Loss to follow-up in cohort studies
Loss to follow-up is when participants leave a cohort study (one that follows groups of people over time to see who develops the disease) before their outcome is known.
Remember
Loss to follow-up threatens a cohort study when the participants who drop out differ from those who remain: the people still observed are no longer a fair picture of the group that started, and the comparison between exposed and unexposed participants can be distorted.
Memory aid
a cohort is a long journey, and the people who get off the bus early are rarely a random sample of the passengers
Picture the cohort as passengers on a long bus journey, and ask whether those who get off early resemble those who stay on.
In a cohort study, groups of people are classified by exposure (for example, air pollution) and then followed to see who develops the outcome (for example, chronic cough). Because follow-up runs for years, some participants die, refuse to continue or cannot be contacted. This is loss to follow-up: they leave before their outcome is known.
Selective loss happens when the reasons for leaving are linked to characteristics that also matter for the outcome. The people who remain then differ from those who left, so the cohort no longer gives a fair picture of the group that was recruited. This is attrition bias: a systematic difference between the people still under observation and the people the study set out to describe. Loss that is unrelated to exposure and outcome does not make the remaining sample systematically different, but it still reduces the information available, because fewer outcomes are observed.
Dropout can also distort the comparison between exposed and unexposed participants, when the reasons for leaving are related to both the exposure and the outcome. A related idea is survivor bias: restricting the analysis to participants who remain available can select a healthier group. In a cohort followed for night-shift work and hypertension, if those who become unwell leave, and healthier ones stay, the harm of night-shift work may look weaker than it really is.
The timing separates loss to follow-up from problems that arise at recruitment. The problem appears during follow-up, after recruitment, and the longer the follow-up, the more people are lost. Researchers should record why people were lost, then check how the loss relates to exposure and outcome.
Key facts
- Loss to follow-up
- participants leave a cohort before their outcome is recorded.
- Selective loss
- if those who leave differ from those who remain, the sample becomes unrepresentative.
- Attrition bias
- loss linked to both the exposure and the risk of the outcome can distort the estimated association.
- Any loss
- even loss that does not introduce bias reduces the information available, because fewer outcomes are observed.
- External validity
- the extent to which findings can be applied beyond the people actually studied, which selective loss limits.
- Long follow-up
- the longer the study, the more people are lost, so the bigger the threat.
- Response
- record the reasons for loss and assess how they relate to both exposure and outcome.
Exam traps
- A healthy volunteer effect (recruiting participants who are healthier than the target population) can make a cohort unrepresentative, but it arises at recruitment, before follow-up begins, so it does not explain people dropping out over time.
- Confounding, a distortion of the exposure-outcome association by a third variable linked to both, can occur even if nobody drops out, so it is a different problem from losing participants.
- Recall bias arises when people remember and report past exposures inaccurately. It affects how accurately exposure is measured, not who is left in the study, so it does not explain the pattern of dropout.
Read the full notes
Tested in 7 SimpleMed+ questions
Practice it in Population and Social Science, with the answer and a full explanation after every question.
Practice these questions