Decision-making in refractory epilepsy: an analysis based on computer simulation

A program developed to simulate management of patients with recurrent seizures was tested in 24 subjects, including 13 experienced neurologists. Identifiable errors included excessive reliance on plasma anticonvulsant levels, misinterpretation of random effects, inefficiency in selection of follow-up intervals, and undue reference to arbitrary data, such as the initial dose, in the selection of "optimum" drug dosage.