Learning from evaluation of clinical outcomes is greatly facilitated by three things that still are often not done: 1) emphasizing clinical significance, not statistical significance. Even with well-done statistics, which usually should not be based on an assumption of random distribution, it really doesn't matter if a finding is statistically significant if its magnitude is so small that changing clinical procedures to obtain this effect would make no practical difference in treatment effectiveness or time, 2) grouping data by problem type, by treatment procedure, and by magnitude of treatment response to show the percentage of patients with minimal, moderate or extreme responses. This is a key to understanding whether findings really are clinically significant and what patients should be told in obtaining informed consent, and 3) reporting data from consecutively treated patients-all of them, no exceptions. Ideally, such data come from a randomized clinical trial, but well-categorized retrospective data are more likely to be available and can be perfectly satisfactory in determining clinical significance. These points will be illustrated with data from selected studies.
Identify clinical significance upon evaluation of outcomes.
Group data to understand findings.
Analyze data to determine its clinical significance.