Page 237 - PhD GT

dependence in the development of psychological and social problems and the reverse: the development of dependence in people with psychological and social problems (Drummond 1992; Robins 1992). Measures of both psychological health and social functioning are commonly used in treatment outcome studies. In the present study, the measure of psychological health used was the General Health Questionnaire (GHQ) and the measure of social satisfaction used was the Social Satisfaction Questionnaire (SSQ).
Since dependence and change in dependence were the focus of the study, an additional measure of dependence was sought. In the absence of a suitable measure of substance dependence, a measure of one component of dependence, namely impaired control, was used. The latter two parts of the Impaired Control Scale performed very similarly to the LDQ and positive correlations between dependence scores and scores on these items were shown. The first part of the scale, consisting of the attempts to control items, performed differently, showing negative correlations with dependence as measured by the LDQ. The first part of the Impaired Control Scale was therefore included in the model constructed to investigate the predictors of change in dependence.
Coping was the main variable in the hypotheses regarding elucidation of the nature of change. Choice of the CBI to measure coping was described in Chapter 6. The five variables: psychological health and social functioning, dependence at intake, attempts to control and coping were subjected to factor analysis and shown to be independent variables. This justified their inclusion in the model to be used for the purpose of predicting change in dependence. Carr-Hill et al. (1987) have described the way that erroneous conclusions can be drawn when variables, assumed to be independent and entered into regression analysis as independent variables, are not in fact independent. They also caution against the arbitrary selection of variables and the making of unfounded assumptions about the direction of the relationship between them which can lead to erroneous conclusions in spite of apparently sound statistical procedures.
In order to examine the predictors of change in dependence, two types of regression analysis were conducted using the model described above. Linear regression analysis was conducted with change in dependence as the dependent variable. This approach was used in order to retain the whole spectrum of change that occurred. The sole significant predictor of change in dependence at both three months and twelve months was dependence at intake. In order to examine the predictors of the dichotomous outcome change/no change independence, based upon the criteria for clinically significant change, logistic regression analysis was conducted. Logistic regression analysis “is especially useful when the distribution of responses on the dependent variable is expected to be nonlinear with one or more of the independent variables” (Tabachnik and Fidell 1996 p. 575).
232