ABSTRACT
The basic story of the paper is a simple and pretty realistic medical problem, because there is no perfect test available. A physician and/or his patient1 must make the (irreversible) decision to treat or not to treat while there is only suspicion that a single (well-defined) illness might be present. This risk is characterized by the a priori probability of disease (p with 0 p 1). Given the medical benefits and costs of the potential treatment for the patient (benefits and costs are assumed to be known with certainty), P.K. describe the critical value of p (i.e, the threshold) above which treatment becomes the best decision. Because the costs and benefits of the treatment are known with certainty while the risk relates to the true health state of the patient, we consider here – as P.K. did – a situation of ‘diagnostic risk’.2 Of course other assumptions corresponding to other medical situations would be possible (e.g. random costs and benefits of the treatment) but they will not be considered here for the sake of brevity.
