The protocol for the "Okkio alla Salute" survey developed by the National Institute of Health and reviewed in collaboration with regional coordinators and members of the technical committee is described elsewhere [16
In this study we will report data regarding the 2010 Tuscany Regional project. In our region, to guarantee the maximum level of territorial coverage, all 12 Local Health Units were invited, and all agreed to join, and collaborate in, the project. Once enrolled, all 12 Local Health Units met for an explanation of the protocol and to arrange the operational formalities of the activities in Tuscany.
Overall, opt-out consent was signed by 95.5% (1983 children) of the parents of the children enrolled in the 99 classes selected; 91.3% (1811 children) of all invited children were present on the day of the study. After data cleaning, 60 more children were discarded (34 out of age range or with missing age); the remaining 88.3% (1,751 children: 922 males and 855 females) out of all enrolled children were analysed.
Students were selected by a stratified one-stage sample with classes as clusters of students and primary stratification by relative health district and municipality size. The number of children to be enrolled was estimated on the basis of an expected prevalence of overweight and obesity of 30%, a precision of 3% using a 95% confidence intervals, and a design effect of 2 [16
Specifically trained personnel using appropriate and standardised instruments measured the children's height and weight. To measure the anthropometric values, we followed the recommendations of the World Health Organization [17
We used electronic scales with a liquid crystal display that measured every 100 grams of weight. Height was measured with a portable stadiometer, with a precision of 0.1 cm; exact decimal age was calculated from the date of birth and day of measurement; BMI was then calculated from weight and height, using the following formula: weight (kg)/height (m²). BMI classes of the children were set using the Cole et al
. Method [18
]; this allowed us to have specific cut-off points for males and females at every age as recommended by the International Obesity Task Force (IOTF). We thus obtained six classes of BMI: thinness grades 3, 2, 1, normal weight, overweight and obesity. According to the Cole's definition, the term "underweight" in children includes thinness grades 3 and 2 (underweight group) while thinness grade 1 and normal weight go into another class (normal group) [19
] as shows in table .
Nutritional status of 8-9 year old school-children
The weight, height and educational level of both parents were recorded in a self-administered questionnaire. Weight (kg) and height (m) were used for the calculation of the BMI (kg/m²); the parents' BMI classes (underweight, normal weight, overweight or obese) were established using international cut-off points for adults [17
]. For 132 (7.4%) mothers and 202 (11.4%) fathers the BMI calculation was not possible due to lack of the necessary data.
Three levels of education were then established: high (university degree), medium (secondary school diploma) and low (middle school, elementary school or none).
Data were analyzed using SPSS (ver.16) and EpiInfo (ver.3.5.1.). Descriptive statistics (e.g., mean, proportion, standard deviation) were used to describe the characteristics of the sample. The χ2-test and χ2-test for trend were used to explore the relationship between: a) children's BMI classes and parents' BMI classes, b) children's BMI classes and parents' educational level.
The Kendall's τb coefficient was used to calculate the measure of association between children's BMI classes and parents' BMI classes and children's BMI classes and their parents' educational level.
The association between children's obesity and their parents' educational level was estimated comparing the Prevalence Ratios (PR), of children whose mothers and fathers had a low educational level with that of children born from mothers and fathers with a high educational level.
All analyses were carried out using the C-Sample routines for complex survey design.