Great advances in medicine were achieved following application of statistical analysis to research findings in the early 20th century. Armed with a formal mechanism to analyze data, investigators could reject empirical claims that treatments such as blistering and bloodletting were beneficial. Structured clinical trials with results analyzed in objective ways paved the way for improved health care by proving that vaccinations and surgical antisepsis were effective.
Statistical analysis of data has proven highly successful for ensuring that medical research findings are objectively analyzed, minimizing the risk of promoting practices solely on empirical evidence. Courses in statistics are a required part of every medical school curriculum. Unfortunately, most physicians only acquire a superficial understanding of statistics. A limited understanding of these analytic techniques can be harmful. Most clinicians equate a P value of less than .05 with the significance of a finding, assuming that any finding with this level of a P value (or less) means that the finding is important and should be accepted. Blindly accepting this logic has caused a great deal of harm.1 Adverse risks of medications have been overstated and some therapies overvalued because of inappropriate acceptance of statistical analyses. Missing in the interpretation of most studies is an easily understood measure of the overall clinical significance of an observation to complement reported statistical significance measurements.
Livingston EH, Elliot A, Hynan L, Cao J. Effect Size EstimationA Necessary Component of Statistical Analysis. Arch Surg. 2009;144(8):706-712. doi:10.1001/archsurg.2009.150