Little loss to follow up

Consider whether outcomes were assessed in all (or nearly all) the people being compared.

In randomized comparisons, participants may drop out because the health action was not working, was making them worse or because of side effects (lost to follow-up). If those people are not followed up to the end of a study, and they are left out of the results, the findings of the study may be misleading.

Explanation

Research has found that loss to follow-up (attrition) in comparisons of health actions can lead to overestimates or underestimates of the effects of the health actions being evaluated. People in comparisons who are not followed up to the end of the study may have worse outcomes than those who completed follow-up. The impact of loss to follow-up on the estimates of the effects of health actions is called “attrition bias”. The best way to prevent attrition bias is to make efforts to keep participants in studies, and to search for information on participants who drop out of a study.

Missing information caused by loss to follow-up can be considered in the statistical analysis of the data in a comparison, for example, by making assumptions about the missing data to give an estimate of the effect of a health action. However, there is still a possibility of bias, and even small numbers of participants lost to follow-up can have an impact on the results of comparisons. Most randomized comparisons report the number of participants lost to follow-up, but many do not report analyses that take account of loss to follow-up.

Example

In a randomized comparison of hip protectors for preventing broken hips in elderly people who are more likely to fall, about 20% of participants were lost to follow-up. The authors looked for the information that was missing due to the loss to follow-up by searching the participants’ general practitioners’ records. They found information about broken hips but not for other important outcomes such as quality of life. By looking at the characteristics of study participants at the start of the comparison, researchers could see that more people with poor health and people with a previous broken hip had been lost from the ‘no-hip protector’ group than had been lost from the ‘hip-protector’ group. Due to the missing information in the participants lost to follow-up, particularly about quality of life, the apparent effect of hip protectors on quality of life is far less certain than the effect on broken hips.

Remember: Be careful about relying on the results of comparisons if many people were lost to follow-up, or if there was a big difference between the comparison groups in the number of people lost to follow-up.

Read more about Key Concepts

Reliably measured outcomes

Outcomes counted in the right group

Back to Top