Agent-based computational models to explore diffusion of medical innovations among cardiologists

Highlights

• Diffusion of innovations is often driven by social contagion or imitation.

• Agent-based models (ABM) are useful to explore diffusion of medical innovations.

• ABM demonstrates the interaction of variables associated to innovation spreading.

• ABM provides a semiquantitative insight on doctors’ emergent collective behavior.

Abstract

Background

Diffusion of medical innovations among physicians rests on a set of theoretical assumptions, including learning and decision-making under uncertainty, social-normative pressures, medical expert knowledge, competitive concerns, network performance effects, professional autonomy or individualism and scientific evidence.

Objectives

The aim of this study was to develop and test four real data-based, agent-based computational models (ABM) to qualitatively and quantitatively explore the factors associated with diffusion and application of innovations among cardiologists.

Methods

Four ABM were developed to study diffusion and application of medical innovations among cardiologists, considering physicians’ network connections, leaders’ opinions, “adopters’ categories”, physicians’ autonomy, scientific evidence, patients’ pressure, affordability for the end-user population, and promotion from companies.

Results

Simulations demonstrated that social imitation among local cardiologists was sufficient for innovation diffusion, as long as opinion leaders did not act as detractors of the innovation. Even in the absence of full scientific evidence to support innovation, up to one-fifth of cardiologists could accept it when local leaders acted as promoters. Patients’ pressure showed a large effect size (Cohen's d > 1.2) on the proportion of cardiologists applying an innovation. Two qualitative patterns (speckled and granular) appeared associated to traditional Gompertz and sigmoid cumulative distributions.

Conclusions

These computational models provided a semiquantitative insight on the emergent collective behavior of a physician population facing the acceptance or refusal of medical innovations. Inclusion in the models of factors related to patients’ pressure and accesibility to medical coverage revealed the contrast between accepting and effectively adopting a new product or technology for population health care.

Introduction

It is commonly accepted that diffusion of innovations is often driven by social contagion or imitation [1]. Specifically, diffusion of medical innovations among physicians rests on a set of theoretical assumptions, including learning and decision-making under uncertainty, social-normative pressures, medical expert knowledge, competitive concerns, network performance effects, professional autonomy or individualism, and scientific evidence [2]. The process of propagation and adoption of medical innovations is not only relevant for marketing research, but also because differences in the performance of medical care may be due to variation in the introduction, diffusion and acceptance of new practices [[3], [4], [5], [6]].

Mathematical representation on the basis of Bass [[4], [5], [6], [7]] or Gamma-Shifted Gompertz [1] models and local interaction models for social networks [8] were traditionally used to explore diffusion of innovations. Low-dimensional differential equations, aggregate regression analysis, and game theory methods are appropriate for some explanatory purposes; however, agent-based computational modeling (ABM) is the main tool in the analysis of spatially distributed systems with heterogeneous autonomous actors connected in social networks, particularly, when decision-making relies on restrained information and limited individual rationality or computing capacity [9]. In ABM, a system is designed as a collection of independent decision-making entities called agents, where each agent individually assesses its situation and makes decisions on the basis of a set of rules. An advantage of ABM is that it computes repetitive competitive interactions between agents to explore population dynamics out of reach of traditional differential equation modeling [10]. As a tool for real-world empirical research, ABM can generate stable macroscopic patterns arising from local interaction of agents, known as emergent phenomena [9]. Agent-based computational modeling applies to cases where people are influenced by their social context, i.e. what others around them do; therefore, it can be used to model diffusion of ideas, products or innovations on social networks. Previously, only a few studies used the ABM approach to model diffusion of medical innovations. These studies provided support to the importance of social networks and external influences in the diffusion process [11], the effect of the interaction between patients and physicians to adopt a new technology [12] and the value of opinion leadership on adopting new evidence [13].

The aim of this study was to develop and test four real data-based ABM to qualitatively and quantitatively explore the factors associated with diffusion and application of innovations among cardiologists.