The present study aims to examine and compare the predictors of changes in self-reported sleep duration in older adults. A total of 2,294 participants over the age of 65 in the Shih-Pai Sleep Study were followed-up for an average of 3 years. According to the self-reported sleep duration at baseline and the results of a follow-up survey conducted 3 years later, participants were classified into three categories: mid-range sleepers (6-7 hours), short sleepers (≤5 hours), and long sleepers (≥8 hours). The main outcome variable was the sleep duration at the follow-up survey. A comparison of the results of the baseline and follow-up surveys revealed that only 45.9% of participants remained in the same spectrum of sleep duration in both surveys, with baseline long sleepers having the lowest consistency rate (27.6%). Only incident diseases, with the exception of prevalent diabetes and physical disability, predicted shortening of sleep duration. In contrast, prevalent morbidities or baseline characteristics correlated with the lengthening of sleep duration. The findings suggested that the self-estimated sleep duration fluctuated in a significant proportion of study participants over time. Predictors of lengthening of sleep duration were essentially different from predictors of shortening of sleep duration.