Study design
This cross-sectional study was conducted in October 2019 in Minhang District, Shanghai.
Setting and recruitment
For an exploratory factor analysis on a 27-item scale, the sample size must be at least 10 times the number of items [19]; thus, a sample size of 270 is required in this study. According to the sample size calculation method of the cross-sectional study, with the standard deviation of oral health behavior intention set at 0.62 [20], an allowable error of 0.062 and a significance level (alpha) of 0.05, the sample size required in this study was 385. therefore, choose a higher sample size requirement among the two, the final sample size of this study was not less than 385. The inclusion criteria were as follows: 1) students in the third grade; 2) no cognitive impairment; and 3) no communication disorder. A convenient cluster sampling method was adopted because of the limitations in resources and the availability of personal contacts. There are 67 primary schools in Minhang district, with approximately 826,000 students. Recruitment invitations were sent to five primary schools in Minhang district, and one primary school, Minqiang primary school was recruited on a first-come, first-served basis. Finally, with the active cooperation of the schools and reasons for cluster sampling, 10 classes were recruited out of a total of 12 classes in the third-grade. There are a total of 426 students in these 10 classes, an adequate sample size, and each student has completed a questionnaire. Informed consent was obtained from the students and their parents prior to the study. A pilot study was conducted before the main study involving 30 students in the third-grade from another primary school in Minhang district to assess readability and comprehensiveness of the questionnaire. According to the results of the pilot study, the students were able to understand most of the items and filled in the questionnaire independently. Approximately 2–3 words were beyond a few students’ comprehension which were replaced with easily understandable synonyms.
Data collection
Questionnaires were distributed to students according to their availability during the break time (30 min). Before students filled out the questionnaire, one investigator explained the purpose of the study and the composition module of the questionnaire. Moreover, the investigator introduced the meaning of the answer options for the Likert scale to the students by drawing the corresponding emoticons on the blackboard in the classroom. As the students filled in the questionnaire, two investigators of the research team were available in the classroom to answer any issues the students might encounter while filling in the questionnaires. After completing the survey, team members were responsible for quality control by checking the questionnaires. If the number of missing values of behavioral intention (three items) was greater than one, or the number of missing values of one factor in attitude (seven items), subjective norms (five items), perceived behavioral control (five items), oral health knowledge (eight items) and past oral health behaviors (nine items) was two or above, the questionnaire can be judged invalid. Finally, 12 questionnaires were deemed as invalid, and 414 questionnaires were included in the study.
Instruments and measures
Demographic characteristics
The demographic characteristics included age and sex.
TPB variables
The initial scale was specifically developed for this study after the research team referred to relevant literature, selected the original items, and modified them appropriately. It consists of 27 items designed to measure four dimensions: attitude (seven items), subjective norms (nine items), perceived behavioral control (seven items) and behavioral intention (four items) [see Additional file 1]. Of these, seven items of attitude and four items of subjective norms were obtained from an oral health behavior-related study [21], while the other five items of subjective norms were obtained from an oral health behavior-related, child-friendly TPB-based questionnaire [12], and seven items of perceived behavioral control, and four items of intention were obtained from a sample TPB questionnaire [22]. The translation and back-translation of the measurement scale used in this study were conducted by two proficient English speakers. The back-translated version was compared with the original English version, and differences were negotiated until the translators agreed with each other. Finally, after a discussion within the research team, the final form of the questionnaire in Chinese was established. All items were standardized on a seven-point Likert scale, ranging from 1 to 7. The mean value of each subscale was determined by dividing the total number of points on subscale items by the total number of subscale items. Therefore, the mean scores of the TPB model variables ranged from one to seven. A higher score indicated a greater level of each dimension.
Oral health knowledge
According to the Chinese National Oral Health Epidemiology Questionnaire, the index for oral health knowledge to reveal the status of oral health knowledge among participants consists of eight items. For example, “It is normal for the gums to bleed a little after brushing”. All items were scored with 1 = correct or 0 = incorrect, and a sum score was computed, the total score ranged from 0 to 8. A higher score indicates a higher level of knowledge about oral health.
Past oral health behaviors
With reference to the Chinese National Oral Health Epidemiology Questionnaire, we used nine items to measure children’s past oral health behaviors, including frequency of brushing, number of surfaces brushed, time of brushing, frequency of changing toothbrushes or replacement head, frequency of rinsing after meals, frequency of flossing, and frequency of eating sweets. Based on the existing ways of assigning weight to children’s oral health behaviors [21] and research team discussions, we initially assigned weights to the answers for each item. For example, the answer (and its weighted score) for the item “the daily frequency of brushing” is: zero (weighted score = 0), one time (weighted score = 1), and two times or more (weighted score = 3). Then, we organized two expert symposiums and invited eight experts in the field of children’s oral health to discuss the rationality of weight and put forth suggestions. Finally, the experts reached a consensus on the allocation of weights. The total score for the past oral health behaviors was added up by a weighted score for each item, which ranged from 0 to 18, with higher scores indicating higher levels of children’s oral health behaviors.
Questionnaire modifications
According to Shanghai’s Chinese curriculum standards for primary schools, students in the third-grade are required to learn about 2000 common words and be able to correctly write more than 1000 words. To ensure the readability of the questionnaire for students in the third-grade, we made appropriate modifications to the items of the questionnaire by referring to the vocabulary required by the Chinese textbook for students in the third-grade in Shanghai (readers can view the questionnaire in Supplementary Appendix 1).
Ethical considerations
Ethics approval of this study was granted by the ethics committee of the Minhang Branch of Fudan University Stomatology Disease Center in Shanghai, China.
Data analysis
EpiData Entry was used for data entry and documentation. The Statistical Package for the Social Sciences (SPSS version 21.0) was used for data management and analysis. Descriptive statistics and frequency distributions were calculated to describe the sociodemographic profiles of the study participants. Construct validity was examined using exploratory factor analysis. A principal component analysis with a varimax rotation was applied to extract the factors. A hierarchical multiple regression was used to explore the relationship of independent variables with the intention to oral health behaviors. Based on strong theoretical and logical framework, hierarchical multiple regression allows the researcher to accumulate input independent variables. Hierarchical multiple regression analysis for the intention to improve oral health behaviors was performed by entering independent variables into 3 separate models cumulatively. Model 1 consisted of sex and age, which were defined as the demographic characteristics and was entered first. Model 2 was comprised of scores from the attitudes, subjective norms, and perceived behavioral control; these were TPB variables. Model 3 consisted of scores from the oral health knowledge and past oral health behaviors; these were extra extended TPB variables. This sequential entry order of variables was based on two priori hypotheses, first, the additional variance may be explained by extended TPB variables after accounting for the variance related to individual factors. Second, the additional variance may be explained by oral health knowledge and past oral health behaviors after accounting for the variance related to individual factors and TPB variables. The model fit of a single regression model was determined by the value of the coefficient R2 of the model (P < 0.05). The model fit of the hierarchical multiple regression was judged by the change in R2 (P < 0.05). All tests were two-tailed.