Predictors of Loss to Self-report Follow-up Data Collection in a Cancer Screening Intervention

Abstract

Objective: Losing participants’ self-report data affects process and outcome analyses, and ultimately, conclusions about results. In this paper, we examine predictors of loss to self-report data collection and attempt to identify potential predictors that can be addressed prospectively. Methods: Data were from a study to increase colorectal and breast cancer screening in women 50-75 years of age (N = 1196). We collected self-report data at baseline (T1), 4 weeks (T2), and 6 months (T3) after consent. Analyses identified predictors of earliest loss (T1 vs T1,T2,T3), intermediate loss (T1 vs T1,T2), and later loss (T1,T2 vs T1,T2,T3). Results: Cancer knowledge and self-reported screening barriers were associated with loss to follow-up. More health problems were associated with less intermediate loss to follow-up, but higher loss to later follow-up. Two intervention groups (Web Only, and Web + Phone) showed greater loss compared to Usual Care. Being overdue for both screenings predicted early loss. Completing the T1 survey by phone was associated with greater loss to follow-up. Conclusions: Knowledge and barriers may have had an early effect, whereas health problems might have had a delayed impact. Intervention characteristics also need to be considered as a source of task demands on participants.