Evaluating agreement between clinical assessment methods
In clinical assessment, it is necessary to determine the extent to which alternative measurement tools yield equivalent information. Using alcohol consumption data obtained from 187 clients of an impaired-driver treatment program, this paper assesses the equivalence of information collected by microcomputer and clinician interviews. First, methods by which equivalence is routinely assessed (i.e., product-moment correlation coefficients and t-scores associated with differences between group means) are evaluated, highlighting the limitations of these approaches. Second, structural equation modeling is examined as an underutilized analytic strategy for examining convergence. Third, the advantages and disadvantages of a relatively novel approach to assessing equivalence, i.e., examination of agreement at the individual level, are discussed. Finally, a general strategy for establishing extent of agreement is recommended.