Page 236 - PhD GT

10.4.3 Constructing hypotheses and models to test them
The nature of change in dependence was the subject of this study; measurement of change in dependence was described and exploratory analysis of the correlates of change was conducted. In order further to investigate which factors might be relevant in influencing such change, in determining whether change occurred or not, regression analyses were used. The question of whether change had occurred or not seemed an appropriate outcome in that investigation could proceed along the lines of distinguishing those characteristics which were or were not associated with this outcome. Presenting the outcome as a dichotomous measure addressed one of the issues raised in the investigation of the response choices, reported in Chapter 4; there was a suggestion that the LDQ might be less sensitive at the lower end of the dependence scale, implying that the measure might not be continuous, that the difference between scores at the high end may not be equivalent to the distance between scores at the lower end. Use of the dichotomous measure of change/no change takes account of this possibility. For this reason, logistic regression analysis was used. On the other hand, computation of a dichotomous variable of change and no change collapses the whole range of change in dependence scores into two categories and may restrict the opportunity to explore the nature of such change. Therefore, the amount of change in dependence was also examined and presented as the dependent variable in linear regression analysis.
The model proposed for investigating the role of different variables in predicting such change was implicit in the design of the study and the measures collected. This model was constructed in the following way: pretreatment demographic variables, functioning variables known to be associated with dependence and misuse and number of treatments attended were included for the reasons given below. Differences in patterns of use between the two main substances, coupled with the finding of a significant difference between the mean scores for dependence for each of the substance groups at intake were suggestive that the substance may effect the course of change in dependence. The difference in the age of the two groups was of such a magnitude that it could not be ignored as a potential source of influence on the outcome. Interestingly in this study cohort, women attending the agency (those in Sample 6) were less likely to be abstinent than men in both substance groups and more likely to be using on seven days of the week. They were using smaller quantities of alcohol but similar quantities of heroin to men. It was because of these observed variations that the three pre-treatment variables of substance, age and gender were included in the model for predicting the outcome in dependence.
Problems of psychological health and social functioning are commonly associated with dependence problems and constitute a focus for treatment. There is evidence for the contribution of
231