4. PREDICTING IMMEDIATE OUTCOMES
In the absence of a comparison group causal conclusions cannot be drawn from the data. Nonetheless it is important to use program and administrative data to identify client and system factors related to client outcomes (Courtney & Collins, 1994; Nash, Kupper, & Fra- ser, 2004). The immediate client outcomes that were analyzed in this part of the evaluation included (1) entering professional alcohol and other drug (AOD) rehabilitation treatment after participating in SAL services, (2) reducing chances of having children removed from the home while participating in SAL services, and (3) having children who were originally in out-of-home placements reunified while parents participated in SAL services. Research questions relating to these outcome variables included ADD FOOTNOTE:
How did parents’ intensity of SAL services relate to their entering professional AOD treatment?
How did parents’ intensity of SAL services relate to having children removed from their home while participating in SAL services?
3.How did parents’ intensity of SAL services relate to having children reunified while participating in SAL services?
Outcomes such as out-of-home placement (Courtney, 1995; Mason, Castrianno, Kessler, Holmstrand, Huefner, Payne, Pecora, Schmaltz, & Stenslie et al., 2003) and reunification (Harris & Courtney, 2003; Mason, Castrianno, Kessler, Holmstrand, Huefner, Payne, Pecora, Schmaltz, & Stenslie et al., 2003) are related to complex combinations of client and system variables that require equally complex analyses.
Using multivariate analyses to understand the relative contribution of several variables on social work intervention outcomes is an appropriate approach to understanding outcomes with complicated and interwoven contributing variables (Nash, Kupper, & Fraser, 2004). The multivariate method used to analyze variables contributing to the outcomes of this evaluation was Logistic Regression (Hosmer & Lemeshow, 1989). This section of this report, addressing the prediction of immediate outcomes, presents the analytic models, and concludes with the results of three models used to answer the three research questions.
The Logistic Regression4 models used were all hierarchical with blocked designs. Such mod- els include criterion, control, and predictor variables.
4 A Logistic Regression is a statistical procedure which determines the odds of obtaining an outcome, or dependent, variable, given a value for the predictor, or independent, variable or group of variables, based on a sample with those characteristics.
Idaho Pre-Treatment Program