Page 84 - PhD GT

alcohol, opiates (which included all opiates, mainly heroin) and “other”; the latter category was so called as a result of insufficient numbers to differentiate further substance specific groups.
Differences between the samples at each site gave a broader base for the validation of the LDQ than was achieved at a single site, and these differences related to age and main substance. Respondents at the Leeds centre were slightly younger with a mean age of 33.2, (SD = 10.9) than the Newcastle centre where the mean age was 34.9 (SD = 10.5). A higher proportion of respondents at the Newcastle Centre attended for problems of alcohol misuse, 60.0% compared with 46.6% at the Leeds centre (p < 0.001), while a higher proportion of respondents at the Leeds centre attended for problems of heroin misuse, 43.1% compared to 29.9% at the Newcastle centre (p < 0.001). For the purpose of the study analysis, samples from the two sites were treated as one.
Principal components analysis of the LDQ item scores for the whole sample yielded a first component which accounted for 53.9% of the variance with loadings greater than 0.5 for all items. As was found in the earlier validation study, loadings were lowest for items 5 and 8, (both at 0.59). When Principal Components Analysis was carried out separately for the substance specific sub-samples, similar component structures were found, though in the opiate sub-sample, the proportion of the variance accounted for by the first component was lower at 47.3% than for the alcohol sub-sample (58.7%) and the “other drug” sample (51.8%). These findings were deemed by the authors to lend support to the suggestion that a unitary construct of substance dependence was identified across the substance domains.
In order to examine the internal consistency of the LDQ, Cronbach’s alpha coefficient was computed for separate substance groups and found to be highest in the alcohol sub-sample at 0.92, with 0.89 for other drugs and 0.86 for opiates. These findings are again similar to the earlier validation study in which Cronbach’s alpha was computed at 0.94 for the whole sample (Raistrick et al. 1994 p. 568). The deletion of single items was not found to raise the value of alpha in the alcohol and the other drugs sub-samples, but in the opiate sub-sample the deletion of item 5 resulted in a slight increase in the value of alpha.
Item total correlation coefficients were found to be satisfactorily high with the relatively lowest loadings on items 5, 7 and 8; in the earlier validation study (Raistrick et al.1994), item total correlation coefficients were also found to be high with relatively lower coefficients for items 5 and 8 only.
The difference in dependence scores between the substance samples in this study reached statistical significance. As in the earlier study, the mean LDQ score for the alcohol sub-sample was lowest at 18.41 (SD = 7.9, range 0-30, n = 821) while the mean LDQ score for the opiate group was
72