Cognitive impairment is an important topic for longitudinal studies of aging, and one that directly affects ability to participate. We study bias in measured cognition due to non-response in the Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA). The much greater use of proxy interviews for impaired respondents in the HRS virtually eliminates attrition bias in measured cognition, whereas there is a noticeable bias in ELSA where proxies are infrequently used. Using Medicare claims data for the HRS we are able to compare cognitive impairment among dropouts post-attrition with that for continuing participants. There again we see the use of proxy interviews virtually eliminates a bias that would otherwise appear.