Observational data bases in neurological disorders: selection bias and generalization of results

Observational data bases in the neurological disorders may be an important source of information on etiology and natural history. The ability to make valid generalizations of results from a data base to the general population of patients with a particular disorder is not, however, guaranteed. The design or analysis of the study may result in selection biases that make inferences untenable. Examples are presented which demonstrate the potential for bias limiting the validity of inferences beyond the sample observed.