Rush regression workbench: An integrated open-source application for regression modeling and analysis in healthcare analytics

Abstract

Regression is widely used in healthcare analytics, whether for examining hospital quality and safety, characterizing patterns of patient volume and healthcare costs, or predicting patient outcomes. Simple linear regression and other basic forms can be conducted with spreadsheet programs and are useful for examining simple linear relationships. However, expert statistical knowledge, computational skills, and specialized tools may be needed to characterize nonlinear relationships and complex interactions, to examine data that fail the assumptions of linear regression, to identify confounding variables and lessen the influence of outliers, and to build and evaluate predictive models. We constructed the Rush Regression Workbench to accomplish these tasks and to automate cautious and sophisticated analyses, provide interpretive outputs, enable reproducible results, and to provide the community with an evolving open-source good containing a diverse set of analyses and a growing library of over 170 preprocessed public healthcare datasets. The Rush Regression Workbench can be accessed via the web or downloaded and used locally.