A model creates the framework for a cost-effectiveness analysis, allowing d
ecision makers to explore the implications of using an intervention in diff
erent ways and under different conditions. To serve its purpose a model mus
t produce accurate predictions and allow for substantial variation in the f
actors that influence costs and effects. This paper considers three aspects
of modelling: validating effectiveness estimates; modelling costs; and the
implications of common statistical forms. Validation procedures similar to
those for effectiveness estimates are proposed for costs. Modellers need t
o pay more attention to ensuring that the pathway of events described by a
model represents costs as well as it does effects. Modellers can also help
improve the epidemiological and clinical research on which cost-effectivene
ss analyses depend by showing the implications for resource allocation of t
he statistical forms conventionally used in these fields. Copyright (C) 199
9 John Wiley & Sons, Ltd.