Representativeness of child controls recruited by random digit dialling.

Recruiting control subjects who are representative of the population from which the cases are drawn is a challenge in case-control studies. This paper examines the performance of random digit dialling (RDD) in obtaining a control sample, and the sample's representativeness of the population with respect to socio-economic status. The study subjects were recruited from 2003 to 2006 for a national, population-based case-control study investigating causes of acute lymphoblastic leukaemia (ALL) in children <15 years of age in Australia. Control families' addresses were linked to Australian Bureau of Statistics Census 2006 Collection Districts and thus to Socio-Economic Indexes for Area scores, which are area-based measures of socio-economic status. These scores were compared with those of all collection districts where families lived. We estimate that 55% of eligible families in the RDD sample agreed to participate in the study. Participation was directly related to socio-economic status with those of highest economic status most likely to participate. Completeness of participation in the components of data collection was similarly related to socio-economic status. This evidence of selection according to socio-economic status indicates that there may also be selection with respect to other factors potentially important in the aetiology of ALL.