Department of Psychology and Health Science Research Institute, Warwick Medical School, University of Warwick, Coventry
Department of Oral and Dental Science, University of Bristol
Department of Psychology, University of Warwick, Coventry
Department of Community-based Medicine, University of Bristol
Department of Child and Adolescent Psychiatry, Institute of Psychiatry, Kings College London
Institute of Health Service Research, Peninsula College of Medicine and Dentistry, Exeter
Department of Psychology, University of Warwick, Coventry, UK
Correspondence: Dieter Wolke, Department of Psychology, University of Warwick, Coventry CV4 7AL, UK. Email: D.Wolke{at}warwick.ac.uk
The UK Medical Research Council, the Wellcome Trust and the University of Bristol provide core support for ALSPAC. This research was specifically funded by the Health Foundation to D.W., R.G., Jean Golding and Mike Beveridge (Grant 265/1981).
Background
Participant drop-out occurs in all longitudinal studies, and if systematic, may lead to selection biases and erroneous conclusions being drawn from a study.
Aims
We investigated whether drop out in the Avon Longitudinal Study of Parents And Children (ALSPAC) was systematic or random, and if systematic, whether it had an impact on the prediction of disruptive behaviour disorders.
Method
Teacher reports of disruptive behaviour among currently participating, previously participating and never participating children aged 8 years in the ALSPAC longitudinal study were collected. Data on family factors were obtained in pregnancy. Simulations were conducted to explain the impact of selective drop-out on the strength of prediction.
Results
Drop out from the ALSPAC cohort was systematic and children who dropped out were more likely to suffer from disruptive behaviour disorder. Systematic participant drop-out according to the family variables, however, did not alter the association between family factors obtained in pregnancy and disruptive behaviour disorder at 8 years of age.
Conclusions
Cohort studies are prone to selective drop-out and are likely to underestimate the prevalence of psychiatric disorder. This empirical study and the simulations confirm that the validity of regression models is only marginally affected despite range restrictions after selective drop-out.
Related articles in BJP: