NIOSHTIC-2 Publications Search

Assessing the effects of nondifferential misclassification of exposures in occupational studies.

Checkoway H; Savitz DA; Heyer NJ

Appl Occup Environ Hyg 1991 Jun; 6(6):528-533

https://doi.org/10.1080/1047322X.1991.10387923

00200889

A discussion was presented of the utility of indirect methods for assessing the potential extent of bias in effect estimates resulting from nondifferential exposure misclassification. Misclassification from retrospective assessment could occur for many reasons, including heterogeneous job groupings, erroneous exposure ratings, measurement instrument error, nonrepresentative exposure monitoring, incorrect imputation of missing values and failure to account for exposure controls. Indirect assessment of effects of misclassification bias was based on correction of observed effect estimates for a plausible range of misclassification rates. Misclassification of a primary study variable and of confounders could be evaluated. Nondifferential misclassification tended to obscure exposure response relationships and bias results toward the null. Hypothetical examples were presented illustrating misclassification for a simple dichotomous comparison and for one using exposure level categories. Equations were presented for correcting observed effect estimates based on cases and person years. Where relevant, time specific exposure misclassification rates were used to estimate corrected summary rate ratios for historical cohort studies. A hypothetical example of this was presented where misclassification rates decreased from 20% to zero over a 49 year time period. Similar approaches could be used to evaluate misclassification of a confounder, such as another occupational exposure, that was highly correlated with the main exposure. The authors conclude that an indirect approach is preferable to a direct approach for evaluating effects of exposure and confounder misclassification because of greater feasibility and nondependence on extrapolated data. Calculating a range of corrected values is recommended.

NIOSH-Publication; NIOSH-Contract; Contract-91-38661; Occupational-health; Epidemiology; Exposure-levels; Occupational-exposure; Analytical-methods; Risk-analysis; Mathematical-models

AOEHE9

19910601

Journal Article

Contract

1991

Contract-91-38661

6

1047-322X

Applied Occupational and Environmental Hygiene