Clinical prediction is at the heart of preventing birth trauma and pelvic floor disorders for individual women.

Abstract

The purpose of this article is to understand that the majority of studies investigating the role of risk factors for maternal birth trauma and pelvic floor disorders are designed using causal inferential statistical methods and have not been designed to investigate the more useful goal of clinical prediction. A review of the literature was conducted to describe notable causal and predictive associations between risk factors and maternal birth trauma outcomes. Examples were obtained to illustrate and contrast differences in clinical usefulness between causal and predictive models. Effects of pregnancy and childbirth on the risk of maternal birth trauma outcomes and subsequent pelvic floor disorders are an area of profound investigation. Numerous observational studies provide evidence that pregnancy and childbirth play a causal role in the increasing prevalence of these outcomes, and clinicians must rely on this observational evidence to guide decisions about preventing maternal birth trauma and pelvic floor disorders. However, there are important differences between the design and evaluation of models for a predictive context including: study design goals, inclusion or exclusion of candidate risk factors, model evaluation and the additional need to assess model error. This article contrasts how causal and predictive modeling approaches are different and argues that indiscriminately modeling risk factors for birth trauma and pelvic floor disorder outcomes is costly to women.