Assessment of exposure misclassification bias in case-control studies using validation data

doi:10.1016/S0895-4356(96)00315-0

Journal of Clinical Epidemiology

Volume 50, Issue 1, January 1997, Pages 15-19

https://doi.org/10.1016/S0895-4356(96)00315-0 Get rights and content

Abstract

It has recently been shown that the bias due to misclassification of a binary exposure in case-control studies can be simply expressed in terms of so-called quality indices. Quality indices are measures of the degree of misclassification that are symmetric with respect to the true and measured exposure. They can be estimated from sub-study validation data and used to estimate the bias. Bias is not present if quality indices are nondifferential with respect to cases and controls. In this article the sampling distribution of these indices is investigated by developing formulae for their standard errors and these formulae are used to derive the standard error of the estimated bias. Simulation studies are used to confirm the validity of the formulae. The formulae are used to assess the bias in two studies of risk factors for Sudden Infant Death Syndrome (SIDS) and can be applied to the assessment of diagnostic tests.

References (11)

E Chouinard et al.
Recall bias in case-control studies: An empirical analysis and theoretical framework
J Clin Epidemiol
(1995)
RJ Marshall
Validation study methods for estimating exposure proportions and odds ratios with misclassified data
J Clin Epidemiol
(1990)
S Greenland
Statistical uncertainty due to misclassification: Implications for validation substudies
J Clin Epidemiol
(1988)
AT Marino et al.
Experimental quantiles of epidemiological indices in case-control studies with non-differential misclassification
Stat Med
(1995)
RJ Marshall
Misclassification of exposure in case-control studies: Assessment by quality indices
Epidemiology
(1994)

There are more references available in the full text version of this article.

Cited by (12)

Bayesian sample size determination for case-control studies with misclassification
2007, Computational Statistics and Data Analysis
Misclassification of exposure is high when interview data on drug use are used as a proxy measure of chronic drug use during follow-up
2004, Journal of Clinical Epidemiology
Citation Excerpt :
We defined time-varying cumulative exposure variables on the basis of pharmacy records from the date of baseline interview or January 1, 1992, until the end of follow-up as described earlier [4]. Misclassification of exposure was assessed and expressed in a quality index in which the ratio of the biased risk estimate and its gold standard plus 95% confidence limits are calculated [13]. The final study cohort consisted of 2,487 persons, with a median of 7.6 years of follow-up.
In many observational studies, the association between drugs and disease is analyzed with information from a baseline interview. We investigated the magnitude and direction of exposure misclassification by comparing interview data at baseline with prospectively gathered pharmacy data.
The study population for this study consisted of a cohort of 2,487 participants aged 71 years or older from the Rotterdam Study. Data on drug use were gathered at the baseline interview and through pharmacies during the follow-up period between January 1, 1991, and January 1, 1999. We assessed the sensitivity, specificity, and positive and negative predictive value of interview data as proxy measures of chronic use of calcium channel blockers (CCB) in comparison with longitudinal medication records from the pharmacy.
Only 3 of the 206 subjects (1.5%) who reported use at baseline did not use CCBs during follow-up. Of the 2,281 persons who reported no use of CCBs at baseline, however, 354 actually used CCBs during follow-up (15.5%). The difference between interview data and pharmacy records corresponded to a misclassification bias of 0.73 (95%CI: 0.52-1.02).
Misclassification of exposure was high when interview data were used as a proxy measure of chronic use during follow-up.
An empirical investigation of exposure measurement bias and its components in case-control studies
1999, Journal of Clinical Epidemiology
This study investigated bias due to misclassification of exposure in case control studies. Using a metaanalysis of data from a number of case control studies, in which a possibly misclassified exposure was validated against a more reliable one, estimates of misclassification indices and bias were obtained. The estimates were used to investigate whether misclassification indices, in particular, sensitivity and specificity and so-called quality indices, are typically nondifferential with respect to cases and controls. It is concluded that quality indices do not show any tendency to be either lower or higher in cases than controls. On the other hand, sensitivity tends to be higher in cases than controls and specificity lower. Estimates of misclassification bias may be either positive or negative and are consistent with random variation; there is little to suggest that bias is present in the studies analyzed.
Proton pump inhibitor use increases mortality and hepatic decompensation in liver cirrhosis
2019, World Journal of Gastroenterology
Hypothesis Testing for an Exposure–Disease Association in Case–Control Studies Under Nondifferential Exposure Misclassification in the Presence of Validation Data: Bayesian and Frequentist Adjustments
2016, Statistics in Biosciences
The effect of misclassification error on risk estimation in case-control studies
2015, Revista Brasileira de Epidemiologia

View all citing articles on Scopus

View full text

Original articleAssessment of exposure misclassification bias in case-control studies using validation data

Abstract

J Clin Epidemiol

J Clin Epidemiol

J Clin Epidemiol

Experimental quantiles of epidemiological indices in case-control studies with non-differential misclassification

Stat Med

Misclassification of exposure in case-control studies: Assessment by quality indices

Epidemiology

Original article
Assessment of exposure misclassification bias in case-control studies using validation data