The impact of measurement error on survey estimates of concurrent partnerships

This paper uses simulation to examine the role that measurement error may play in survey estimates of concurrent partnerships. Concurrent partnerships accelerate the spread of a sexually transmitted infection, making data on concurrency important for modeling and public health purposes. Methods for collecting data on concurrency typically rely on the reporting of dates. Little is known about the accuracy or reliability of estimates from such data. We examine the possible impact of two types of date reporting error here: unit heaping, which is often imposed by the survey instrument, and recall error. The impact of these errors depends on the underlying behavior pattern. When the interval that persons spend single between partners is small compared to the interval when partnerships overlap, it is easier for errors to create a concurrent partnership where none exists than to eliminate one that does. Under these conditions, measurement errors introduce a slight positive bias in estimates of the prevalence of concurrent partnerships, and a slight negative bias in the length of the overlap.