Attrition and use of proxy respondents and auxiliary information in the sicilian neuroepidemiologic study

Two-phase prevalence surveys with screening (phase 1) and examination (phase 2) are useful for some chronic diseases. Attrition, which may bias estimates, occurs in either phase because some eligible subjects die before contact, some refuse to cooperate, some are incapacitated, and some are unreachable. This investigation relates to a survey of neurologic diseases conducted in three municipalities of Sicily (prevalence date, November 1, 1987) and considers the attrition experienced and the use of proxy respondents in phase 1 and auxiliary information in phase 2 to offset, in part, this attrition. Regarding case finding, the salvage effort was more productive for deceased and incapacitated subjects. The age, sex, and household size of the subject were related to phase 1 attrition, but only age was related for all four attrition groups-deceased, refusing, incapacitated, and unreachable subjects. On the basis of information from proxy respondents, the educational levels of refusing and unreachable subjects were compared with those of subjects screened directly. Refusing subjects were less educated, and unreachable subjects were more educated. The proxy respondent performance, as indicated by "don't know" responses, was better with screening items concerning facial paralysis and mouth drooping (and not limb sensory abnormalities or impaired consciousness), better with younger subjects, and worse with refusing or incapacitated subjects. Am J Epidemiol 1994;139:219-28.