Method
This cross-sectional study was conducted in 2016 with the support from the Education Bureau of Yunnan Province. Ethical approval was obtained from the Kunming Medical University Institutional Review Board. Invitation letters were sent to legal guardians and written consents were collected. The report of this study follows the Statement of Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) (Additional file 1) [15].
Sample size calculation and sample selection
The sample size calculation was based on formula: n=Z2×P×(1-P)/d2, in which P was the prevalence rate of oral diseases, the Z statistic was based on the confidence level and d was the precision [16]. The prevalence rate of oral diseases (P) was estimated to be 50% to yield the maximum sample size. With the confidence level of 95% and the confidence interval of 5% (confidence interval: 45 to 55%), sample size was calculated to be 384. Estimating a response rate of 80%, at least 480 children needed to be invited in the study.
This study adopted a multistage cluster sampling method. Yunnan has eight cities and eight autonomous prefectures. Most of the Lisu population resides in the western region, which includes three cities (Baoshan City, Lijiang City and Lincang City) and five autonomous prefectures (Chuxiong Yi Autonomous Prefecture, Dali Bai Autonomous Prefecture, Dehong Dai and Jingpo Autonomous Prefecture, Diqing Tibetan Autonomous Prefecture and Nujiang Lisu Autonomous Prefecture). Only a few Lisu are scattered in the eastern region, which consists of five cities and three autonomous prefectures. The distribution ratio of the Lisu population in the western region and eastern region was 8:1 [17]. Accordingly, the ratio of children invited from these two regions was determined. At least 427 children should be invited from the western region and 53 should be invited from the eastern region. Lists of primary schools in western and eastern regions were obtained and the schools were numbered sequentially. Then, schools were selected by a simple random sampling method with a list of random numbers generated by a computer. All 12-year-old Lisu children in the selected schools were invited until the number of invitations in that region was fulfilled. The inclusion criteria were 12-year-old Lisu children who were generally healthy without special care needs, prolonged use of medications and severe chronic diseases and had parental consents. Children who were on long-term medications were excluded.
Questionnaire survey
Self-administrated questionnaires were distributed to the study children and collected with parental consents one day before the oral examination. A research assistant collected the parental consent together with the questionnaire and checked the responses in school. If the research assistant found any missing or inappropriate answers, the participating children were required to filled in or confirmed their answers on the spot. If the study children failed to answer their parents’ education levels, the research assistant would confirm the data with their parents by phone. The questionnaire was adapted from the one used in a previous epidemiological study [17]. It consisted of two parts as follows:
i. Child’s sociodemographic background information: sex, parental education levels and child’s monthly-pocket money; and
ii. Child’s oral health-related behaviours: toothbrushing frequency (daily), sugary snacking habit, sour food snacking habit and dental attendance experience.
Oral examination
Two calibrated dentists (Y.L and S.Z) conducted the oral examination in the primary schools during school hours with the aid of Community Periodontal Index (CPI) probes, dental mirrors and light-emitting diode headlights for intra-oral illumination. S.Z is an experienced dental epidemiologist. Y.L was calibrated with S.Z in the same setting before the study commenced. To assess the color of tooth crown, the children were required to remove food debris and dental plaque before the oral examination. No special toothbrushing technique was introduced. The research assistant reminded the children to keep their mouth clean on the examination day by following their daily oral hygiene practice routine.
The diagnosis criteria of dental caries followed WHO recommendations [11]. Dental caries experience was measured by the decayed, missing and filled teeth (DMFT) index. If a tooth had an unmistakable cavity, undermined enamel or detectable softened floor or wall, it was recorded as a decayed tooth (DT). If a tooth was extracted due to dental caries, it was recorded as a missing tooth (MT). If a tooth was permanently filled without dental caries, it was recorded as a filled tooth (FT). Children’s periodontal status were assessed by the presence of gingival bleeding followed the WHO recommendation and diagnosis criteria. Periodontal pockets were not assessed as the study children were younger than 15 years old [11]. Examiners inserted the tip of CPI probe into the gingival sulcus to explore the full extent of the sulcus with the sensing force of no more than 20 gram. The tip of CPI probe was first placed as close as possible to the distal contact point with the probe paralleled to the long axis of the tooth. Then the probe was move gently with short upward and downward movements along the sulcus to the mesial contact point. Gingivae of all teeth present in mouth were examined. Children who had bleeding on probing in any tooth site was diagnosed as having gingival bleeding [11]. Dean’s index criteria were used to assess dental fluorosis as recommended by WHO [18]. Enamel surface that had white flecks, occasional spots, paper-white areas or brown staining was regarded as dental fluorosis. Basic Erosive Wear Examination (BEWE) criteria were adopted in this study to assess the status of dental erosion [19]. All tooth surfaces in six sextants (17-14, 13-23, 24-27, 37-34, 33-43 and 44-47) were screened and the most severely affected tooth surface in each sextant were scored. The scoring consists of four levels: (0) no loss of the tooth surface, (1) initial loss of enamel texture, (2) distinct defect with hard tissue less than 50% of the tooth surface area, and (3) hard tissue loss of more than 50% of the tooth surface area. The cumulative scores of all six sextants were calculated to assess the patient’s risk level. There were four risk levels including none (BEWE score≤2), low (BEWE score = 3 to 8), medium (BEWE score = 9 to 13) and high (BEWE score ≥14). Tetracycline-stained teeth were diagnosed by assessing the tetracycline discoloration of the tooth’s crown in the daylight [20]. A 10% random sample of the study children were selected by a dental assistant for duplicate examination without notifying the examiners on the same day. The duplicate examination was performed to assess the inter- and intra-examiner agreements.
Data analysis
Data were analyzed by IBM SPSS version 25.0 (IBM Corp., NY, USA) and SAS® OnDemand for Academics (SAS Inst., NC, USA). The intra- and inter-examiner reliability was assessed by intraclass correlation coefficient (ICC). A chi-square test was conducted to test the association between independent variables and the prevalence of oral conditions (dental caries and gingival bleeding). The Mann–Whitney U test was performed to analyze the distribution of DMFT scores according to different independent variables because the distribution of DMFT score was not normal. This study considered the zero-inflated negative binomial regression model (ZINB) to study the relationships between the DMFT scores and covariates. The Vuong’s test was employed to test the model fit. A multivariate logistic regression model was considered to study the relationship between the prevalence of gingival bleeding and covariates. Independent variables with p-value less than 0.1 in the univariate analysis (chi-square test, Mann-Whitney U test) were studied as covariates in the regression models. Doing so can minimize the influence of the irrelevant variables and prevent missing important variables [21]. Insignificant variables were removed from the models by backward stepwise selection until all remaining variables had a p-value less than 0.05.