Introduction: Understanding the role of risk factors associated with accelerated cognitive decline has important public health relevance for the aging US population. To test hypotheses that mid-life cardiovascular (CV) risk factors increase risk of dementia in old age, the Atherosclerosis Risk in Communities (ARIC) cohort was evaluated for dementia in 2011-2013. Anticipating challenges of complete dementia ascertainment in an elderly population, supplementary data sources were collected to aid analysis accounting for potentially informative attrition. Methods: ARIC enrolled 15,792 participants 45-64 years old at baseline (1987-1989) from four US communities and collected baseline CV exposure data. A complete neurocognitive battery including informant interview was completed 2011-2013, yielding an algorithmic diagnosis of dementia, mild cognitive impairment (MCI) or normal cognition. Cognitively impaired and a random sample of cognitively normal were selected for further clinical evaluation. Syndromic diagnoses (dementia/MCI/normal) and etiologic diagnoses were made by a panel of experts using standardized criteria. Standardized protocols, timely reports of reviewer reliability, ongoing training of reviewers and a web-based data management system were developed to ensure reliability, consistency and efficiency of data collection and review. Living participants not attending the visit were asked to complete the Telephone Interview for Cognitive Status (TICS). Based on dementia discharge codes and death codes identified during cohort surveillance, interviewer impression of hearing loss or cognitive impairment or need for a proxy during semi-annual calls, or random sampling, participants refusing the TICS are eligible for proxy dementia interview. Medicare claims are being examined to identify missed dementia discharge codes. Results: Of 11,017 participants alive as of 2011, 6495 (59%) completed the neurocognitive battery; 2937 of these (45%) were selected for, and completed, additional clinical assessments and classification by committee (algorithmic or reviewer diagnosis: dementia 5%, MCI 21%). Of the 4522 participants who did not undergo the neurocognitive battery, 1463 (32%) completed the TICS (dementia 7%), while 1627 of the remaining participants were eligible for proxy calls (903 completed to date, dementia 58%). Conclusions: Participants who were not examined were more likely to be cognitively impaired, highlighting the importance of collecting supplementary data to support analyses of midlife CV risk factors accounting for informative attrition. Collection of exam data in an elderly cohort is difficult and requires a multi-pronged approach, especially for an outcome highly correlated with attrition. The strategy described here could be applied in other settings.