PLOS ONE Medicine&Health

An agent-based model to advance the science of collaborative learning health systems

by Michael Seid, David Bridgeland, Christine L. Schuler, David M. Hartley

Improving the healthcare system is a persistent and pressing challenge. Collaborative Learning Health Systems, or Learning Health Networks (LHNs), are a novel, replicable organizational form in healthcare delivery that show substantial promise for improving health outcomes. To realize that promise requires a scientific understanding that can serve LHNs’ improvement and scaling. We translated social and organizational theories of collaboration to a computational (agent-based) model to develop a computer simulation of an LHN and demonstrate the potential of this new tool for advancing the science of LHNs. Model sensitivity analysis showed a small number of parameters with outsized effect on outcomes. Contour plots of these influential parameters allow exploration of alternative strategies for maximizing model outcomes of interest. A simulated trial of two common health system interventions – pre-visit planning and use of a registry – suggested that the efficacy of these could depend on LHN current state. By translating heuristic theories of LHNs to a specifiable, reproducible, and explicit model, this research advances the scientific study of LHNs using tools available from complex systems science.