The terminal drop hypothesis: Fact or artifact?

Abstract

Findings from eight longitudinal investigations were reviewed in order to evaluate the terminal drop hypothesis. The concept of terminal drop was seen to be more useful when partitioned into its component parts: the relationship between the level of cognitive performance with survival and with distance from death, and the relationship between changes in cognitive performance and death. Health status and age at death emerged as important intervening variables in understanding the terminal drop concept.

Some determinants of attrition in prospective studies on aging

“…This finding is thus an example of what has come to be known as "terminal drop" (Kleemeier, 1962; Riegel, Riegel, & Meyer, 1967). The reasons for this so-called terminal drop in cognitive activities have been discussed previously at length (e.g., see Siegler, 1975). Also the finding that all of the OARS subscales, except economic resources, had moderately high correlations with the first variate suggests a rather general impairment in both mental and physical health in this group, that was not apparent in the disinterested group.…”

Population Inference with Mortality and Attrition in Longitudinal Studies on Aging: A Two-Stage Multiple Imputation Method

“…Because these individuals typically go unmeasured, the results will be upwardly biased towards inferences of less decline in the population than is actually true. Findings of decline across a range of cognitive abilities associated with time to death provide evidence for "terminal decline" or "terminal drop" (for reviews see Berg, 1987Berg, , 1996Siegler, 1975;Small & Bäckman, 1999;Bosworth & Siegler, 2002). Of the longitudinal studies that have evaluated this hypothesis, significant findings have typically included spans between measurement occasions of less than 3 years (e.g., Deeg, Hofman, & van Zonneveld, 1990;Johansson & Berg, 1989;Smits, Deeg, Kriegsman, & Schmand, 1999).…”

Scale of reference bias and the evolution of health

“…Results of this study would be biased if survey drop-out is correlated with (subjective) health status.If respondents in poor health are more likely to cease participating in the survey,the survivors in the survey would be a selective group of persons with an above average health status.Consequently results would be upward biased. In the literature this phenomenon is referred to as "terminal drop" (see [18 ]). …”