With escalating healthcare costs and increasing concerns about optimizing use of medicine, there is an unresolved debate over years around the potential impact of pharmaceutical promotion on physicians' prescribing behaviors. What should be the appropriate balance of promotion dollars to physicians? We use three major brands in the US antibiotic universe to explore this issue, presenting a theoretical framework for better understanding the cause-and-effect relationship between common promotional spending and prescription responsiveness. Using simulations we demonstrate that neural networks guided by genetic algorithm-partial least squares is able to provide managers with better understanding of physicians' prescribing activities without an appreciably lower predictive accuracy when compared to that obtained by a standalone neural network modeling.

He has previously worked with the development and application of computational methods to problems in drug delivery research. Much of his work was centred on pharmacokinetic analysis of drug disposition in skin and prediction of skin permeability. Currently, he is working at the International Marketing Department in Daiichi Pharmaceutical Co., Ltd. His role combines market research with special interest in promotional effectiveness.

Purpose -The aim is to explore the potential of a hybrid genetic algorithm-partial least squares (GA-PLS) modeling approach to understand the important promotional spending variables that influence physicians' prescribing habits and to help leverage managers' insight to plan better spend on their promotional activities. Design/methodology/approach -A GA was used as a variable-selecting tool, and PLS analysis was employed for correlating these variables with the observed variation in the volume of prescriptions. This approach is illustrated using database from a marketing consultant on four major brands in the US antibiotic universe. Findings -Good statistical models were derived that permit simpler and faster computational prediction of the effects of physician-directed promotion. All final models established had r 2 values ranging from 0.835 to 0.922 and cross-validated r 2 (q 2 ) values ranging from 0.791 to 0.911 whereas the mean absolute percentage error (MAPE) values were confined within 5 percent range on averaging all brand models. Further, thorough statistical analyses revealed the usefulness of promotional spending as a variable and the robustness of GA-PLS as a correlation tool. Research limitations/implications -Modeling frame was comprised of only antibiotic category, which may limit its general utility. Practical implications -Managers can become more adept at interpreting the effects of promotion on prescribing behaviors of physicians and are able to build predictive models that would help identify where and how their curious blend of promotional cocktail would yield the highest future returns. Moreover, if the impact of individual promotional spending element can be measured, then this is perhaps a testament to the way the efficacy of interventions to reduce the harmful consequences of pharmaceutical marketing could be validated given a growing number of public beliefs that physician-directed promotion has grown too heavy-handed and is undermining medical professionalism. Originality/value -This area of research has not received much attention in the pharmaceutical marketing literature until recent years, and hopefully this study will stimulate some interest.

No abstract