Study design and study population
The study was designed as a cross sectional survey conform to the STROBE guidelines; data were obtained from the second Italian Pathfinder on oral health, a population-based cross-sectional survey of Italian children aged from 4 to 12 years [4]. The survey protocol was approved by the ethical committee of the University of Sassari (Italy) (AOUNIS: 29/16). In 2017 the Italian population was of 60,589,445 (29,445,741 males and 31,143,704 females); 1.62% were of 3–4 years old.
Sample procedure and data collection were previously described [4].
A multistage cluster sampling was performed as previously reported [14], considering the Italian sections as strata: North-West, North-East, Central, South and Insular Italy. In the second stage the counties of the sections and then the schools (kindergartens) were chosen at cluster level with proportional random selection of participants. A sample size for each stratum was calculated based on an assumed prevalence of dental caries [14] with a standard error of 0.05 and a design effect of 2.5 [15]. The number of 5,100 Italian children aged 4 years was increased by 10%. This strategy provided a sample that was self-weighting. In total, 7,051 children were recruited and 6,825 were examined; 173 children with no parental consent and 53 not present in the classroom at the moment of the examination were excluded. The study sample represented 1.16% of the total Italian population aged 3–4 years attending kindergartens (585,985 children with a frequency of 59.7%.) [13].
Subjects were examined at school by calibrated examiners [16], using a plain mirror (Hahnenkratt, Königsbach, Germany) and the WHO ballpoint probe (Asa-Dental, Milan, Italy) under artificial light. Caries data was recorded using the two-digit codes related to ICDAS for each tooth surface.
Clinical outcomes
Two units of analysis were selected: subject and tooth. The frequency of caries was the outcome variable analyzed by subject, while the most severe ICDAS score observed was the variable used at tooth level. Each tooth was recorded as caries-free (ICDAS 0), enamel lesion (ICDAS 1/2), pre-cavitated lesion (ICDAS 3/4), cavitated lesion (ICDAS 5/6), filled teeth due to caries (Ft) and missing teeth due to caries (Mt). At subject level, the sum of teeth caries affected (ICDAS ≠ 0), filled or extracted due to caries was calculated as Caries Experience (CariesEx).
Socioeconomic status/explanatory variables
European children were defined as children whose both parents were born in one of the European countries, while non-European children were coded as those whose at least one parents born outside Europe. Children were stratified based on their Socioeconomic status and behavioral habits of children/parents/caregivers (prolonged breastfeeding, the use of pacifier at night, brushing frequency, the frequency of a cariogenic diet and smoking habit of the parents) were estimated by mean a standardized self-submitted questionnaire [17, 18] before the clinical assessment.
Statistical analysis
Data were analyzed separately for European children and those with an immigrant background. Caries experience was first analyzed through cross-tabulations and adjusted for age and sex through regression analysis. Both the questionnaire and oral examination forms were manually checked for the completion of the required information.
An algorithm about the household income was developed derived by the modified Organization for Economic Co-operation and Development (OECD) equivalence scale (https://www.oecd.org) [19]. The household was assembled via allocating scores to each person in a household and then summing up the equivalence points of all household members [14]. The factor was then categorized in 4 levels, with quartile 1 being the lowest and quartile 4 being the highest. The five Italian areas were ordered following the mean Gross National Product (GNP) per capita starting from the highest (North-West, North-East, Central, South, Islands). As proxy measurements of family’s socio-economic status, the occupational profile and the educational level of both parents and the number of offspring were recorded. The parents’ occupational profile was categorized into 3 levels: un-employment/unskilled jobs as the lowest, skilled jobs/qualified jobs and white-collar jobs as the highest. The educational level was categorized in Low (no education/primary education/lower secondary education), Intermediate (upper secondary) and High (degree, master or doctorate) [20]. Oral health habits (frequency of toothbrushing and fluoride supplements beyond toothpaste), diet (duration of breast-feeding, use of sweetened pacifier at night) and life-style behaviors (parents/guardians smoking habits) were also collected. For each explanatory variable, the group with the most favorable situation was chosen as a reference point from which to measure disparities. The regression between the caries experience (CariesEx) in each household quartile was computed and the Slope Index of Inequality (SII) [21], as of the mid-point value of CariesEx score in each household group, was calculated.
As a proxy for health inequality, a social gradient was generated as the weighted sum of the worst circumstances deriving from social explanatory variables. Children were stratified into four social gradient levels based on their number of worst options: ‘‘best,’’ with up to two; ‘‘good,’’ with three; ‘‘bad,’’ with four; and ‘‘worst,’’ with five or more. Multivariate regression models were used to elucidate the associations between all explanatory variables and health outcome (namely the caries prevalence). Assuming an overdispersion and excess of zero in total sample and in children with European background, the Zero-Inflated Negative Binomial logistic regression was used; while, a logistic regression was used in children with non-European background, assuming a high caries prevalence. Different additional analyses were performed and Mantel Haenszel trends of odds adjusted by immigrant background and area of living were calculated to study the existence, dimensions and direction of the social gradients in oral health.
Stata/SE 16.1 for Mac (Intel 64-bit) and SPPS 27 were used to account for the complex data structure.