State of the Art
Leading scientific journals have highlighted a need for improved reproducibility to increase confidence in results and reduce the number of retractions. In a recent survey, 90% of researchers acknowledged that there is a “reproducibility crisis”. It has been proposed that open science where both data and source code are shared could enhance reproducibility. In emergency medicine, very few published studies adhere to these principles.After the data collection stage, research -- whether involving randomized controlled trials, prospective observation studies or retrospective analyses -- can be viewed as a series of computational steps (data cleaning, processing, analysis, visualization) that should be readily reproducible because all of the steps can be scripted into a machine-readable format. Computational reproducibility in this context is the ability to exactly reproduce results given the same data, as opposed to replication, which requires an independent experiment. This article represents an overview of reproducible research methods for emergency medicine. We provide a framework and example of computational reproducibility and highlight specific features including open code, open data, executable documents, and scientific containers. Specifically, we will discuss open code, open data, executable documents, and scientific containers. We will highlight simple steps that researchers can take including best coding practices, versioning, and public repositories for code storage. We will discuss how to de-identify your data to make it suitable for public posting, IRB issues around the posting of data, and data storage in public repositories.We will provide live demonstrations in R studio. We will discuss the basics behind the scientific container concept and point users to available software and areas for further learning.