Services Research Outcomes Study (SROS)

COMPARING BEHAVIOR BEFORE AND AFTER THE SROS TREATMENT EPISODE

The effects or outcomes of treatment are evaluated in SROS through two methods: "before/after" (or "pretest/post-test") comparisons and regression analysis. This section presents an overview of the two methods. A more detailed description of the methods is presented in Appendix A.

Before/After Analysis. The before/after design compares behaviors measured before and after an intervention. Specifically, SROS compares the behaviors and characteristics of clients discharged from the index SROS treatment episode (i.e., the episode selected from records of clients discharged between September 1, 1989, and August 31, 1990) by comparing group rates of behavior (e.g., drug use, criminal activity, employment, living arrangements, or physical health) during the five years before and after the index treatment episode.

Every outcome examined is one on which individuals could change for the better or worse, and therefore, the group and subgroup rates could increase or decrease — i.e., the group become better or worse off — after treatment.

Regression Analysis. To help relate outcomes to associated variables, two types of regression analyses are used. For continuous outcome variables, "ordinary least-squares" regression models are used. For dichotomous (binary) outcome variables, SROS uses logistic regression models.

The following model is employed for continuous variables:

Y_AFTER = _ + _Y_BEFORE + _ ß_i X_i + _ ß_ii X_ii + ....._ ß_kX_k + e

where Y_AFTER denotes the value of a continuous outcome variable reported for the five years after the SROS treatment period; Y_BEFORE denotes the value of the same variable reported by the same individual for the five years before the SROS treatment period; X_i‘s are other explanatory variables; Greek letters represent regression coefficients; and "e" is a random error term.

The model is a conditional change model, relating an individual's outcome after treatment to his/her status before treatment.

For dichotomous variables (i.e., variables that alternate between two values [e.g., used drugs in five-year period after treatment or did not use drugs in five-year period after treatment]), logit analysis is employed, using the following "unified model":

logit(D2) = b0 + b1*D1 +.... b2*X + b3*Z

Where D2 denotes the after-treatment measurement of the dichotomous outcome (i.e., D2=0 if no and D2=1 if yes), and D1 denotes the before-treatment period of the same dichotomous outcome, (i.e., D1=0 if no and D1=1 if yes). More detailed descriptions of these models are presented in Appendix A.

This page was last updated on June 03, 2008.

SAMHSA, an agency in the Department of Health and Human Services, is the Federal Government's lead agency for improving the quality and availability of substance abuse prevention, addiction treatment, and mental health services in the United States.