Registry-based analysis of participator representativeness: a source of concern for sickness absence research?