Social problems from drinking in the Swedish general population: measurement and reliability

Abstract

Aims: The paper constructs and tests summary measures for different areas of alcohol‐related problems, using general population data. The main emphasis is on the rather unexplored area of measuring social harm from drinking.

Methods: The data analysed is a Swedish national survey on drinking behaviour and related consequences, collected in 2002 by means of telephone interviews with 5469 adult Swedes. In total, 38 items on both personal and social problems by respondents attributed to their drinking were subjected to factor analysis in order to identify different problem areas for which summary measures could be constructed. The psychometric properties of the measures were then tested.

Results: In line with others' findings, the problem items tended to load on a single factor. After Varimax rotation, 11 factors were built, only in part fitting a logical conceptual pattern. In the light both of the dimensions identified in the factor analysis and of face validity, measures of five areas of alcohol‐related problems were constructed: impaired self‐control, chronic health problems, public disorder, interpersonal problems and alcohol‐related social problems. The last measure is a summary scale including both items covered by public disorder and interpersonal problems scales, as well as a few other items. In terms of internal consistency and test–retest reliability (analysing a smaller sample from a test–retest pilot study), the five measures showed satisfactory psychometric properties.

Conclusions: In the light of others' findings, the measures developed here seem to be more consistent and reliable than a number of other scales. In order to establish comparability, it seems reasonable to develop and test similar measures across different drinking cultures and, perhaps, modify them thereafter.

• Dimensionality of harm from drinking
• social consequences
• summary measures
• general populations

Notes

1. Although questions were asked about frequency of experiencing problems, the variability beyond the once in the past 12 months was rather low.

2. A correlation matrix (Pearson's r) shows that the four ‘core’ scales are certainly correlated to each other, but the relationships are not particularly strong (0.36–0.51).

3. Three questions from the main study had not been included in the pilot: talking to a doctor or some other professional, quarrels with partner and being arrested for drunken behaviour. A few items included in both studies had slightly different wordings and have been adjusted for this analysis to be as similar as possible.