Medical procedure choice and social interactions- the role of patient heterogeneity

To identify possible explanations for high regional varia-tion in medical procedure choices, we consider a dynamic model where physicians base treatment choices on two factors: the characteristic of the patient, drawn according to a random variable describing the pa-tient population and on a variable that depends on the past choices of the physicians' network neighbors, the physician's own past choices or a combination of these. We characterize the long run treatment fre-quencies in some particular cases. This characterization shows that two behavioral assumptions, social interactions or personal reinforcement or a combination can reinforce di erences between regions in a similar way. This highlights a possible explanation for the observed variations other than that of social interactions which has previously been suggested in the litterature. In the model, we then compare steady state frequen-cies of treatments to those that would prevail without social or personal in uence. We relate the direction and magnitude of the reinforcement e ects to the distribution of patient characteristics.