Patterns of pharmaceutical prescribing

Stern, Philip (1994) Patterns of pharmaceutical prescribing. Doctoral thesis, University of London: London Business School.

Abstract

This is an empirical study of how general practitioners prescribe ethical pharmaceuticals in the UK. The focus is describing and modelling rather than explaining the prescribing patterns. The study draws upon two stochastic models, the Negative Binomial Distribution and the Dirichlet. These were originally developed in response to empirical regularities observed in grocery markets and have evolved into a substantive theory of buyer behaviour. Panel data for two pharmaceutical product fields and eight different medical diagnoses are analysed. In order to describe the patterns, various measures of behaviour are calculated: the number of doctors prescribing each drug and the rate at which they prescribe; the distribution of light and heavy prescribers; the incidence of sole prescribing; and the way in which doctors spread their prescription needs across the different drugs available. The analysis consists of five elements. Firstly the empirical regularities of prescribing behaviour are explored and then these are modelled with the NBD-Dirichlet at the product field level and then at the diagnosis level. Fourthly, prescribing behaviour of branded pharmaceuticals is compared with that of their generic equivalents. Finally, a single brand which exhibits consistent deviations from the model is studied in some detail to explain why it is different. There are many similarities between pharmaceutical prescribing and buying groceries despite the extensive differences in the structure of the markets. For example, doctors neither pay for nor consume what they choose to prescribe, and they are trained to make particular choices. The Dirichlet model predicts the components of a drug's prescription share very well. It also predicts product field prescribing behaviour by aggregating data at the diagnosis level. The analysis reveals differentiation with respect to diagnosis which varies between the two product classes. There is also a tendency for doctors who prescribe a specific drug for one diagnosis to be slightly more inclined to prescribe it in another diagnostic situation. In general there are no differences in the way that branded and generic drugs are prescribed.