning making and Internet use behaviors (P 0.05). However, there was no statistical difference between left-behind and non-leftbehind adolescents in the scores of the HAPA model variables.

Correlation analysis of the HAPA model variables
The results of the correlation analysis showed that there was no statistical correlation between outcome expectancy and risk perception, maintenance self-e cacy, as well as Internet use behaviors, while other variables were all correlated (P 0.05) ( Table 3). Risk perception was negatively related to action self-e cacy, intention, planning, maintenance self-e cacy, and Internet use behaviors (P 0.01). The correlation coe cients between other variables were positive (P 0.05). Note: *P 0.05,**P 0.01 ***P 0.001

Model tting and analysis results
The initial model was obtained by adjusting action self-e cacy and risk perception, outcome expectancy, as well as maintenance self-e cacy, all of which were found relevant in previous studies (20,34). The modi ed model (model 1) was obtained by appropriate adjustment according to the modi cation index (MI) (Figure.1).
The tting indices of model 1 were as follows:  Note: * P 0.05, ** P 0.01. The starting of the arrows were independent variables, and the end of the arrows were dependent variables.

Discussion
The results of this study illustrated that interventions based on the HAPA model can effectively improve psychological cognitive factors of the excessive Internet use behaviors of Chinese rural adolescents. After the interventions, the rate of excessive Internet users reduced by 30%. Among participants, the rate of average daily Internet time on weekends ≥ 4 h dropped from 57.2-39.1%, and the rate of usually daily game time ≥ 2 h reduced from 51.1-35.2%. Although no improvement was observed in average daily Internet time on weekdays ≥ 2 h and online overnight in the past thirty days, the rates after intervention did not increase. Meanwhile, the interventions improved outcome expectancy, intention, and maintenance self-e cacy of rural adolescents.
Previous school-based studies were designed to improved Internet use behaviors of urban adolescents. The studies focused on health education courses have positive effects on game time, while no impact on daily online time (17,35). Different from urban adolescents, Internet use behaviors of rural adolescents have not received enough attention and lacked targeted intervention measures. This may be the reason when facing the rapidly developing Internet rural adolescents lack su cient correct knowledge and lack of self-e cacy, which may lead to excessive Internet use. As far as we know, this is the rst intervention study aimed to improve rural adolescents' excessive Internet use. And the results illustrate that the HAPA model can be used to develop effective interventions for rural adolescents.
In this study, the changes in outcome expectancy, intention, and maintenance self-e cacy were key indicators to the reduction of Internet use. Previous studies have claimed that outcome expectancy is the strongest predictor of intention (36)(37)(38). Adolescents are inclined to generate the intention of reducing online time based on the awareness that reduction of online time could bring bene ts for themselves such as improving the eyesight and academic performance.
Meanwhile, intention is the most predictable factor for future behaviors (20). The generation of intentions is a prerequisite for planning and transforming planning to behaviors. Besides, maintenance self-e cacy is fundamental in the production of behaviors. Individuals with high maintenance self-e cacy are more likely to convert planning into behaviors and believe that they can overcome the potential obstacles when adopting aim behavior (39). Interventions based on the HAPA model lead to changes in behaviors by enhancing the cognition of these three key variables of the participants, so as to investigate the applicability and effectiveness of HAPA model in intervening excessive Internet use. Interestingly, there were certain differences in the HAPA variable scores of adolescents with different genders and different grades after the intervention, which denotes differences in their perception corresponding to targeted intervention. It meant that boys should strengthen the interventions of action self-e cacy, and younger adolescents are supposed to focus on planning and maintaining self-e cacy.
The results of SEM identi ed that the HAPA model has good applicability in rural adolescents' Internet use behaviors. In model 1, most paths had been effectively veri ed, apart from risk perception to intention, and maintenance self-e cacy to Internet use behaviors. We found that there was only an indirect impact between maintenance self-e cacy and Internet use behaviors, while the direct effect was not observed. It demonstrated that for Internet use behaviors of adolescents, the role of maintaining self-e cacy was to promote planning convert into behaviors and to maintain the implementation of planning. Further, the strong correlation between action self-e cacy and maintenance self-e cacy had been veri ed (20).
Different from the HAPA hypothesis, this study did not nd the association between risk perception and intention similarly to other studies (36,40,41). It might be related to the description of the items. In this study, the items of risk perception were described as "If you compare your current Internet use behaviors to that of individuals with the same gender and age, how likely do you think…".They might not deem that their Internet use behaviors were worse than that of same-gender peers, nor they would encounter the risks mentioned in the items (42). Risk perception having little effect on intention and behaviors have been reported by previous studies applying the HAPA model in different behaviors (20,36,41,43). Behaviors such as taking medicines and vaccination could reduce the risk of diseases, closely associated with risk perception. On the contrary, the risks brought by daily lifestyles (including Internet use behaviors, physical activity, and diet) were not easy to be perceived, which may attenuate the impact of risk perception on behaviors in the general population. It was di cult for adolescents to perceive the risk of excessive Internet use, for whom the comprehension and perception abilities were slightly lower than those of adults.
It is worth mentioning that the ideal results of interventions should be obtained from the comparison between the intervention school and the control school. However, because of the heavy academic pressure, the control school had added a strict system of the residence in the research process, and forbidden students to bring mobile phones and any other electronic products into campus. And students can go home for only 2-day holidays each month. While the experimental school didn't have such rules. Therefore, it is impossible to directly compare the change degree of excessive Internet use in the experimental school with the control school. Nevertheless, it could be seen that the psychological cognitive factors from HAPA in the experimental school were improved more than those in the control school (Appendix Table 1).
There were some limitations to this study. First, the data came from the self-reports of the participants, and there might be reporting bias. Second, previous studies con rmed that planning could be divided into action planning and coping planning. Action planning is important for the beginning of behaviors, and coping planning is crucial to the maintenance of behaviors (44). However, this study did not re ne them. Third, in the research process, the subjects of the control school were affected by the external environment such as the strict management system, which made it di cult to directly compared the status of excessive Internet use in the experimental school with the control school. It is hoped that future researches can improve this de ciency in this study and obtain more realistic results to illustrate the effectiveness of interventions based on the HAPA model.
Despite these limitations, the strengths of this study could not be ignored. It was the rst study applying the HAPA model to intervene in adolescents' excessive Internet use behaviors, which could provide new ideas for understanding the cognitive process of Internet use behaviors in adolescents. Further, longitudinal data in this study were used for analysis to better illustrate the continuous process of cognitive, intention, and behavioral change. Since the transformation of intention, planning and behaviors takes time, the longitudinal design was able to describe the transformation process. Besides, targeted intervention measures were formulated based on the HAPA model, and a positive effect had been obtained in the Internet use behaviors of adolescents. This study not only veri ed the effectiveness of the HAPA model in guiding behavioral interventions but also proposed several effective measures to improve rural adolescents' excessive Internet use behaviors.

Conclusion
This study applied the HAPA model as the theoretical framework to examine its applicability and effectiveness for improving excessive Internet use behaviors in Chinese rural adolescents. The results of con rmatory research revealed that the HAPA model was effectively veri ed in Internet use behaviors in rural adolescents.
The results of intervention research suggested that interventions based on the HAPA model could effectively improve excessive Internet use behaviors in rural adolescents, mainly through the enhancement of outcome expectancy, intention, and maintenance self-e cacy.