Introduction Epidemiological studies on occupational exposure to pesticides commonly use self-reported questionnaire or interview data, so insight into recall accuracy is crucial to understand study findings.

Objectives To evaluate farmers’ and farmworkers’ recall of occupational exposure to pesticides and other exposure determinants, and to estimate the size of any recall bias.

Methods We used data from the IMPRESS project (www.impress-project.org), which includes three occupational cohorts of farmers’ and farmworkers’ exposure to pesticides in the UK and one in Uganda. Participants were surveyed at baseline to ascertain the frequency of their pesticide use, personal protective equipment (PPE) practices, and other information that may affect their exposure to pesticides; re-assessment occurred 2–14 years later, depending on the cohort. To assess recall, we examined the percentage of overall agreement, sensitivity, specificity, and any trends by demographic characteristics using regression analysis.

Results Across the four cohorts, 899 participants provided responses at two time-points. Preliminary analysis identified no overall trends in the recalled frequency of pesticide application in the Ugandan cohort, where reassessment occurred 2 years later. Initial results in the UK, where re-assessment took place many years later, suggested overestimation of years working with pesticides (Geometric Mean Ratio=1.25 [95% CI: 1.16 to 1.36]). The UK participants reported more frequent use of PPE items than in Uganda, which was mostly limited to long-trousers, shirt, and boots; however, recall ability was similar (mostly >70% agreement). The reporting of specific crops involving pesticide use was low among Ugandan farmers, who cultivated many crops (e.g., only 13% could recall all three major crops). Sensitivity of reported crop use in the UK appeared to be lower (<40%) with a longer recall period.

Conclusion Our results to date suggest limited evidence of recall bias, which appears to differ based on the specific exposure determinant and length of recall period.