![IconGems-Print](images/IconGems-Print.png)
Predicting Panel Attrition on a National Study: How We Can Optimize Locating Resources and Methods
Lauren McNamara
NORC at the University of Chicago
Melissa Heim Viox
NORC at the University of Chicago
Ned English
NORC at the University of Chicago
Papia Paul
NORC at the University of Chicago
Katie Dekker
NORC at the University of Chicago
Ron Hazen
NORC at the University of Chicago
Longitudinal studies depend on the retention of sample members across an extended period of time. The propensity of sample members to attrite is not randomly distributed. Thus, longitudinal studies run the risk of introducing bias via differential attrition due to movement and nonresponse. To gain a better understanding of who is likely to become lost to follow-up, analysis and modeling comparing lost to retained participants were performed using data from the National Children's Study (NCS) Vanguard Study, a national longitudinal pilot study of children's health and development. Our paper uses logistic regression to understand panel attrition and predict the propensity of a participant to become lost to follow-up. In so doing we pursue how attrition is dependent on recruitment strategy, demographics, geography, and number and distance of moves. Our results indicate that it may be most effective to focus resources in the form of extra locating effort or interim follow-ups on a subset of participants, decreasing efforts spent on "stable" participants. Our findings shed light on the relationship between participant characteristics and panel attrition in longitudinal studies.