Page 231 - PhD GT

detected.
Tests of the reliability of individual instruments used were reported in the chapters in which
the instruments were described.
10.3.4 Reliability of the interview method
The implementation of those recommendations made by Babor et al. (1987b) and discussed in Chapter 3 which were relevant to the behaviour of the researchers were described in the procedures section 7.2 of Chapter 7. Respondents were guaranteed confidentiality by separation of the research requirements from the clinical process; on no occasion were data collected by the staff member responsible for the treatment of the patient.
As there were three research assistants in addition to myself who collected the data (one at intake and three at three months and twelve months), weekly meetings were conducted during the follow-up periods for discussion of data collection protocols. Checks were made on the wording of interview questions and the recording of information in order to standardise the process and to maximise the reliability of this method of data collection.
The short interview schedule (see Appendix 5) enquired about use of alcohol or heroin in the past week. The decision to enquire about the past week was that this time frame matched that of the main study measure, the Leeds Dependence Questionnaire and the relationships between use and dependence were the ones of greatest interest. A quantity-frequency method was adopted for simplicity and speed of collecting the information in the present study. This latter requirement was considered to be important particularly at the point of initial data collection in order to minimise the intrusiveness of the study procedures; study participants were attending the agency for treatment of their alcohol or drug problem and thus provided an opportunistic sample for the study. Even though
their consent was invariably obtained before administration of any of the study procedures, it was deemed important to keep to the minimum the amount of their time these procedures took up. Miller and Del Boca (1994) suggested that this strategy may “significantly underestimate consumption in problem drinkers..... and be insensitive to fluctuations that occur before and after treatment” (Miller and Del Boca 1994 p.113); this consideration was thought to be relevant to treatment outcomes studies in which the quantity and frequency of drinking was a main outcome measure and therefore required maximum accuracy. As this was not the purpose of the present study, it was thought acceptable to sacrifice maximum accuracy for the sake of minimal intrusion, within reason. The simplicity of the two questions about use during the previous week was thought to have the further advantage of greater inter-rater reliability. The sorts of correlations among use
226