Differential postmorbidity mortality in observational studies of risk factors for neurologic disorders.

Selection bias may be introduced in case-control or cross-sectional studies, and the impact on the observed associations may be dramatic. Many authors have examined this issue primarily in the context of subject source (e.g. referral bias). The potential bias encountered when a risk factor associated with outcome is also associated with an increase in mortality among cases greater than that among those not developing the disease (e.g. selective survival) is examined here. The potential for selective survival bias arising in observational studies is demonstrated in studies of the etiology of neonatal seizures and Parkinson's disease.