Aging in disguise: age, period and cohort effects in mobility and edentulousness over three decades

By corroborating cross-sectional with longitudinal analyses, this study illustrates how cohort effects can confound trends over age and time by demonstrating that age effects dominated mobility, whereas cohort effects dominated edentulousness.

Abstract

By corroborating cross-sectional with longitudinal analyses, this study illustrates how cohort effects can confound trends over age and time. Mobility (walking difficulties) and edentulousness (toothlessness) were studied from 1968 to 2002 in a nationally representative panel aged 18-75 (5 waves, n ≈ 5,000) and ages 77+ at later waves (2 waves, n ≈ 500). Three analyses were done: cross-sectional 10-year age group differences in 5 waves, time-lag differences between waves (shifts across time) for age groups, and within-cohort differences between waves for 10-year birth cohorts followed over time. Complementary age-period-cohort models using logistic regression analysis evaluated differences. Both mobility and edentulousness have earlier been shown to be strongly related to age cross-sectionally. For mobility, cross-sectional and longitudinal analyses showed large changes, whereas time-lag analysis indicated no or marginal changes. Both cross-sectional and longitudinal results showed an exponential curvilinear age dependency for mobility limitations, with limitations becoming more usual in older ages. In contrast, cross-sectional and time-lag analyses of edentulousness showed large differences, whereas longitudinal analysis indicated no or marginal changes. Rates of edentulousness became increasingly lower for successively later cohorts in a curvilinear fashion. These patterns demonstrate that age effects dominated mobility, whereas cohort effects dominated edentulousness. Age-period-cohort models confirmed these findings. The cohort effect of edentulousness implies that the cohorts' movement through time gives a false impression of age and period effects in cross-sectional data.