Reliability and validity

The Composite Abuse Scale has the ability to classify women according to type and severity of abuse, a common criticism of other current measures of abuse. The original scale, developed in 1995, contained 74 items comprised four sub-scales (Severe Combined Abuse, Physical Abuse, Emotional Abuse and Harassment) and was validated on a convenience sample of nurses (n=427). Further validation on a sample of general practice patients (n=1896) and emergency department patients (n= 345) has resulted in the current 30 item version. The second study confirmed the four CAS sub-dimensions i.e. the same four factors as in the preliminary validation study on a nurses sample, although the order of the factors was different with the Emotional Abuse factor rather than the Severe Combined Abuse factor accounting for the majority of the variance in this sample.

These four factors would seem to exhibit good internal reliability i.e. the items reflect common underlying constructs. Bland and Altman (1997) discusses how for comparing groups, Crohnbach’s alpha values of 0.7 to 0.8 are regarded as satisfactory but that for clinical applications much higher values of alpha are needed (>0.9). The internal consistency reliability of the CAS was 0.85 or above and for the majority of sub-scales greater than 0.9 and the corrected item-total correlations were generally high (more than 0.5).

The Composite Abuse Scale has demonstrated face, content, criterion, and construct validity. There is strong evidence in the literature that to weight items for multi item scales does not produce advantages in measurement accuracy. The second CAS validation study investigated how upset women were by the individual acts by their partners. As the degree of upset was found to be highly correlated with the frequency of the behavior, it is recommended that CAS not be weighted by the upset scale.