Individual- and contextual-level predictors of participant retention in the American Cancer Society Nutrition and Physical Activity Study

Participant retention, which may be influenced by both individual and contextual factors, has been a challenge in interventions aimed to change lifestyle behaviors. Although this is a common problem, predictors remain poorly understood. This study aims to examine specific demographic and contextual variables associated with evaluation retention in an intervention study.

The ACS Nutrition and Physical Activity Study is designed to promote physical activity, healthy diet, and weight management utilizing an individually-tailored counseling intervention conducted through the National Cancer Information Center (NCIC). Participants are randomized to the control or intervention group at enrollment with three follow-ups during the one-year study period. Currently, over 1,600 are randomized with a target of 2,000 by June 2007.

Once due for follow-up evaluation, the NCIC must successfully contact individuals who consent to follow-up interview. We hypothesize that both individual (i.e., age, gender, ethnicity, baseline BMI, and feeling sad or blue) and contextual (i.e., urbanicity, proximity to grocery stores, restaurants, and parks) variables relate to retention in the study evaluation.

In logistic regression, obesity at intake (BMI≥30 kg/m²) was associated with decreased retention in the study at the first follow-up (4 months) in comparison to those with normal BMI (Χ²₂ =6.584, p=0.0371) after controlling for other individual-level variables. Among participants who responded to the first follow-up (N=284), non-Caucasian (N=82) were less likely to remain in the study at the second follow-up (7 months) than Caucasians (Χ²₁ =8.241, p=0.0041), controlling for all individual variables and changes in BMI during the first 4 months. Baseline BMI was not associated with 7-month retention. Subsequent analyses will simultaneously examine contextual and individual characteristics using multilevel logistic regression.

Although all study variables have significant implications for research and policy, examination of contextual variables will be fruitful in achieving optimal levels of participant evaluation retention and obtaining valid best estimates of the effect of behavior-change interventions. Furthermore, contextual variables garnered from sources other than the participant enables researchers to examine probable reasons for follow-up evaluation withdrawal despite loss of individually-reported information. A thorough understanding of these variables may aid researchers in understanding and improving evaluation retention. This is particularly critical in studies of demanding interventions such as those required for effective weight management and lifestyle change.