Primer for Clinical Researchers on Ordinal Outcomes for Emergency Medicine

Ordinal outcomes are becoming increasingly common in clinical research because they can incorporate multiple clinical states into a single outcome, offer increased statistical power compared to binary outcomes, and can be applied across a range of illness severities. However, ordinal outcomes may be unfamiliar to many clinicians and researchers. In this paper, we aim to provide a practical conceptual overview of ordinal outcomes in the context of emergency medicine, discuss their pros and cons, and describe the associated effect measures and statistical methods that can be used to estimate them. We describe the different types of ordinal outcomes, including hierarchical composite endpoints (e.g., Desirability of Outcome Ranking or DOOR); outline the target parameters that may be of interest, such as the common odds ratio and win ratio; and describe the statistical methods for estimating these parameters, including their assumptions and limitations. Ordinal outcomes offer a flexible, efficient, and nuanced way to measure treatment effects but require careful planning for how treatment effects will be estimated and communicated. We hope this review will make these outcomes more accessible for clinical researchers.