Place and Attrition from Substance Abuse Treatment

Article excerpt

High rates of attrition are one of the greatest problems interfering with the effectiveness of substance abuse treatment programs. Though unexamined to date, this paper argues that where a client lives, where treatment is delivered, and the relationship between these locations may influence both voluntary and involuntary dropout. The notion of a "treatment ecology" is presented, and theories of mechanisms by which location may impact attrition are developed through a review of related literature. In other fields, lack of attention to theory prior to empirical investigations of contextual effects has prompted criticism and made interpretation of empirical results difficult. Neighborhood disadvantage, drug availability, community resources, restorative qualities of the neighborhood, and travel burden are identified as promising areas for study. Current conceptual models of treatment outcomes are first reviewed to identify how they might accommodate attention to place. Policy implications and special considerations for future empirical work on this topic are discussed briefly.


Substance abuse treatment programs in the U.S. that receive public funds logged more than 1.1 million admissions involving an illicit drug problem annually from 1992 to 1999.' However, completion rates and engagement on the part of the participant in these programs are generally low. For example, statewide estimates from the California Alcohol and Drug Data System suggest that nearly 70% of drug treatment clients admitted to residential programs stay less than 90 days (Maglione, Chao, & Anglin, 2000), and the growing body of retention studies typically reports rates of attrition from about 25% to 75%, depending on the treatment modality and definition of dropout (e.g., Joe, Simpson, & Broome, 1998; De Leon, Hawke, Jainchill, & Melnick, 2000; Hiller Knight, & Simpson, 1999; Veach, Remley, & Kippers, 2000; Lang & Belenko, 2000; Stahler, Cohen, & Shipley, 1993). Client attrition is considered by some to be "one of the greatest problems interfering with treatment effectiveness in substance abuse programs" generally (Stahler et al., 1993) and "one of the main challenges" facing clients coerced into treatment by the criminal justice system (Sia, Dansereau, & Czuchry, 2000). Indeed, previous investigations into the causes of attrition have been motivated not only by high dropout rates, but by the finding that a longer stay in treatment is among the few consistent predictors of better posttreatment outcomes (Hubbard, Craddock, Flynn, Anderson, & Ethridge, 1997; Simpson, Joe, Broome et al, 1997; see Anglin & Hser, 1990 for a review). Although prior studies have examined a broad range of factors related to client background, severity and nature of abuse, and components of the treatment program, as a rule they have come up short in explaining much of the observed variation in retention among treatment clients.

One aspect that remains unexamined by the literature on retention to date is the role of geographic and neighborhood context. This paper argues that there is good reason to believe that many of the contextual factors that define the treatment client's living and treatment environments and their interactions may play a causal role in both voluntary and involuntary dropout decisions. Despite passing mention of the importance of neighborhood context and treatment facility location in the literature on treatment outcomes more generally (e.g., Joe, Simpson, & Sells, 1994; Davis & Tunks, 1990-1991; Rohsenhow, Niaura, Childress, Abrams, & Monti, 1990-1991; Tucker, Vuchinich, & Gladsjo, 1990-1991; lguchi & Stitzer, 1991), thoughtful consideration of causal mechanisms to guide hypothesis generation and testing in the area of retention has been lacking. Attention to causal mechanisms is particularly important in neighborhood effects studies, however, where self-selection into neighborhoods makes endogeneity and confounding a real concern and where potential, systematic differences in variation across subjects within neighborhoods raise problems for traditional regression models. …