Comparability of Information on Pesticide Use Obtained from Farmers and Their Proxy Respondents

A methodologic study in Iowa was conducted to compare information onagricultural factors obtained by interview from farmers and their proxy respondents.In-person interviews were sought from 100 farmer surrogate pairs and obtained from85 pairs. Respondents provided information on farm commodities produced, details onpesticide use, and pesticide application practices. The sensitivity and specificity ofsurrogate responses were over 90% for information such as years operating a farm oramount of corn or livestock production in recent years. Accuracy was lower forindividual pesticides used and details about pesticide application practices (sensitivitiesfrom 40 to 70% and specificities from 50 to 90%). Surrogate respondents were unableto provide as much information and reported lower prevalences of pesticide use thanthe farmers themselves. The correlation between prevalences of individual pesticidesreported by farmers and by surrogates was 0.91 for herbicides and 0.78 for livestockinsecticides. The loss of information from using proxy respondents would typically biasestimates of relative risk downward toward the null value and could result in falsenegative interpretations. Although use of surrogates undoubtedly introducesmisclassification of agricultural exposures, the degree of misclassification is generally inthe range observed for other factors obtained by interview in epidemiologic studiessuch as information on diet and details of tobacco and alcohol use.