Can we predict retention in longitudinal studies of substance use? Findings from the Australian Treatment Outcome Study

Background: Longitudinal studies are often threatened by difficulties with sample attrition, high rates of which threaten the validity of study findings. The present study examined methodological and participant characteristics associated with sample retention in the Australian Treatment Outcome Study (ATOS) across 3 years. Method: Follow-up interviews were conducted at 3-, 12-, 24-, and 36-months post baseline, with follow-up rates of 89%, 81%, 76% and 70%, respectively. Structured interviews measuring past-month drug use, mental health, criminal involvement and demographic characteristics were administered to participants at baseline and each follow-up. Data were analysed using multinomial logistic regression and generalised estimating equations to produce odds ratios with 95% confidence intervals. Results: Completing all follow-up interviews was associated with being in treatment (OR 3.62), using other opiates at baseline (OR 3.45), more years of schooling (OR 1.20), and having completed the previous interview (OR 35.04). A history of incarceration was independently associated with not completing follow-up interviews (OR 0.47). Conclusion: Retention can largely be predicted at study entry, and is unaffected by changes that occur in the interim. These findings highlight the importance of obtaining and maintaining comprehensive locator information, maintaining strong relationships with treatment agencies, as well as the necessity of patience, perseverance and flexibility. (C) 2015 Elsevier Ltd. All rights reserved.