The main basis for the analyses are the data from the medical check-up of the cohort starting school in 2004, provided by the Department for Health and Environment of the City of Munich, Germany. The city of Munich was chosen for several reasons: The results from the medical check-ups are available in a data set (which is not the case for all communities in Germany); some information on the social structure is available per school district (which is rarely the case in Germany); Munich is a prosperous city and public awareness of health inequalities is rather low. The data set includes the whole age cohort; every child has to be presented to a school nurse before starting school. If any health problems are observed by the nurse, a school physician is consulted as well [22
For the school year starting in 2004, a total of 9,883 children were presented to the school nurse [23
]. We excluded those children who could not be clearly linked to a school district (missing or wrong school district number). Children signing in for the German-French school were excluded too because this school (being a private school) does not have its own school district. Finally, we could include 9,353 children (94.6%).
The dataset includes just a few variables on health related factors for each child. We focussed our analyses on those that were assessed in a rather objective way: overweight (defined by the Body-Mass-Index, BMI), participation in preventive health check-ups for children (in Germany called 'U1 to U9'), and vaccinations. Other indicators included in the check-up (such as impaired vision or hearing disorders) were disregarded, as the number of cases per school district is very low, and as they were assessed on a rather imprecise way.
For adults, overweight is commonly defined by a simple BMI value such as 25 [24
]. For children, age- and sex-specific limits should be used. For Germany, these limits have been defined by Kromeyer-Hausschild et al. [25
], based on several studies conducted between 1985 and 1999 in different German regions. We have chosen this definition as the study focuses on children living in Germany. We have also conducted additional analyses based on a definition that is more common in the international literature, i.e. the definition proposed by Cole et al. [26
]. The results were very similar, though, and are not reported below. The limit for overweight is defined as '90% or above' the age and sex specific distribution. We applied this limit and defined two groups (i.e. normal weight, overweight). Body height and weight were measured during the medical check-up, probably yielding reliable data.
The nine preventive health check-ups 'U1 to U9' are offered to every child between birth and six years of age. The aim is to identify any health problem of the child, thus giving some information on the 'health awareness' of the parents. The costs are covered by the statutory health insurance, i.e. there are no extra out of pocket payments for the parents. It is strongly recommended to visit all nine check-ups; the consultations are documented in the so called 'yellow booklet'. This information can be included here, as the parents are asked to show this booklet to the school nurse. It is assumed that parents who presented their children to all check-ups are more concerned about the health of their children than other parents. Two groups are differentiated for the analyses: a) children visiting all nine preventive check-ups; b) children missing at least one of these check-ups.
Concerning vaccinations, we focused only on those recommended by the German 'permanent vaccination commission' of the Robert Koch Institute, i.e. measles, mumps, rubella, diphtheria, hepatitis B, meningitis, tetanus, pertussis, polio [27
]. Children having received all nine vaccinations are categorized as 'completely immunized'. If one or more vaccination is missing, the child is placed in the category 'not completely immunized'.
These three dependent variables are combined with social variables on the individual and on the regional level. On the individual level, the dataset includes:
- sex of the child (boy/girl)
- mother tongue of the parent
- child had visited the Kindergarten for at least twelve months (yes/no)
The mother tongue of the parents is coded for the mother and the father. It is not further specified if this language is the only spoken language by the parent but it can be seen as a more reliable indicator for the migration background than the indicator 'nationality', as having a German passport does not necessarily imply good understanding of the German language. For this analysis we distinguished two groups of children: both parents 'German language', at least one parent 'other language'.
Concerning the school districts level, data for two variables could be acquired from the Statistical Office of the City of Munich; they can be understood as surrogates for variables not available at the individual level:
- percentage of single-parent-households (grouped into tertiles: low, medium, high percentage)
- percentage of households with at least one adult having the lowest educational level (grouped into tertiles: low, medium, high percentage)
The German school system differentiates between three main educational levels. The lowest is the 'Hauptschulabschluss' which is usually gained after nine years of school education. The next is the 'Mittlere Reife' and the highest is the 'Abitur'.
The statistical analyses were calculated with the SAS software (Version 9.1). To identify associations between the dependent and independent variables, we first calculated cross tables (including chi2
-tests). Then we performed the following multivariate analyses: For each of the three dependent variables logistic regressions were calculated. The first model only includes the individual social variables (i.e. sex of the child, mother tongue of parents, Kindergarten visit of the child), and the second model only includes the regional social variables (i.e. percentage of single-parent-households, of households with lower educational level). In a third step, all independent variables were included simultaneously in a multilevel-analysis (on 'level 1' the three individual social variables, and on 'level 2' the two regional variables). A multilevel analysis is necessary for controlling auto-correlations when the data can be split into different 'levels' [28
]. The statistical analysis was performed with SAS using the glimmix procedure (with random intercept). In order to gain results that can more easily be communicated to policy makers, we performed a separate analysis for each of the two 'level 2' variables. In addition, the regional distribution is visualized by including a thematic map, produced with the ESRI software ArcGIS (Version 9.1).