Risk factors and sources of variation in horse falls in steeplechase racing in the UK.

We identified risk factors associated with falling during steeplechase racing. We used retrospective data from all steeplechase runs on UK racecourses during 1999: 10,866 starts with 647 horse falls. The relationship between continuous variables and falling was assessed using generalised additive models (GAMs). Polynomial fits then were included in a multilevel, multivariable logistic-regression model. The number of runners had a linear, positive association with the risk of falling. The distance of the race had a non-linear relationship with the risk of falling; the risk steadily increased in races up to 23 furlongs (1furlong approximately equals 198 m), and then decreased in longer races. Age also had a significant, non-linear relationship with the risk of falling: a decreasing risk up to 12 years of age followed by an increasing risk in older horses. Horses that wore visors and had raced previously were associated with a decrease in the risk of falling. Intra-class correlation coefficients (ICCs) showed that although most of the variation resided at the start (level 1), a proportion of variation in the risk of falling could be attributed to horse and race. Trainer and jockey contributed very little to the variation in the risk of falling.