Participation bias in longitudinal studies: experience from the population study of women in Gothenburg, Sweden