Journal of Biometrics & Biostatistics

Estimating and Comparing Diagnostic Laboratory Performance with Weighted Estimating Equations

Abstract

Patricia Cooper Barfoot*, Stefan H. Steiner and R. Jock MacKay

Patient test outcomes from diagnostic testing laboratories can be indicators of laboratory performance. One method of proficiency testing compares test results across laboratories to flag non-compliant laboratories. Under this type of proficiency test, the statistical approaches for estimating the test result and comparing these across laboratories have important implications. The proficiency test of fecal occult blood testing laboratories in Ontario compare estimates of pass rate for a particular test by laboratory based on data observed in the present (latest) month. Estimates by laboratory are compared to an acceptance interval determined by data across all laboratories. A laboratory is classified as non-compliant when its monthly rate is outside the acceptance interval. We note that monthly sample sizes have an important impact on the probability that a laboratory is classified in error and differences in the number of tests conducted across laboratories should be considered. We present an alternative approach (Weighted Estimating Equations, or WEE) for combining historical data to improve the precision of the estimate of present performance in the case that performance changes slowly over time. The WEE approach uses all available historical data through estimating functions that down-weight past data. We compare the WEE approach to current practice through a real dataset of patient fecal occult blood test outcomes at laboratories in Ontario as well as simulated data. The study approach improves precision of the estimates and the power of a hypothesis test to compare estimates in order to reduce the risk of classifying a laboratory as compliant or non-compliant in error.