Survey data and statistical analysis

This study uses survey data from farms in the Farm Accountancy Data Network (FADN) and additional veal calf farmers.

Survey data

The FADN contains a stratified sample of around 1,500 agricultural and horticultural farms in the Netherlands (Vrolijk et al., 2009). Records are made of the economic data and technical key figures of these farms. Every year a number of farms are replaced by other farms to ensure that the database of the Data Network remains representative of Dutch livestock farming. On these farms all animal-medicine data and veterinary services are recorded. This provides information about the true exposure of farm animals to antibiotics, and gives insight into the underlying factors that could explain changes in antibiotic use. In cooperation with the veal calf sector the use in veal calves is monitored in an additional random sample. On the veal calf farms detailed data were collected on the number of animals present and the amount of antibiotics used.

To ensure that the farms in the sample are representative of the whole population and to make the sampling as efficient as possible, a disproportional stratified random sampling strategy is used (Vrolijk et al., 2009). A stratified sample implies that the population is divided into a number of homogeneous groups. Subsequently, farms are selected from each of the groups. For strata with larger variation in the use of antibiotics, relatively more sample farms are selected. In the FADN sample the strata are based on both farm size and animal category. The additional sample of veal calf farms is additionally stratified for 'large integration' versus 'small integration or non-contracted farms'.

Data of 187 pig, broiler and dairy cattle farms in the FADN were used to estimate the antibiotic use in 2012, based on the first semester. As from 2007, data from veal calf farms were collected in an additional sample. See Table 2.1 for details.

Table 2.1

Statistical analysis

Data from the FADN farms and the additional sample for veal calves are used to estimate the use in the whole population to obtain insight into the amount of and trends in antibiotic use on the national level. Antibiotic use per species on the national level is expressed as average number of daily dosages per average animal present on an average farm. Since the stratification is disproportional, the results have to be weighted to be representative. For each stratum the average daily dosages per animal year is determined. Then the weighted average for an animal category is calculated, based on the number of farms in the population in each stratum.

The aggregated usage data are considered to be representative for the total exposure of Dutch food-producing animals to antibiotics. The 95% confidence intervals (CI) indicate that with 95% certainty, the average antibiotic use per animal on a national level, expressed by the number of daily dosages per animal year, will lie within the upper and lower limits given. The confidence interval also indicates the variation in antibiotic use amongst farms.

In this report the usage data of all sample farms are used for statistical analysis on not only antibiotic use in each year, but also on the changes in antibiotic use over a period of two or more years. Comparing average uses of antibiotics between two years can be done in two ways: one using only farms that are in the sample for both years,  the other using all farms in the sample in both years. The first method usually gives better results if the majority of the farms are in the sample for both periods. This usually is the case in two consecutive years. However, if the years of comparison are further apart, the number of sample farms available in both years will be more limited. In that case, testing for significant differences can better be done by using all farms in the sample to increase the statistical power of the comparison.