Predictors of Study Attrition in a Randomized Controlled Trial Evaluating a Perinatal Home-Visiting Program with Mothers with Psychosocial Vulnerabilities

OBJECTIVE Randomised controlled trials evaluating perinatal home-visiting programs are frequently confronted with the problem of high attrition rates. The aim of the present study is to identify predictors of study attrition in a trial evaluating a perinatal home-visiting program in France. MATERIALS AND METHODS CAPEDP is a French randomized trial comparing a perinatal home-visiting program using psychologists versus usual care (N = 440). The first assessment was at inclusion into the trial at the 27th week of pregnancy and the final assessment when the child reached the age of two. Attrition rates were calculated at 3 and 24 months postpartum. Stepwise logistic regression was used to identify predictors of early (between inclusion and 3 months postpartum) and later (between 3 and 24 months postpartum) attrition among social, psychological and parenting factors. RESULTS Attrition rates were 17% and 63% at 3 and 24 months respectively. At 24 months, there was significantly more attrition in the control arm (70.6%) compared to the intervention arm (55.2%). Five independent predictors of early attrition were identified: having already had an abortion; having greater attachment insecurity as measured with the Vulnerable Attachment Style Questionnaire (VASQ); having lower global severity of psychiatric symptoms as assessed with the Symptom Check-List (SCL-90) at inclusion, being neither currently employed nor studying; and declaring no tobacco consumption during pregnancy. Being randomized into the control arm, having undergone early parental loss before age 11 and having lower global severity of psychiatric symptoms (SCL-90) at 3 months postpartum were the only variables associated with later attrition. CONCLUSION This study provides key information for identifying mothers who may require specific support to avoid study attrition in trials evaluating a home-visiting program.