Representing 62% of the treatments centers, hospitals overwhelmingly constituted the largest group, with only 31% of centers classified as for-profit. Centers of both types averaged approximately 115 clients each, ranging, on average, over four or five levels of care. The demographic composition of the couselor workforce was 57% female, 13% minority group members, 50% with graduate degrees, 56% certified, and 37% in recovery. Counselors' salaries averaged $30,134, with a range of $15,000 -- $62,640. With respect to the client population, an average of 54% had experienced a relapse. Twenty-nine percent will in a program due to a court mandate. Twenty-six percent of clients were covered by Medicaid and 18% were covered by managed care. Discussion

McNulty et al. tested ten hypotheses in four conceptual blocks of variables. Their "key hypotheses" (p. 173) concerned the role of management practices. They were careful, in formulating hypotheses, not to assume causal relationships between management practices and center and workforce characteristics. They did note, however "the model considers center characteristics first because…workforce and client composition are perhaps shaped in part by the structural/operation features of treatment centers (p. 173). Factors under study in their hypotheses are not independent of one another, as the tabulation of multivariate sustainability indicates.

In discussing reliability of the study, there are several points that are worthy of consideration. The first is with respect to the nature of the factors considered in forming the hypotheses and then subsequently measured and studied. Examination of the factors shows that they are objective measures and, in some instances, binary. The nature of the facility, for example, is binary; it is either for-profit or hospital-based. Factors such as counselor-client ratio, size of the facility, classification of facility (inpatient, outpatient, or both), salary ranges, and opportunities for mobility are generally stable but can, dependent upon circumstances, be changeable to an extent. Factors such as education levels and attainment of certification may also be relatively stable in that facilities may have established mandates for hiring. Factors such as gender composition of the workforce, racial and ethnic composition, and recovery status among members of the workforce has the most potential for change within and across institutions; as with education and certification attainment, however, organizations may have mandated quotas to meet in hiring practices.

The subjective factors in the study are the scaled-response survey questions completed by counselors in the facilities studied. The study by McNulty et al. does not have any way of controlling or accounting for the emotional states of counselors or personal or professional experiences they may have had that will influence their responses.

Survey questions cannot control for individuals who, as a matter of practice, never select the strongest response in a set. Responses to questions about participatory management will be influenced by each employee's experiences. For example is the statement "The center's management makes sure that employee concerns are heard before decisions are made" (p. 188). Responses could be affected by an employee's own experience, the experience or a colleague, hearsay, or an employee's perception of a typical management response. There is no provision for factoring in an employee's length of service and whether dissatisfaction, as expressed in this response or others, reflects a single incident...

As above, responses are subjective and depend, to a small extent, on individual's customary response habits (e.g., choosing to avoid strongest responses or tending to select "middle-of-the-road" responses). To a greater extent, responses may be skewed depending upon whether respondents answer based on overall feelings or in reaction to a specific incident. For example, the first item in the category of organization commitment states "As soon as I can find a better job, I will leave this center" (p. 188). An individual's response may be affected by a negative encounter with another staff member or client; if the staff member or client left the facility, the individual's response might be different. An individual's response might be a reflection or management behavior or practice, or a personal agenda which management may not be aware of and over which it has no control.
Results from the study by McNulty et al. were based on a large sample population in two waves, so it is reasonable to conclude that the questions raised about the subjectivity of certain responses would not invalidate the findings. McNulty et al. used WLS regression equations to account for the roles of participatory management and organizational commitment in counselor turnover and controlled separately for each additional block of variables.

The methodology of the study is solid. McNulty et al. carefully evaluated a number of factors that could affect turnover. The model used a dependent variable and examined factors categorized in four blocks. Analysis of the data revealed "significant heteroscedasticity in the models" (p. 177); the researchers compensated for this variance in statistical measures by weighting scores, since it was determined that prediction errors were tended to be greater with smaller centers that employed fewer workers.

Careful methodology and statistical analysis resulted in a valid study, the ramifications of which are important to substance abuse treatment centers that are faced with high turnover rates as well as for centers who endeavor to take a proactive approach in developing and executing management practices that support counselor retention. The study could easily be replicated for any group of centers, provided that a sufficiently large sample was studied. Important considerations in addition to the large sample size would be inclusion of hospital-based and for-profit centers and composition of client base (inpatient, outpatient, Medicaid, and managed care).

McNulty et al. noted that their analysis had some limitations and made recommendations for future research. Key would be studies that "would help to sort out causal processes and establish the nature of direct and indirect effects of predictors on turnover rates" (p. 187). As has been shown, the study by McNulty et al. And future studies could be used to inform management decisions. Another application of the studies, those which have been undertaken as well as those suggested for the future, is to inform educational programs and professional development opportunities within the fields of social work, psychology, and administration as they pertain to substance abuse treatment centers. Awareness of potential factors influencing turnover could be a first step in creating not only effective policies but also effective programs to build and maintain a supportive culture in which counselors can work to provide…