The objective of this study was to compare sick leave data obtained from questionnaires with data from company records. During a period of 12 months, questionnaires were completed monthly for 6 months and then at 9 and 12 months. The sensitivity and specificity of questionnaires for detecting an episode of sick leave were determined, using the company records as a reference standard. In addition, the duration of sick leave episodes reported in the two data sets was compared. In this analysis, company records were not assumed to be superior, and agreement was assessed with intraclass correlation coefficients (ICCs). The sensitivity of questionnaires for detecting an episode of sick leave was 55% (95% CI = 0.50-0.60) and the specificity 83% (95% CI = 0.72-0.94). The ICC for all episodes was 0.58 (95% CI = 0.47-0.67). The only satisfactory ICC (0.87; 95% CI = 0.74-0.93) was found for the questionnaires at 9 and 12 months. No large systematic differences were found between the duration of episodes reported in the two data sets. In conclusion, in our study, the sensitivity of questionnaires for detecting an episode of sick leave was very low. Furthermore, when episodes were recalled, there was little agreement on the duration of the episode between questionnaire data and data in the company records. Based on these results and considering the risk of missing questionnaires, data on sick leave gathered from company records are clearly preferable as an outcome measure in research.