*Poster 13: Strategies for Identifying Factors Predicting Participant Attrition and Wave Nonresponse in the Traumatic Brain Injury Model Systems Program

Objective: To put forth an analytic strategy for identifying participant and nonparticipant factors predictive of attrition and wave nonresponse in the Traumatic Brain Injury Model Systems (TBIMS) program, a multi-site, longitudinal study of outcomes in individuals with traumatic brain injury (TBI). Preliminary findings will also be presented and discussed. Design: Retrospective analysis of TBIMS data collected from 1987 through 2009. Setting: Data were gathered through semi-structured interviews conducted in-person or via telephone by TBIMS staff. Participants: Individuals 16 years or older with TBI ranging from moderate to severe, hospitalized within 72 hours of injury, and given acute care and inpatient rehabilitation in a model system hospital (N=9844). Interventions: Not applicable. Main Outcome Measures: For demographic variables (eg, age), appropriate items from TBIMS Form I assessment. For psychosocial participant variables (eg, employment status) and nonparticipant factors (eg, site), appropriate items from TBIMS Form II assessment. Results: The relation between single variables and attrition will be determined using chi-square tests for categorical variables and t tests for continuous variables. Selected variables will include those previously associated with attrition in TBIMS (eg, socioeconomic status) and in longitudinal studies of other populations (eg, functional status). Exploratory analysis of participant factors (eg, community integration) and nonparticipant factors (eg, recruitment benchmark changes) will also be conducted. Variables significantly associated with attrition will then be included in a logistic regression to determine the strongest predictors and begin developing a predictive model. A similar set of analyses will investigate factors related to wave nonresponse. Conclusions: Determining the constellation of participant and nonparticipant factors associated with attrition and wave-nonresponse is a necessary first step toward improving the validity of longitudinal research results regarding TBI outcomes. Specifically, such information could help identify participants at high risk of attrition who require additional investigator effort for retention or provide empirical bases for weighting cases in future statistical analyses.