Another limitation in our data was theabsence of some measure of cognitive functioningas a covariate or a risk factor. To control for this variable, we required interviewers toproceed with questioning only if they had no doubt about the ability of respondents to understandand respond to questions. Our data therefore reflect responses of a cognitively intact,community-residing subpopulation of older adults; prevalence and risk factors shouldbe considered in that context. Generalization of our results to what may be the group mostat-risk for mistreatment, the cognitively impaired elderly, is not appropriate. For thisat-risk group, and particularly for members of this group living in residential settings, alternativemethods are required and will probably resemble the sentinel approach used with children. We did not include individuals who did not have a landline (i.e., those with only cell phones), introducing the potential for bias; however, our data were nationally representative and were weighted by census estimates,
increasing the generalizability of our findings.