higher percentage (nearly one half) of staff in recovery compared to outpatient programs. Total staff per patient ratios in Target Cities programs and correctional programs were about half of those in the Critical Populations programs. Other important associations include the significant positive relationships between staffing levels per patient and the provision of medical services, particularly in the use of contract MDs and RNs (p =0.0043), as well as weaker relationships between total staff per patient (p=0.0623) and staff MDs and RNs (p=0.0993).
Returning to Table 1, it is clear that not all of the measures of structure and management listed there have been discussed in this paper. In the simple descriptive analyses, for example, there were no particularly interesting relationships discerned among therapeutic emphasis and other management/structure/service technology variables; most SDUs stressed individual counseling and no drugs, alcohol or marijuana across treatment modalities and CSAT program types. In addition, information about ownership and affiliation is only available for a subset of the SDUs with matching patient-level data in NTIES. Among this subset, the most common structure was a free-standing, non-residential private nonprofit SDU. The influence of these structural variables and the relationship of leadership variables to treatment outcomes will be more fully explored in the multilevel analyses (discussed below).
From these exploratory analyses, it is clear that there are a number of interrelationships among these structural, management, and primary work factors that may have direct or indirect effects on substance abuse treatment outcomes. These direct or indirect effects are also likely mediated through patient characteristics, which have not been discussed in this analysis. Gaining a fuller understanding of how financial management influences human resource management and in turn service technology (and ultimately treatment outcomes), controlling for patient characteristics, will be an important goal of the next phase of multilevel, multivariate analysis.