Results should be described transparently

Just words without certainty

Treatment effects that are only reported descriptively are not adequate. When statistically comparing groups, confidence intervals should be provided to determine the level of uncertainty about a finding.

Words can be easier to understand than numbers however they are subjective and can mean different things to different people. For example, one person’s definition of a “small effect” could be very different from that of another person. Reporting numbers is more objective, allowing you to decide for yourself what they show.

Language can also be used to directly mislead people. For example, someone selling a treatment may describe it as “natural” which is often used to insinuate that it is safer or better for patients. However, this is not always true.

In addition to the calculated statistical outcome and p-value, confidence intervals should also be stated. This is because there is always a level of uncertainty surrounding whether the values measured in our sample population are an accurate representation of the entire population. Confidence intervals, therefore, provide an upper and lower limit around our point estimate (e.g., mean) which will encapsulate the true value for the entire population and provide an estimate of how certain we are of the size of the effect. Most studies will use a confidence interval of either 95% or 99%.

REMEMBER: Don’t be tricked by the words that are chosen to describe treatment effects. Where possible consider if the numbers support the words.