Missing data and errors of measurement are common problems in studies of the elderly. The two problems are related because investigators must frequently choose between detailed assessments, which may be more difficult to obtain from the oldest old and the impaired, and brief assessments obtainable from proxies or computerized records which may increase measurement error. A particular problem with nonresponse occurs when missing data are related to measures of interest, such as cognitive function. Measurement error may also be greater in studies of the elderly both because of greater variability of some physiologic measures and because self-reported information may be affected by memory, attention or depression. The long-term objective of this application is to extend and apply recently developed statistical methods to account for missing data and measurement error in studies of the elderly. Specific questions to be addressed include: l) adjustment for nonresponse in estimating cognitive decline in longitudinal studies; 2) accounting for measurement error in evaluating the consequences of cognitive impairment; 3) adjustment for greater variability of systolic blood pressure in the elderly when determining the risk of cardiovascular outcomes; 4) accounting for the difficulty in obtaining objective measures of physical function from the most impaired; 5) assessment of the relative accuracy of self-report versus claims data to identify myocardial infarction; and 6) development of improved measures of comorbidity in linked Medicare and Medicaid databases. A key feature of several of the approaches to be applied is incorporation of data from a sample of nonrespondents who are followed up or from an external validation study to improve estimation of effects of interest. The methods will be applied to data from the Established Populations for Epidemiologic Studies of the Elderly and to linked Medicare and Medicaid claims data available for over 10 years in New Jersey. Approaches described will be applicable to the design and analysis of many studies of aging.
Showing the most recent 10 out of 30 publications
