Noncollapsibility of the odds ratio unraveled

Abstract

Purpose of review The concept of noncollapsibility of the odds ratio has intrigued the epidemiology community for decades. Despite often being overlooked or negatively perceived, which can lead researchers to prefer collapsible measures, this study highlights the beneficial aspects of noncollapsibility of the odds ratio. Recent findings This study links recent diagnostic test research with effect measures. Noncollapsibility is a useful characteristic, and its understanding can have important implications in medical research. Summary We demonstrate that a change in effect could be due to confounding, noncollapsibility, or a combination of both and that noncollapsibility can be a powerful tool for data analysis and interpretation when properly understood. This study aims to help researchers demystify this complex issue and offers updated recommendations.