This cross-sectional study was carried out on 906 twelve-year-old schoolchildren in Kerman, Iran in 2012. In Iran, trained school nurses regularly examine the health information for pupils in primary schools. The dental examination was carried out by school nurses for detection of dental caries in accordance with World Health Organization criteria [22]. The dental examination was noninvasive and x-rays were not used. These health records were used and socio-demographic information was recorded for selected pupils in 2012. The subjects were selected through a stratified cluster random sampling. At first, the whole target area was stratified according to two geographic areas, the north and south area; the northern area generally has a better economic situation than the southern area. Then each area was stratified according to gender. In the next step, several schools were considered as clusters, which were selected randomly in each geographic area and gender strata; all the schoolchildren in these schools were included in this study as samples. The subjects were from ten boy schools and nine girl schools, representing 10% of the target population.

In this research, the dmft/DMFT index was used as the dependent variable because it defines the condition of caries in children [22]. This index was calculated as the sum of total decayed (d/D), missing (m/M) and filled (f/F) teeth for primary and permanent dentitions, respectively.

Pupil’s BMI (Body Mass Index) was calculated as weight in kilograms divided by the square of the height in meters (kg/m²). Standardized percentile curves of BMI in Iran were used. According to these curves underweight condition is defined as under the 5th percentile curve, normal as between the 5th and 85th percentile, at risk of overweight as higher than the 85th and lower than the 95th percentile and overweight as higher or equal to the 95th percentile [23].

The categorical variables for pupils, studied here, included gender (boy/girl), family size (three, four and five or more) and BMI (underweight, normal, at risk of overweight and overweight). Parental data included father’s and mother’s education (high school degree or lower/ university), father’s job (employee, worker and self-employed), mother’s job (employee/housewife) and mother’s age at delivery (less than 35/more than 35 years).

For assessing inter-examiner reliability (i.e. school nurses), three school nurses and twenty pupils were selected. Then, dental examinations were carried out by each school nurse on all the twenty pupils. The intraclass correlation (ICC) between the results of examinations (dmft/DMFT) for school nurses was 0.91, demonstrating a high agreement rate between school nurses. In addition, for assessing validity, a dentist carried out dental examinations on the same twenty pupils. Then the results were compared with those of school nurses. Repeated measures analysis of variance did not show any significant differences between the dentist and school nurses in relation to dental examinations (ρ = 0.26). In addition, Pearson’s correlation coefficients (ρ) between the dentist and each school nurse in relation to dental examinations was high (ρ = 0.92 for Examiner 1, ρ = 0.91 for Examiner 2 and ρ = 0.91 for Examiner 3), indicating that the results of examinations by school nurses were valid.

Frequency and percentages of caries-free (dmft/DMFT = 0) pupils and categorical variables were reported. The mean ± SD for caries severity, dmft/DMFT index, among subjects with any caries was reported. A multilevel negative binomial hurdle model (MNBH) was used to investigate the effect of the variables evaluated on dental caries, which has recently been suggested as an approach that gives a better fit to these types of count data [19].

A backward elimination (B.E) approach was used to exclude independent variables from the adjusted model. The final model contained only factors that were significant at a significance level of p ≤ 0.05. The odds ratio (OR) and its 95% confidence interval were reported in relation to the logistic component of the model, corresponding to the caries-free state. The incidence rate ratio (IRR) and its 95% confidence interval were reported in relation to the truncated negative binomial component of the model, corresponding to caries severity, i.e. dmft/ DMFT index [19].

The Akaike information criterion (AIC) index was used to compare the model used here with Poisson hurdle model, as well as with Poisson and negative binomial regression model. Data were analyzed using programs written with SAS 9.2.