Does Field Substitution Impact the Educational Profile of the Belgian Health Interview Survey Net Sample?
Abstract Background Although controversial as a sampling technique, field substitution of non-respondents is applied in the Belgian Health Interview (BHIS) since its start in 1997. The target number of participants to obtain is predefined and set at 10,000 individuals. Based on data derived from the National Register, non-participating households are substituted by at most three households matched on statistical sector, age group of the households’ reference person (administrative contact of the household) and household size, thus creating a cluster. In this study, the impact of field substitution on the educational composition of the net sample is assessed. Methods The educational level of the household’ reference person derived from the Census 2001 and Census 2011, was used as a proxy for socio-economic position and was linked with respectively BHIS 2001 and BHIS 2013 paradata on the use of field substitution using a unique identifier. Given the high level of missing data on the educational level (+/-16%) in the Census, regression based multiple imputations (m=5) procedures were applied, presuming missingness at random. Response rates by educational level at any stage of the substitution process stage were calculated. Differences in response rates were assessed by applying the Delta method. Results At any stage of the substitution process, the participation rate was the lowest in the lowest educated households and significantly higher in the middle and highest educated households. Throughout the substitution process, the participation rate dropped from 51.6% to 42.7% for low educated households and from 61.7% to 46.3% for high educated households. Conclusions It is concluded that field substitution introduces higher levels of non-participation but does not affect the educational composition of the net sample.