Combining Direct and Proxy Assessments to Reduce Attrition Bias in a Longitudinal Study

• DOI: 10.1097/wad.0b013e31826cfe90
• Publication Date: 2012-09-19T08:34:09Z
• Authors: Qiong Wu; Eric J. Tchetgen ...; Theresa L. Osypuk; Kellee White; Mahasin Mujahid; M. Maria Glymour
• Keywords: Proxy (statistics); Statistic; Longitudinal Study

Retaining severely impaired individuals poses a major challenge in longitudinal studies of determinants dementia or memory decline. In the Health and Retirement Study (HRS), participants complete direct assessments biennially until they are too to interview. Thereafter, proxy informants, typically spouses, assess subject's cognitive function using standardized instruments. Because there is no common scale for assessments, reports often excluded from analyses. The Aging, Demographics, Memory (ADAMS) implemented full neuropsychological examinations on subsample (n=856) HRS participants, including respondents with prior core Using data ADAMS, we developed an approach estimating probability composite score basis either interviews. prediction model achieved c-statistic 94.3% DSM diagnosed ADAMS sample. We applied these scoring rules sample born 1923 earlier (n=5483) biennial 1995 2008. Compared estimates excluding cohort, incorporating information increased estimated prevalence by 12 percentage points 2008 (average age=89) suggested accelerated rates decline over time.