Riders’ prediction of results at Fédération Equestre Internationale (FEI) endurance rides and sources of bias in questionnaires completed by riders

Abstract

Information acquired from endurance riders and its relationship with the results of the ride has not been investigated. The aims of this study were to assess associations between data provided by riders and data obtained from the Fédération Equestre Internationale (FEI) website at FEI endurance rides and to identify whether data provided by riders in pre- and post-ride questionnaires was subject to response and/or information bias. Variables were collected from the FEI website and from self-completed pre-ride and post-ride questionnaires at 20 FEI endurance rides in 2011 and 2012. Kappa statistics and Mann–Whitney U-tests were used to assess the relationship between FEI website and questionnaire data and between the riders’ predictions and experiences. Univariable logistic regression was used to investigate association between completion of the ride and riders’ predictions. Response bias was assessed using multi-level logistic regression models. At least one questionnaire was completed for 236 (18.8%) of 1254 competition starts included in the study. There was excellent agreement on signalment between questionnaire and FEI data, except for data on breed. There was moderate agreement between the mean speed category predicted by riders and reported in the official results. Riders with the aim of qualification or ‘competition, achieving the best possible results’ were less likely to complete the ride than riders aiming for ‘training’. Female riders, riders competing in a young rider class, older riders, riders competing in rides of shorter distances and riders that had completed the distance of the study ride more than three times were associated with increased likelihood of completing the questionnaires. In conclusion, some riders’ predictions were associated with the outcome of the ride. A larger study is needed to assess these variables as risk factors for eliminations. The response biases and the difficulties of data collection identified in this study can help when designing future studies.