Objective: To test reliability and construct validity of an alcohol-related harm scale widely used in North American surveys.
Method: Data base: three representative general population household telephone surveys in Ontario, Canada (1994: N = 2,022, response rate 63%; 1995: N = 994, response rate 63%; 1996: N = 2,721, response rate 64%).
Statistical analysis: psychometric analysis of internal consistency (Cronbach); Mokken scaling to test homogeneity of underlying construct; tests for construct validity by measuring associations with similar scales.
Results: The scale showed high internal consistency and homogeneity of the underlying construct. The correlations with the CAGE and ICD-10 criteria for dependence ranged between 0.5 and 0.7.
Conclusions: The harm scale is measuring a unidimensional construct, but one which is not distinct from that measured by the CAGE or dependence criteria.