Drug and Alcohol Dependence 83 (2006) 218–224 Predictors of retention in methadone programs: A signal detection analysis Steven W. Villafranca ∗ , John D. McKellar, Jodie A. Trafton, Keith Humphreys Center for Health Care Evaluation, Veterans Affairs Palo Alto Health Care System & Stanford University School of Medicine, CA, USA Received 21 June 2005; received in revised form 14 November 2005; accepted 15 November 2005 Abstract Retention in Opioid Agonist Therapy (OAT) is associated with reductions in substance use, HIV risk behavior, and criminal activities in opioid dependent patients. To improve the effectiveness of treatment for opioid dependence, it is important to identify predisposing characteristics and provider-related variables that predict retention in OAT. Participants include 258 veterans enrolled in 8 outpatient methadone/l-alpha-acetylmethadol (LAAM) treatment programs. Signal detection analysis was utilized to identify variables predictive of 1-year retention and to identify the optimal cut-offs for significant predictors. Provider-related variables play a vital role in predicting retention in OAT programs, as higher methadone dose (≥59 mg/day) and greater treatment satisfaction were among the strongest predictors of retention at 1-year follow-up. © 2005 Elsevier Ireland Ltd. All rights reserved. Keywords: Methadone; Retention; Signal detection 1. Introduction Opioid Agonist Therapy (OAT) is an evidence-based treatment for opioid dependence [Department of Veterans Affairs/Department of Defense (VA/DoD, 2001); Practice Guideline for the Treatment of Patients with Substance Use Dis- orders: Alcohol, Cocaine, Opioids (PGTPSUD, 1995)]. OAT has been shown to reduce heroin and other opioid use by preventing withdrawal, reducing craving, and blocking the “high” associ- ated with use of illicit opioids (Faggiano et al., 2003; VA/DoD, 2001). In addition, OAT has significant public health benefits including reductions in criminal activity, HIV-risk behaviors, and the negative medical, legal, and social consequences of sub- stance use (Ball et al., 1988; Metzger et al., 1993). Although OAT consistently produces long-term reductions in opioid use (Newman and Whitehall, 1979; Marsch, 1998; Strain et al., 1999; Johnson et al., 2000), patient dropout is a problem (Hubbard et al., 1989; Simpson, 1981; De Leon, 1991; Marlatt et al., 1997; Simpson et al., 1997; Rabinowitz and Marjefsky, 1998; Mertens and Weisner, 2000). Retention is a critical issue because discharge from OAT programs is followed by relapse and other adverse outcomes in the majority of opioid dependent patients (Gerstein et al., 1994). Treatment dropout can lead to overdose, HIV and Hepatitis C infection or transmission, dan- ∗ Corresponding author. Tel.: +1 650 493 5000x27131; fax: +1 650 617 2736. E-mail address: Steven.Villafranca@va.gov (S.W. Villafranca). gerous criminal behavior, and premature mortality (Davoli et al., 1993; Caplehorn et al., 1996; Zaric et al., 2000). The behavioral model of health care utilization (Aday and Anderson, 1974; Aday et al., 1999; Anderson, 1995; Anderson and Newman, 1973) provides a framework for examining factors associated with patient utilization of health care services. Most studies utilizing this model focus on variables falling into three patient-centered categories: predisposing, enabling, and need. Predisposing characteristics occur before illness onset and include demographic characteristics such as age, ethnic back- ground, and employment status. Older age, which has been posited to be related to increasing dissatisfaction with addict life- style and health concerns, has been associated with retention in methadone programs in several studies (Mertens and Weisner, 2000; Saxon et al., 1996; Stark, 1992). Enabling characteristics traditionally refer to “means” factors such as insurance coverage or income, but have more recently been extended to include social support (Beckman and Kocel, 1982). Enabling characteristics, such as marital status, have also been found to predict retention rates (Babst et al., 1971). Need characteristics stem from the severity of the illness. Low composite scores on the Addiction Severity Index (ASI), such as legal problems (Babst et al., 1971; Saxon et al., 1996) and psychiatric problems (McLellan, 1983), are common examples of need characteristics predictive of retention. A recent study by authors of this model (Phillips et al., 1998) notes, however, that provider-related variables have been largely neglected. Phillips et al. (1998) define provider-related variables 0376-8716/$ – see front matter © 2005 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.drugalcdep.2005.11.020