Attrition in an exercise intervention: A comparison of early and later dropouts

OBJECTIVE: To identify reasons for dropout and factors that may predict dro pout from an exercise intervention aimed at improving physical function in frail older persons. DESIGN/SETTING: An 18-month randomized controlled intervention in a communi ty setting. The intervention comprised 2 groups: class-based and self-paced exercise. PARTICIPANTS: 155 community-dwelling older persons, mean age 77.4, with mil dly to moderately compromised mobility. MEASUREMENTS: The primary outcome measure was dropout. Dropouts were groupe d as: D-0, dropout between baseline and 3-month assessment, and D-3, dropou t after 3-month assessment. MEASUREMENTS: Measurements of demographics, health, and physical performanc e included self-rated health, SF-36, disease burden, adverse events, PPT-8, MacArthur battery, 6-minute walk, and gait velocity. RESULTS: There were 56 dropouts (36%), 31 in first 3 months. Compared with retained subjects (R), the D-0 group had greater disease burden (P = .011), worse self-perceived physical health (P = .014), slower usual gait speed ( P = .001), and walked a shorter distance over 6 minutes (P < .001). No diff erences were found between R and D-3 Multinomial logistic regression showed 6-minute walk (P < .001) and usual gait velocity (P < .001) were the stron gest independent predictors of dropout. Controlling for all other variables , adverse events after randomization and 6-minute walk distance were the st rongest independent predictors of dropout, and self-paced exercise assignme nt increased the risk of dropout. CONCLUSIONS: We observed baseline differences between early dropouts and re tained subjects in disease burden, physical function, and endurance, sugges ting that these factors at baseline may predict dropout. Improved understan ding of factors that lead to and predict dropout could allow researchers to identify subjects at risk of dropout before randomization. Assigning targe ted retention techniques in accordance with these factors could result in d ecreased attrition in future studies. Therefore, the results of selective a ttrition of frailer subjects, such as decreased heterogeneity, restricted g eneralizability of study findings, and limited understanding of exercise ef fects in this population, would be avoided.