The agreement between workers and within workers in regard to occupational exposure to mercury in dental practice assessed from a questionnaire and an interview

We examined the concordance between workers by comparing answers to the questionnaire given by persons working in the same clinics during the same time spans and the agreement within workers by comparing answers to the same questions from a questionnaire and from an interview. Other aims were to see if there was a difference in the answers to the questionnaire across job titles and to study the impact of missing information on the response rate in a detailed questionnaire.There was a marked difference between the pairs of employees working in the same clinic regarding the start and termination years for the different preparation methods, and this was partly independent of their occupation. Kappa values for using different preparation methods in the questionnaire and at the interview varied between 0.41(moderate) to 0.88(very good).The results of this study indicate that a mailed questionnaire will cause misclassification of exposure. The observed occurrence of false positive exposure classifications from the questionnaire compared to the interview was higher than for false negative. This is important and may result in serious bias if the prevalence of exposure is low. Due to missing information, detailed questionnaires may also be inefficient if the goal is to construct exposure measures from combinations of several answers in the questionnaire.Exposure assessment based on questionnaires can introduce different biases involving both differential and nondifferential misclassification. Differential misclassification of the exposure occurs when the likelihood of being classified as exposed is dependent on the outcome, while nondifferential misclassification is a question of sensitivity and specificity of the tool used. In epidemiological studies, differential misclassification may bias an observed relative risk in both directions, while nondifferential misclassification most often biases the relative risk towards the null effect [1,2]. Detailed job-specific questionnai