Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Minimizing Bias in a Case-Control Study of Farm Injury

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  Journal of Agricultural Safety and Health. 11(2): 175-184. (doi: 10.13031/2013.18184) @2005
Authors:   L. Day, J. Langley, D. Voaklander, M. Sim, R. Wolfe, J. Dosman, L. Hagel, J. Ozanne-Smith
Keywords:   Agricultural injuries, Bias, Case control, Recruitment, Risk factors

We report on our strategies to minimize bias in the FIRM study, a prospective case-control study of risk factors for serious farmwork-related injury. The study base is adult males working on farms in the catchment regions of 14 larger regional hospitals in one Australian state. Cases are identified on presentation to the emergency departments, while age-matched controls are recruited via random telephone survey. Eligibility criteria for cases include a maximum abbreviated injury severity score of at least 2, to minimize the potential for selection bias against those with less severe injuries treated outside the hospital system. An audit at one hospital showed that 93% of eligible patients identified in the electronic surveillance system had been approached regarding participation. Results to date show that 38% of those approached decline to have their contact details made available to researchers. Those who decline are asked to complete two key questions to enable comparison with those who participate. Control recruitment relies on telephoning regional households until an individual from the study base, satisfying the matching criteria, is identified. This process minimizes the potential for selecting against farm workers who may live off-farm. Ninety-four percent of age-matched eligible controls have participated to date. We are testing a dynamic pool of individuals identified as study base members but not matched on the first call to determine its effect on the probabilities of selection. Our strategies appear to be minimizing detection, selection, and response bias, thereby enhancing the validity of the study results.

(Download PDF)    (Export to EndNotes)