Text this: Does attrition during follow-up of a population cohort study inevitably lead to biased estimates of health status?