Epidemiology and Infection
- Epidemiology and Infection / Volume 140 / Issue 08 / August 2012, pp 1515-1524
- Copyright © Cambridge University Press 2011
- DOI: http://dx.doi.org/10.1017/S0950268811001932 (About DOI), Published online: 23 September 2011
Methods
Association between covariates and disease occurrence in the presence of diagnostic error
F. LEWISa1 c1, M. J. SANCHEZ-VAZQUEZa2 and P. R. TORGERSONa1
a1 Vetsuisse Faculty, University of Zürich, Zürich, Switzerland
a2 Epidemiology Research Unit, SAC (Scottish Agricultural College), King's Buildings, West Mains Road, Edinburgh, UK
SUMMARY
Identification of covariates associated with disease is a key part of epidemiological research. Yet, while adjustment for imperfect diagnostic accuracy is well established when estimating disease prevalence, similar adjustment when estimating covariate effects is far less common, although of important practical relevance due to the sensitivity of such analyses to misclassification error. Case-study data exploring evidence for seasonal differences in Salmonella prevalence using serological testing is presented, in addition simulated data with known properties are analysed. It is demonstrated that: (i) adjusting for misclassification error in models comprising continuous covariates can have a very substantial impact on the resulting conclusions which can then be drawn from any analyses; and (ii) incorporating prior knowledge through Bayesian estimation can provide potentially more informative assessments of covariates while removing the assumption of perfect diagnostic accuracy. The method presented is widely applicable and easily generalized to many types of epidemiological studies.
(Accepted September 04 2011)
(Online publication September 23 2011)
Key Words:
Correspondence:
c1 Author for correspondence: Dr F. Lewis, Vetsuisse Faculty, University of Zürich, Winterthurerstrasse 270, Zürich, Switzerland, CH 8057. (Email: fraseriain.lewis@uzh.ch)
