Reliability of baseline self-reported information in the AGRICAN cohort

An important challenge in epidemiology is to ensure the reliability of collected data. Very few studies have been conducted in farming populations. We assessed the reliability of self-reported data on lifestyle, reproductive history, health and agricultural activities and tasks from the AGRICAN cohort.

Our analysis focused on 739 individuals from the 181,842 cohort members who completed the questionnaire twice between 2005 and 2007 with a median time interval of 452 days. Consistency in the responses to questionnaire items (lifestyle, health and agricultural activities including pesticide treatments) was assessed by the percentage of exact agreement (PA), Cohen's Kappa value (K) and the intraclass correlation coefficient (ICC).

Agreement was substantial to almost perfect for education, smoking, reproductive history and most health indicators (K/ICC > 0.61). Agreement was moderate for alcohol consumption and fair for diet. Agreement was substantial for animal and crop farming activities and tasks such as pesticide use on crops and protective equipment use (PA 81-99%, K/ICC 0.61-0.96). Most tasks showed moderate to substantial agreement, except a few with low agreement. Substantial to perfect agreement was observed for the duration of tasks, based on exact years of beginning and ending.

Farmers' answers appeared reliable for most occupational data, including data used to assess individual exposure to specific pesticides, and for most potential confounders.