We used data from the first and second Dutch national surveys of general practice, which were performed by the Netherlands Institute for Health Services Research (NIVEL) in 1987 and 2001. In the Netherlands, general practices have a fixed list size, and all non-institutionalised inhabitants are listed in a general practice, and GPs have a gate-keeping role. Usually, the first contact with health care, in a broad sense, is the contact with the general practitioner. Each survey included a representative sample of the Dutch population.
In 1987 practices were randomly sampled from a list of all Dutch practices, per stratum defined by region and degree of urbanization. Sampling fractions differed between strata. 161 GPs in 103 practices participated in the first national survey [11
]. With respect to age and gender the participating GPs and practices were representative of Dutch GPs and practices in 1987. The GPs were divided into four groups, and each group used registration forms to register data (e.g. diagnosis, prescription and referrals) on all contacts between patient and practice during one of four consecutive 3-month periods. Baseline characteristics such as age and gender were derived from patient records. Other socio-demographic characteristics such as socio-economic status (SES) and ethnicity were obtained by a questionnaire and filled out by parents, or by the children themselves if they were older than 12 years (response rate 91.2%). SES was based on the father's occupation, which was categorized into five classes "non-manual work high (class I)", "non-manual work middle (class II)", "non-manual low and farmers (class III)", "manual work high / middle (class IV)" and "manual work low (class V)". Ethnicity was derived from the country of birth of either parent. If either parent was born in Turkey, Africa, Asia (except Japan and Indonesia) and Central or South America, their children were considered to be children of non-Western origin (in accordance with the classification of Statistics Netherlands). All other children were defined as Western. The degree of urbanization was derived from the general practice's postal code and categorized into four classes 'under 30,000 inhabitants', '30,000–50,000 inhabitants', 'over 50,000 inhabitants' and 'the three large Dutch cities Amsterdam, Rotterdam and The Hague'. The Netherlands were divided into a Northern, Central and Southern region. Season was divided into four categories: spring was defined as months April-June, summer as July-September, autumn as October-November and winter as January-March.
The diagnoses made by the GPs were coded afterwards by clerks using the International Classification of Primary Care (ICPC) [12
In 2001, 195 GPs in 104 practices registered data about all physician-patient contacts over 12 months [13
]. They registered all health problems presented within a consultation, and coded the diagnosis themselves using the ICPC. Patient demographic characteristics such as age and gender were derived from the GP's computerized patient files. As in 1987, SES and ethnicity were obtained by a questionnaire (response rate 76%). Degree of urbanization, region and season were derived as in 1987.
In both surveys each contact with the GP was defined as one consultation. All health problems presented within one consultation were recorded separately. Both surveys were episode orientated, meaning that a consultation on a new health problem marked the beginning of a new episode. If there were multiple consultations in a single episode, the diagnosis made during the last consultation was regarded as the episode-diagnosis. In order to decide whether two consultations with the same problem belonged to the same episode or were different episodes, the latter was arbitrarily decided upon if the interval between two consultations was at least four weeks (28 days).
There were 20 practices that participated in both surveys. In 2001 eight practices were excluded from analyses for the following reasons: two practices had software problems; one practice registered only over a three-month period; five practices showed insufficient quality of the morbidity registration.
The study was carried out according to Dutch legislation on privacy. The privacy regulation of the study was approved by the Dutch Data Protection Authority. According to Dutch legislation, obtaining informed consent is not obligatory for observational studies.
This study analyzed data from both surveys for children aged 0–17 years presenting with skin diseases, classified by ICPC codes. Incidence rates in general practice were calculated for all combined skin diseases and for each skin disease separately with a distinct ICPC code. We calculated the incidence rate by dividing the total number of new episodes (numerator) by the study population at risk multiplied by the follow-up time (denominator). In 1987 the denominator was calculated by multiplying the number of all patients listed in the participating practices by the follow-up time (person years). In 2001, persons that moved into or out of the participating practices during the registration period were assumed to contribute for half a year to the follow-up time. The so-called mid-time population was calculated as the mean of all listed patients of all participating GPs, aged 0–17 years, at the beginning and at the end of the registration period, irrespective of health care use. Data were stratified for age categories, gender, urbanization level, region, season, SES and ethnicity.
Further we assessed the changes in incidence rates of all skin diseases between 1987 and 2001. Incidence rates were expressed per 1000 person-years; 95% confidence intervals (CI) were calculated assuming a Poisson distribution. Skin diseases which contributed less than 0.5 percent to the total skin morbidity were not analyzed in detail and were combined into one residual group.