Participant retention is a significant challenge in conducting research in traditionally underserved populations. Many studies experience problems retaining participants, and reasons for loss to follow-up are not routinely reported. Understanding retention is important for understanding bias and potential for generalizability. We examined retention in the context of a longitudinal study of colon cancer screening behaviors and attitudes among adults from medically underserved areas.

This survey study followed participants for 12 months and conducted surveys at baseline, 6-months, and 12-months. Participants were recruited from federally qualified health centers in an urban area. Surveys included self-report questions about demographics, health and healthcare, and attitudes and barriers to colon cancer screening. We examine correlates of retention from baseline to the six-month survey.

Despite several measures to promote retention, the study experienced substantial attrition by six-months. About four in ten baseline participants did not complete the six-month survey. Few participants actively withdrew from the study; most non-completers were unable to be contacted due to wrong or disconnected telephone numbers, change in mailing address, or loss of contact with the “locator” name they provided. Few demographic characteristics were significantly associated with non completion of the six-month survey. Non-completers were significantly (p< .05) more likely to have been unemployed at baseline compared to responders (64.9% of non-responders vs. 39.1% of responders) and to fall in the lowest income category (i.e., <$400/month) compared to responders (60.4% vs. 34.1%). Having a high school education was slightly but not significantly associated with completing the six-month survey. Having health insurance and self-rated health were not associated with responder status.

In this very socioeconomically disadvantaged population, many participants were lost to follow-up despite collecting multiple contact numbers, alternate contacts, and mailing addresses and the use of giftcards as incentives. Collecting the name of a “locator” contact person was helpful in some cases, but in many cases the locator had their information change and may not have been in consistent or reliable contact with the participant. There were few detectable demographic differences between completers and non completers, but non completers were more likely to report unemployment, lack of health insurance, and lower monthly incomes. Surprisingly, a history of homelessness was not associated with completion. However, as most (>70%) respondents had been homeless at some point, it would have been difficult to detect differences. Conducting longitudinal research in underserved populations is challenging, as many researchers have found. Multiple and novel strategies are likely necessary in many cases in order to maximize retention and protect study findings from resultant biases.

Citation Information: Cancer Epidemiol Biomarkers Prev 2011;20(10 Suppl):B17.