Quantifying the Social Influence on Information Technology Adoption in Healthcare: A Hierarchical Bayesian Learning Model Analysis

Information technology adoption is a critical continuing challenge in healthcare. This paper presents a Bayesian learning model to quantify social learning effects on information technology adoption by physicians from a community health system. The model estimates confirm that both self-learning and social learning effects reduce the uncertainty around the perceived value of a new information technology solution, with subsequent positive impact on users' adoption of the technology. However, self-learning is not easy to influence, so this study is specifically focused on quantifying the impact of social learning effects, opinion leader effects in this community health system context, on technology adoption behavior by physicians in medical practice groups. Simulations of potential policy interventions indicate that if opinion leaders doubled their monthly technology use, their peers' technology adoption probability will increase by 4.4%, on average; however, if opinion leader effects variability decreases by 50%, i.e. opinion leaders send more consistent signals or precise information to their peers, their peers' technology adoption probability will increase by 24%, on average. More importantly, the simulated addition of opinion leaders to medical practice groups which had no opinion leaders, using the new technology 5 instances each month, shows that the adoption probability of their peer physician users will increase by a significant 44%, on average. These simulation results suggest that the more the opinion leaders use the new technology or the lower the variability of opinion leaders' signals, the higher their peers' adoption probability, and opinion leaders' consistent signals more than use frequency have stronger influence on peers' adoption behavior. Also, the addition of an opinion leader to groups that did not have one significantly impacts users' adoption probability. A Markov Chain Monte Carlo method is used for model estimation and the hierarchical Bayesian 3 structure of the model also shows that substantial heterogeneity exists across physicians even after controlling for demographic variables.