Purposes: We examined whether the percentage of items missing and the factors related to item missing differ across follow-up surveys, Variables targeted to examine missing items included health indicators (activities of daily living, cognitive function, self-rated health, Center for Epidemiologic Studies--Depression, and PGC Morale Scale), health habits (cigarette smoking, alcohol consumption, physical exercise, and relative weight), and socioeconomic indicators (educational attainment, income, and social networks).
Methods: Longitudinal data were collected at intervals of three years since 1987 through a national survey of Japanese adults aged 60 and over, At the baseline survey, a total of 2,200 interviews were completed from the list of 3,288 names. At the following three follow-up surveys, 1,671, 1,369, and 1,068 persons were reinterviewed respectively. Possible factors related to appearance of a missing item consisted of five aspects; 1) demographic variables (age and sex), 2) social status (educational attainment, existence of a spouse, and job status), 3) health status (activities of daily living and cognitive function), 4) cooperative attitude toward a survey, and 5) whether an item had been missing at the previous survey (s). Those factors were analyzed for each variable respectively. If a group with scaled or collective items had one or more missing items, we classified that group as a missing item group.
Results: 1. The percentage of cases with items missing was 5 percent or more for four variables; CES-D, PGC Morale Scale, income, and health habits. Those percentages were almost constant over the four surveys. 2. Factors related to appearance of items missing differed by psychological variables such as, CES-D and PGC Morale Scale, income, or health habits. Those factors had constant impact on appearance of items missing over follow-up surveys. 3. Regarding CES-D, PGC Morale Scale, income, or health habits, persons with an item missing at a previous survey, or who did not have a cooperative attitude toward the survey had a significant impact on an increase in the percentage of missing items.
Conclusion: Characteristics of persons with items missing differs among the variables, and those characteristics may contribute to the incidence of items missing in subsequent surveys.