摘要：Objectives. We conducted a 3-year cohort study of 407 youths aged 9 to 18 years to develop multivariable risk prediction models of agriculture-related injuries. Methods. Data were obtained via participant event monitoring, with youths self-reporting injuries and exposures in daily diaries over a 13-week period. We evaluated data quality by comparing injury self-reports with other injury data. Results. Semilogarithmic plots of rates of all unintentional injuries combined (US data from 2000) as well as of agriculture-related injuries (US and Canadian data from 19 previous studies) graphed as a function of injury severity exhibited linearity, as did plots based on the present results. Severity-specific unintentional injury rates were 1.4- to 4.3-times higher than national rates, suggesting that our methodology can significantly reduce injury underreporting. In addition, at each severity level, estimated agriculture-related injury rates were 5.8- to 9.3-times higher than rates from previous national, regional, and state-based studies. Conclusions. Our approach to participant event monitoring can be implemented with youths aged 9 to 18 years and will yield reliable daily data on unintentional injuries. Unintentional injuries are a significant childhood health problem in the United States 1 , 2 ; for example, in 2001 such injuries accounted for almost 45% of all deaths in the population aged 1 to 19 years. 3 , 4 Unintentional injury is the leading cause of death in the 1- to 34-year age group 4 and has been for some time. Of the 30 million nonfatal injuries treated in US hospital emergency departments in 2001, 35% involved individuals aged younger than 20 years. 5 A significant methodological problem in injury epidemiology is collecting accurate data on events causing tissue damage as well as the tissue damage itself. 6 , 7 Other factors that complicate this process include the varieties of injury-producing events (e.g., slips, trips, falls), the multidimensionality of injury severity, and the question of what constitutes a reportable event. In studies of unintentional injuries, the reporting threshold has commonly been defined in terms of the nature of medical attention or the degree of interference with normal activities. Combined with the relative rarity of high-severity injuries, the ascertainment problem hampers accurate estimation of injury rates and identification of important risk indicators. It has been suggested that the reporting threshold be lowered, when appropriate, to include minor injuries, which would increase the number of injury events available for study, and that such data be collected on a weekly or even daily basis, which would shorten the recall period and thereby reduce underreporting and misclassification. 8 – 11 Some have argued that minor injuries may serve as proxies for more severe injuries. 7 We conducted a study to develop multivariable risk prediction models of work-related injuries among young people aged 9 to 18 years who were exposed to agricultural hazards. 12 We report on the quality of data obtained from youths who completed daily diaries during a 13-week reporting period.
