Primary schools
The schools included in the present study were participating in a research project exploring school absence in primary education in the West-Brabant region of the Netherlands. The number of schools approached was based on a power analysis carried out for another study in that research project, for which 10 MPSs were needed.
Regular education schools were included in this study. Special needs schools were excluded as these are intended for pupils with severe physical or psychiatric problems [
30], which could seriously influence attendance patterns. A random sample of 16 out of 265 MPSs in the region was selected using a random sample of cases procedure in SPSS. Ten of these schools agreed to participate, eight of which were able to provide data on absence. Seven of the MPSs also provided data on pupil characteristics. All seven SSPEs in the region were asked to participate in the study, six of which agreed to participate and provide all data.
The eight participating MPSs had a total of 2216 pupils at the end of the school year. Three SSPEs did not supply the total number of pupils at the end of the year meaning that the total number of pupils in these three SSPEs at the end of the school year had to be estimated. This was done by taking the official total number of pupils in October 2015 and adding the average increase in pupils (9%) found in the other three SSPEs. This resulted in 24 additional pupils bringing the total estimated number of pupils in the six SSPEs to 1000.
The median age of pupils in the eight MPSs was 7.4 years. In the municipalities where the eight MPSs were located, 50% of pupils were boys. The median age of SSPE pupils was 9.4 years and 64% were boys [
26,
31].
Measures
The participating schools used a digital school absence registry to record each pupil’s absence daily. The school year 2015–2016 was analysed retrospectively. The school absence registry only contained those pupils recorded as absent, meaning that the number of pupils who were not absent in the chosen school year was not recorded. In order to determine the one-year prevalence of absence, the total number of pupils attending the school at the end of the school year was used.
Three types of authorised and two types of unauthorised absence were categorised: authorised comprised sickness absence, doctor’s visits and other authorised absence (such as family holidays or events requiring approval from the principal); unauthorised comprised tardiness and all other unauthorised absence, e.g. truancy.
When reported sick, pupils were labelled as sick either
occasionally or
extensively based on the definition of extensive sickness absence by Vanneste et al. of more than nine school days or more than four periods in a school year [
29]. A period of absence is a single continuous span of time during which a pupil is absent. As soon as a pupil is registered as back in school, this period ends.
The frequency and duration (in half days) of all types of sickness absence, other authorised absence and other unauthorised absence were analysed. Additionally, the absence rate each of these absence types was determined, based on an estimated total of 180 possible school days in a school year. The absence rate is the ratio of absence days to possible school days. The duration of doctors’ visits and tardiness is not recorded by schools, therefore, only the frequency of these types of absence was analysed.
The month and year of birth, sex, years, and parental educational score of MPS pupils were collected from the school absence registry. Only the sex and the date of birth of pupils were available from SSPEs.
Age was calculated at the end of the school year based on the pupil’s month and year of birth.
For years MPS groups were made by combining lower years (Dutch school years one and two when pupils are normally four or five years-old), middle years (three, four and five) and senior years (six, seven and eight). Several schools had combination classes with different years in one class. These were allocated to the group of the highest year in each combination class.
The
parental educational score was based on the parents’ highest educational achievement [
32]. It was converted into a binary variable: category zero for parents with an education up to, or the equivalent of, prevocational education in the Netherlands, and category one for parents with a higher educational achievement than prevocational education.
Analysis
Due to the variation in selection methods, data from MPSs and SSPEs were analysed separately. χ2 and Mann-Whitney U tests were used to analyse differences in occurrence of absence in MPSs and SSPEs. Univariate and multiple logistic regression analyses were used to determine the association between extensive sickness absence and (i) pupil characteristics (ii) other types of absence, and compared with occasional sickness absence.
The data were structured hierarchically, with pupils (first level) nested within schools (second level). In order to test if it was necessary to control for school effect in the analyses, the intra-class correlation coefficient (ICC) was determined [
33,
34]. The ICCs for sickness absence ranged from .02 for MPSs to .07 for SSPEs, thus less than 8% of the variation in sickness absence in this sample was due to differences between schools, indicating that controlling for school effects was unnecessary.