Death by p-value: the overreliance on p-values in critical care research

Abstract

The p-value has changed from a versatile tool for scientific reasoning to a strict judge of medical information, with the usual 0.05 cutoff frequently deciding a study’s significance and subsequent clinical use. Through an examination of five critical care interventions that demonstrated meaningful treatment effects yet narrowly missed conventional statistical significance, this paper illustrates how rigid adherence to p-value thresholds may obscure therapeutically beneficial findings. By providing a clear, step-by-step illustration of a basic Bayesian calculation, we demonstrate that clinical importance can remain undetected when relying solely on p-values. These observations challenge current statistical paradigms and advocate for hybrid approaches—including both frequentist and Bayesian methodologies—to provide a more comprehensive understanding of clinical data, ultimately leading to better-informed medical decisions.