Page 64 - PhD GT

correlation between reports varies depending on the nature of the information sought, for example, hospitalisations or imprisonments due to alcohol consumption, total days drinking or abstinent, quantity of alcohol consumed (for example O’Farrell et al.1984; Maisto et al. 1983; Hesselbrock et al. 1983). With reference to specific information, both Hesselbrock et al. (1983) and Polich (1982) reported high correspondence between subject and collateral reports on any drinking and when errors did occur, they were in the direction of the respondent reporting higher consumption of alcohol than the collateral, further lending support to the caution that collaterals are less likely to have regular access to information about drinking, nor is their report usually validated in its own right.
In their review of studies of self-report in drug users, Maisto et al. (1990) also found that the inconsistency between collateral and self-report was in both directions and could not necessarily be determined by the nature of the collateral. In one study, probation officers reported less use than the drug using clients and in another, treatment staff reported drug use highly consistently with their clients. Their findings emphasise the way that no single factor contributing to potential bias should be taken into account in isolation but each addressed in the context of the others.
Sobell et al. (1979) compared the accuracy of mental health professionals’ judgement of patient intoxication with breath alcohol level and found agreement in 40-78% of cases; false negatives were more common than false positives. Babor and colleagues (1987b) suggest one source of such discrepancy may be the presence of tolerance which masks common signs of intoxication in regular heavy users. It could be argued that a positive breath or blood alcohol level in tolerant respondents may affect the accuracy of their report to a lesser degree than in non-tolerant respondents but there would be considerable variation between individuals. Blood and breath tests are the objective measures most commonly used for assessing current state and methods for immediate reading are available. In the present main study such tests were deemed to be excessively intrusive on first or subsequent contact and were therefore ruled inappropriate for use. Judgement by the person administering the questions and questionnaires was relied upon, recognising the potential difficulties outlined above.
3.3.2 Other measures used in the validation of self-report
Babor et al. (1987b) cite several studies in which consistency was found when self-report was tested by using multiple measures for alcohol related behaviour. Different methods of obtaining information, for example a retrospective or prospective diary method and an estimate of quantity and frequency (Redman et al. 1987), a time-line interview method (Sobell et al. 1982; Sobell et al.
52