Estimating age incidence from survey data with adjustments for recall errors.

In this paper we propose a method and discuss the type of data required to estimate age incidence rates from population survey data. While surveys are typically designed to estimate the prevalence of a disease or medical condition, they can also be used to estimate incidence rates. A limitation of survey data, however, is that recall is prone to errors. Three types of errors are common: telescopic, false negative, and false positive reports. Telescopic reports are thought to be the most common. We propose a method to adjust for recall errors by modeling the reported age of onset (ONST) and the time interval since the reported first occurrence of the medical condition (LAG). A number of models were examined using migraine headache data from over 10,000 subjects in Washington County, Maryland. Population surveys should be considered as a relatively inexpensive means for estimating the age incidence of medical conditions, especially for symptom based problems like back pain, asthma, mental illness, and serious headache. We recommend that data be collected on variables which can be used as surrogates for the different types of recall errors. Specifically, the age at interview, the date when the condition was cured or in remission, the severity of the condition, and possibly a specific inquiry as to how certain the respondent is in reporting the date of medical events, should be considered for this purpose.