P-Values, Effect Sizes and Whether Results Apply to Your Patients

A p-value below 0.05 only suggests a result may be distinguishable from chance, and its true false positive rate is likely higher than 5%; Neil recommends David Colquhoun's papers on the topic. What matters is the scale used, the size of the difference between groups and its precision, shown by 95% confidence intervals. A 0.3-point drop on a 10-point pain scale is unlikely to change anyone's life. Whether an effect is clinically important should be decided with patients, weighing baseline pain, cost, effort and risks. Applying results locally draws on clinical expertise: who was recruited or excluded, the care setting, whether the intervention is feasible and acceptable, and whether outcomes are patient-centered. Avoid hunting for flaws just to dismiss a paper you dislike. Neil's takeaways: be prepared to hear any answer, invest in critical appraisal as much as new techniques, choose authorities such as Cochrane reviews, NICE or bloggers critically, and stop arguing from papers you have not read. Systematic reviews and good clinical guidelines are efficient starting points, and the best way to begin is simply to start.