Incomplete Ascertainment of Mortality in a Nationally Representative Longitudinal Study of Community-Living Older Americans

Background In longitudinal studies of older persons, complete ascertainment of mortality is needed to minimize potential biases. To ascertain mortality in the National Health and Aging Trends Study (NHATS), investigators are advised to use its Sensitive files, which include month and year of death on most decedents who had not dropped out of the study. Because losses to follow-up are not insubstantial, ascertainment of mortality is likely incomplete. Methods We used linked Medicare data as the reference standard to determine the extent by which mortality is underestimated in NHATS through use of its recommended strategy. Ascertainment of mortality was compared between the two strategies over 10 years for 7,608 members of the 2011 cohort and 5 years for 7,498 members of the 2015 cohort. Results The Sensitive files did not identify a large number of decedents, leading to suboptimal sensitivity, ranging from 61.3% (2011 cohort, 10 years) to 75.5% (2015 cohort, 5 years). Some non-decedents were also misclassified as dead using the Sensitive files. Cumulative mortality rates were modestly lower for the recommended strategy, although the number of participants at risk decreased markedly over time. Mortality incidence rates were also modestly lower for the recommended strategy, with incidence rate ratios ranging from 0.88 (2011 cohort, 10 years) to 0.94 (2011 cohort, 5 years). Conclusions The strategy recommended by NHATS leads to incomplete ascertainment and, to a lesser degree, misclassification of mortality. Caution may be warranted when interpreting results of longitudinal analyses in NHATS that evaluate mortality using the recommended strategy.