This review illustrates the use of computer-based Markov models to estimate cost-effectiveness and prognosis in a complex problem in clinical cardiology. Decision analysis and cost-effectiveness analysis were used to assess whether to implant a permanent cardiac pacemaker, treat with drugs, perform electrophysiologic studies or observe patients who have two clinical features--syncope and bifascicular block--that may or may not be causally related. Using a Markov process model, a computer program simulated the prognosis of five cohorts of such patients--one treated conservatively, one given empiric antiarrhythmic drug therapy, one receiving a pacemaker, one treated with empiric drugs and pacing and one tested with electrophysiologic studies. On the basis of data from published reports and expert opinion, quality-adjusted life expectancy was calculated by summing the average time a member of each cohort would survive with and without symptoms for each initial treatment choice. The costs were estimated from 1985 hospital charges. For patients with normal left ventricular function, electrophysiologic testing provides a benefit of 14 quality-adjusted months of life over observation, at an additional cost of $24,200. Empiric pacing would add 2.5 additional months, at a further cost of $14,300. In patients with poor left ventricular function, empiric drug therapy offers 1.5 additional quality-adjusted months over observation, at a cost of $6,900. Electrophysiologic testing provides a further 16.5 months at an additional cost of $16,900. These results hold when the relation between symptoms and arrhythmia is not firmly established. Varying the probabilities of underlying ventricular tachyarrhythmias, bradyarrhythmic conduction defects or noncardiac causes of syncope affects the cost-effectiveness relative to the alternative treatments.