X hits on this document

Word document

Improving the Organization, Management, and - page 28 / 42





28 / 42

important to include as patient-level factors in the multilevel models.  

The second box of “treatment variables” that are added to the Schildhaus et al. model to predict outcomes include: treatment modality, treatment revenue, completion of treatment, length of treatment, counselor relationship, and use of drugs during treatment.  In their multilevel models using the NTIES data, Orwin and Ellis (2000) also included the use of case managers, staff training specialists, patient involvement in treatment planning, the matching of clients and tailoring of services by SDUs, counseling intensity, and the provision of vocational and academic services. All of these variables are represented in Table 1 of this paper and constitute a subset of the treatment/program-level factors that will be tested in forthcoming analyses.  A third box in the Schildhaus et al. diagram contributes measures of “post-treatment variables,” or circumstances and factors that will further aid in predicting patient outcomes in the follow-up period.  As in the SROS data, information about treatment services, patient activities, and other environmental factors in the follow-up period that might intervene to influence outcomes is limited in the NTIES data.

The pre-treatment, treatment and post-treatment variables are depicted in this diagram as having both direct and sequential effects on patient/program outcomes.  Schildhaus et al. (2000b) estimated linear and logistic regression models to analyze the relationships of these variables to reported drug use or non-use and the frequency of drug use.  As noted earlier, they found weak effects of the program variables in most models but cited limitations of both their measures and methods.  Multilevel models will allow us to examine not only direct or sequential effects of structural, management, and primary work variables but also potential cross-level or interactive effects with patient characteristics that might influence patient outcomes.  Our own diagram of a multilevel modeling strategy is presented in Table 8.  It shows the two levels of data (program

Document info
Document views57
Page views57
Page last viewedMon Oct 24 05:50:58 UTC 2016