Read Missing Outcome Data Before Trusting a Study Result
Missing outcome data can affect a study's conclusion when the missingness is related to the unobserved outcome. The percentage missing is only part of the question; reasons and group differences also matter.
The worked examples are teaching aids. Check them against your own source material and the requirements of your specific task.
Start with the task
Cochrane's trial guidance explains why missing results need a contextual assessment.
Follow the participants
For the outcome you intend to cite, record how many participants were assigned or enrolled, how many supplied that outcome, and why others did not. Use the relevant time point rather than a single overall completion count.
In a fictional teaching study, participants who found a programme difficult might be less likely to complete the follow-up questionnaire. Analysing only completed questionnaires could tell a different story from the full assigned group.
This possibility is a question to investigate, not proof that the study is biased.
Compare the groups and assumptions
Were losses similar in number and reason? What method was used to handle the missing data? Did sensitivity analyses examine plausible alternatives?
Do not assume that replacing missing values or using a sophisticated model resolves the problem. Those methods depend on assumptions that need to fit the situation.
Avoid a universal rule that any particular missing percentage is always acceptable. A small amount can matter for a rare outcome, while a larger amount may have different implications under other conditions.
Preserve the uncertainty in the rationale
A useful note names the missing outcome, the reported explanation, and what it limits. “Follow-up was incomplete” is less informative than identifying which result is affected.
Use the evidence-source guide and keep the qualification attached to the finding in Nursevi. Do not translate a result from respondents into a guaranteed effect for everyone in the assignment's target group.