In epidemiologic studies, it is important to avoid systemic errors that would make it difficult or wrong to generalize findings to larger populations. One error, selection bias, is generally thought of as a bias introduced when selecting individuals or groups for data collection or analysis, as the name implies, in a way that results in inherent differences relating to the outcome of the study.
However, when establishing a prospective cohort to begin a long-term study, enrollment procedures generally do not result in selection bias because the outcomes of interest have not yet occurred, and researchers can create enrollment criteria that will allow for a representative population.
Selection bias in prospective cohort studies more commonly occurs during the follow-up period with differences in retention over time among the groups of interest. Research participants leave the cohort for a number of reasons, including:
- Moving without providing updated contact information;
- Losing interest in participating;
- Becoming sick or unable to participate; or
Participant Retention is Key for Accuracy of Long-Term Studies
If these events were to occur randomly among the study population groups so that events happened equally with respect to the outcomes of interest, this would be non-differential and may not bias the results. But if the events tend to happen more often in the ‘exposed’ or ‘unexposed’ groups, this would result in differential loss to follow-up and would cause a selection bias.
To prevent this selection bias, researchers need to maintain a cohort of at least 80-90 percent of those who originally enrolled. But this takes time, effort, training, and regular communication with research participants as well as staff over the course of the entire study, which can last 10 to 20 years or more.
The Lip Balm Effect: Small Gestures Can Encourage Research Subjects to Stay Engaged
Common steps that are vital to study retention include tracking and tracing efforts, such as obtaining data from cancer registries or the National Death Index, as well as internet tracking resources for updated contact information. While these tools can be effective, especially to ascertain valuable end-points, such as cancer diagnoses and vital status, I would argue that the best efforts can be made proactively through regular contact with research participants.
A first step is to ensure that staff who have direct contact with participants are well trained and feel comfortable answering questions about the study, or know who to contact if the participant requests additional resources. Research staff should collect multiple types of contact information at the participant’s initial visit and update this information annually, including information on two or three close family members or friends who staff would have permission to contact if the participant becomes lost to follow-up. Regular communication through mail, email, and newsletters allows staff to quickly know if someone’s contact information has changed. For instance, the post office only provides a forwarding address for one year.
Additionally, sending small tokens such as pens, magnets or lip balm provides an opportunity to show appreciation as well as provide study contact information. Research findings and study updates should be given to the participants throughout the course of the project. This helps keep participants engaged over time by allowing them to see how they are part of the bigger project and to know that their time is valued. In this technical age, a well-maintained and regularly updated website should serve as a hub for these efforts as well as for data collection.
A large, prospective cohort study has the opportunity to elucidate research answers that other study types, such as retrospective or case-control studies, cannot answer, but only if the proper methods and procedures – including some innovative ones – are used to maintain the validity of the study.