A prediction interval is different from uncertainty around the average

Find what quantity the interval surrounds

Cochrane distinguishes the mean-effect confidence interval from a prediction interval representing the spread of underlying study effects. Prediction intervals rely on assumptions and can be problematic when the study set is small. These statements concern methods; they do not establish an actual vaping effect or guarantee a result in a new setting.

Read the figure label and the corresponding methods. A review may show intervals for individual studies, a pooled mean, and a prediction range. Do not borrow a line from one layer to explain another. Write the exact interval type beside the result before interpreting whether the graphic shows precision, variation, or both.

Cochrane: Cochrane Handbook, Chapter 10: Analysing data and undertaking meta-analyses

Use an original average-and-spread picture

Imagine an invented collection of support-study estimates that vary across settings. A narrow interval around their average could coexist with substantial spread among underlying effects. Draw a small bracket labeled uncertainty around average and a different bracket labeled modeled spread. Give the brackets no clinical numbers or claimed quitting probabilities.

The picture illustrates two questions rather than a contradiction. Knowing an average more precisely does not force every setting to have that average. Equally, spread between settings does not prove that the average cannot be informative. The task is to see which question each interval addresses before deciding what a headline omitted.

Retain the model and comparable-setting condition

A prediction interval does not remove limitations in study design, missing information, or how well the included studies fit the reader’s question. Read the assumptions behind its calculation and the review’s discussion of the study count. Avoid constructing your own prediction range from the widest visible study line.

If the range crosses the no-difference value on the relevant scale, keep that feature alongside the mean rather than choosing whichever graphic looks more favorable. Its interpretation still needs the exact measure, uncertainty, and evidence appraisal. A range over study effects is not a range dividing individual participants into future successes and failures.

Write two scoped sentences instead of one promise

Your note can describe the reported mean estimate and its uncertainty, then separately describe the reported prediction interval and its stated limitations. If the review provides only the first, do not invent the second. If the interval type is unclear, mark that ambiguity and consult the full explanation before sharing a forecast.

For individualized care, seek qualified interpretation of the actual review. An introductory illustration cannot determine which intervention is suitable for you. Your own progress chart also cannot supply the missing evidence about between-study variation. The useful output is recognizing the difference between precision of an average, variation across settings, and an individual outcome that neither interval establishes.