A new statistical trend in clinical research – Bayesian statistics

Abstract

Background: The emphasis on evidence-based practice in physical therapy has increased the number of clinicians who perform and interpret clinical research. Unfortunately, the traditional statistical analysis (frequentist approach) used most often in clinical research (except meta-analysis) has been criticized by biostatisticians for potential bias and misleading results if used with data from single studies. Alternatively, Bayesian inference can be used instead of the traditional frequentist approach although this trend has yet to be seen in rehabilitation research. Used for at least three decades, the Bayesian approach provides a formal framework for researchers to incorporate prior knowledge and current evidence to derive new probabilities for various hypotheses. Since the results are presented in terms of probability, clinicians can interpret and apply research findings to clinical practice directly.Objectives: The objectives of this review are to discuss the common misconceptions among users of the frequentist approach, the inherent limitations of the frequentist approach, as well as to introduce the characteristics and limitations of the Bayesian approach using illustrated examples.Conclusions: The Bayesian approach can be used as an alternative or adjunct to the frequentist method in future studies. This approach is also robust in situations that are unfavourable to traditional statistics such as sequential clinical trials. However, biostatisticians may have to be consulted for some sophisticated Bayesian analysis. As the Bayesian approach may gain popularity, a good understanding of this method will benefit clinicians in interpreting research papers and planning their future clinical studies.