This type of analysis will allow us to identify the strongest factors or effects, among these many interrelated variables, that will enable policymakers to take effective action in improving substance abuse treatment program outcomes.
FORTHCOMING ANALYSIS: LINKING PROGRAM AND PATIENT-LEVEL FACTORS TO OUTCOMES
The preceding analyses are limited to program-level data, and more specifically to variables representing structural, management and primary work components in the substance abuse treatment process. The most important questions for substance abuse treatment policymakers and program administrators concern how these components or factors relate to patient outcomes. Working with researchers from the National Opinion Research Center (NORC) who designed the NTIES study, in the next phase of analysis, we will construct more fully specified models. Using these models, we will test a variety of hypotheses concerning causal relationships between structural, management, treatment and individual factors (including interactive and cross-level effects) to substance abuse treatment outcomes at the patient and program levels.
In their analysis of the SROS data using individual- and program-level variables, Schildhaus et al. (2000b:1882) presented a helpful diagram representing “assumptions that structure the regression models” of treatment outcomes. In this diagram, “pre-treatment variables” are shown in one box to illustrate that the “presence, absence or level of a[n individual] characteristic is predictive of the same characteristic after treatment.” These types of patient-level variables — age, gender, race/ethnicity, prior drug use, criminal behavior and legal pressure to enter treatment — are just some of the individual characteristics that might be