An item reduction analysis of the Group Questionnaire.

There is a constant tension between having measures short enough for daily practice and long enough to provide useful information. Although shorter measures are more convenient for clients, fewer items necessarily mean less information, a loss of psychometrics, and possible floor and ceiling effects. This study examined the effects of shortening the Group Questionnaire (GQ) on its clinical utility and psychometric integrity. Creation of a 12-item GQ (GQ-12) was done using archival data with 1,087 participants gathered from counseling centers, nonclinical process groups, outpatient psychiatric hospitals, and an inpatient state hospital. Testing of the loss of clinical utility was conducted using archival data from 432 group counseling center patients. Analysis for creation of the GQ-12 was done using multilevel structural equation modeling. Items were selected using clinical judgment and statistical judgment considering interitem correlation and factor loading. Model fit was analyzed in comparison with the standards in the literature and with the full-length GQ. Loss of clinical utility was analyzed by comparing alerts generated by the GQ-30 with alerts generated by the GQ-12. The GQ-12 has good model fit and acceptable reliability. A significant number of alerts were lost by reducing the items (status alerts: sensitivity 82-93%, specificity 97-99%; change alerts: sensitivity 52-67%, specificity 99-100%). This study suggests that although it is possible to create a psychometrically sound, shortened version of a feedback measure, clinicians should be aware that helpful information is lost. (PsycINFO Database Record