Diffusion and social networks: revisiting medical innovation with agents

Pascal Perez, Nazmun Ratna, Anne Dray, Quentin Grafton, David Newth, Thomas Kompas

    Research output: Contribution to conferencePaper

    Abstract

    In this paper, we reanalyze Medical Innovation, the classic study on diffusion of Tetracycline by Coleman, Katz and Menzel (1966). Medical Innovation articulates how different patterns of interpersonal communications can influence the diffusion process at different stages of adoption. In their pioneering study, individual network (discussion, friendship or advice) was perceived as a set of disjointed pairs, and the extent of influences were therefore, evaluated for pairs of individuals. Given the existence of overlapping networks and consequent influences on doctors' adoption decisions, the complexity of actual events was not captured by pair analysis. Subsequent reanalyses (Burt 1987, Strang and Tuma 1993, Valente 1995, Van den Bulte and Lilien 2001) failed to capture the complexity involved in the diffusion process and had a static exposure of the network structure. In this paper, for the first time, we address these limitations by combining Agent-Based Modeling (ABM) and network analysis. Based on the findings of Coleman et. al. (1966) study, we develop a diffusion model, Gammanym. Using SMALLTALK programming language, Gammanym is developed with CORMAS platform under Visual Works environment. The medical community is portrayed in an 8 X 8 spatial grid. The unit cell captures three different locations for professional interactions: practices, hospitals, and conference centers, randomly located over the spatial grid. Two social agents-Doctor and Laboratory are depicted in the model. Doctors are the principal agents in the diffusion process and are initially located at their respective practices. A doctor's adoption decision is influenced by a random friendship network, and a professional network created through discussions with office colleagues, or hospital visits or conference attendance. A communicating agent, Laboratory, on the other hand, influences doctors' adoption decisions by sending information through multiple channels: medical representatives or detailman visiting practices, journals sent to doctors' practices and commercial flyers available during conferences. Doctors' decisions to adopt a new drug involve interdependent local interactions among different entities in Gammanym. The cumulative adoption curves (Figure 1) are derived for three sets of initial conditions, based on which network topology and evolution of uptake are analyzed. The three scenarios are specified to evaluate the degree of influences by different factors in the diffusion process: baseline scenario with one seed (initial adopter), one detailman and one journal; heavy media scenario with one seed but increasing degrees of external influence, with five detailman and four journals; and integration scenario with one seed, without any external influence from the laboratory. (Graph Presented) Averaged over an ensemble of 100 runs, clustering coefficient and average shortest path length indicate that social networks depicted in Gammanym are random graphs. Evolution of uptake suggests that although the degree of external influence in terms of marketing strategies adopted by the pharmaceutical company does not have impact on the network structure, the speed of diffusion is largely determined by it.
    Original languageEnglish
    Pages1639-45
    Publication statusPublished - 2005
    Event2005 International Congress on Modelling and Simulation: Advances and Applications for Management and Decision Making, MODSIM 2005 - Melbourne, Australia
    Duration: 12 Dec 200515 Dec 2005

    Conference

    Conference2005 International Congress on Modelling and Simulation: Advances and Applications for Management and Decision Making, MODSIM 2005
    CountryAustralia
    CityMelbourne
    Period12/12/0515/12/05

    Fingerprint Dive into the research topics of 'Diffusion and social networks: revisiting medical innovation with agents'. Together they form a unique fingerprint.

    Cite this