standardised mean difference

Trials of the same treatment often measure the same outcome on different instruments. Depression research uses several rating scales, insomnia research several more, and their raw scores cannot simply be averaged together. The standardised mean difference solves that by dividing the difference between the group means by the standard deviation of the measurements within the studies, which is the spread of individual scores around their average. The result is a number expressed in units of that spread rather than in points on any one scale.

The Cochrane Handbook describes it as the size of the intervention effect in each study relative to the between participant variability in outcome measurements observed in that study. Values near 0.2 are conventionally called small, 0.5 moderate and 0.8 large, but those labels are conventions rather than facts, and none of them says whether a change is noticeable to the person experiencing it.

The measure carries an assumption worth knowing. It treats a difference in standard deviation between trials as a difference in the measuring scale rather than as a real difference in how varied the participants were. Where trials recruited quite different populations, that assumption weakens, and the pooled figure becomes harder to read.