Deploying Healthcare Simulation Models Using Containerization and Continuous Integration

Abstract

Methods or approaches from disciplines outside of OR Modeling and Simulation (M&S) can potentially increase the functionality of simulation models. In healthcare research, where simulation models are commonly used, we see few applications of models that can easily be deployed by other researchers or by healthcare stakeholders. Models are treated as disposable artifacts, developed to deliver a set of results for stakeholders or for publication. By utilising approaches from software engineering, M&S researchers can develop models that are intended to be deployed for re-use. We propose one potential solution to deploying free and open source simulations using containerisation with continuous integration. A container provides a self-contained environment that encapsulates the model and all its required dependencies including the operating system, software, and packages. This overcomes a significant barrier to sharing models developed in open source software, which is dependency management. Isolating the environment in a container ensures that the simulation model behaves the same way across different computing environments. It also means that other users can interact with the model without installing software and packages, supporting both use and re-use, and reproducibility of results. We illustrate the approach using a model developed for orthopaedic elective recovery planning, developed with a user-friendly interface in Python, including a clear set of steps to support M&S researchers to deploy their own models using our hybrid framework.

Towards Sharing Tools and Artefacts for Reusable Simulations in Healthcare

“…Although not essential to the practice of sharing models, very few authors paid attention to the accessibility of their computer models. Accessibility is an important facet of open science, and could be enhanced via contemporary web technology to enable healthcare professionals or other researchers to use models without the requirement to install software (Dagkakis & Heavey, 2016;Harper, Monks, & Manzi, 2023;R. Smith & Schneider, 2020;Wojciechowski, Hopkins, & Upton, 2015).…”

“…Documentation can be enhanced with notebooks that combine code and explanation (Ayllón et al, 2021;The Turing Way Community, 2022), and reporting guidelines to improve explanation of models and data (Monks et al, 2019). Finally, to enhance the accessibility and maintainability of models, contemporary web technology can enable online execution, interaction, and experimentation (Dagkakis & Heavey, 2016;Harper, Monks, & Manzi, 2023;.…”

“…What is clear is that more complex open science methods require more effort than simpler efforts Harper, Monks, & Manzi, 2023;. For example, in a small trial, the journal Nature Biotechnology reported that it took a median of nine days for authors to set up an online executable version of their computational artefacts using the containerisation and compute services provided by Code Ocean (Editorial (2019), Nature Biotechnology).…”

“…Although we provide no formal guidance, a variation of STARS could also be considered where modellers deploy their simulations via containerisation and continuous integration technology (Harper, Monks, & Manzi, 2023;Moreau et al, 2023). For example, when a model is updated a Docker image is automatically built and deployed to DockerHub (a container registry).…”