Abstract Developmental Orthopaedic Disease (DOD) in limb joints is frequent, can cause lameness and is a major cause of economic losses for the horse breeding industry. Studies on risk factors for DOD usually dichotomise the outcome to presence/absence without considering the extent of the disease, which can be appreciated by the number or the severity of the lesions. The aim of this study was to assess the effects of growth, exercise conditions and feeding practices on the number of radiographic findings (RF) of DOD in a cohort of horses, assuming that the risk factors associated with the presence of the disease and the number of RF may differ. We recruited broodmares on 21 stud farms in the Normandy region, France in 2002–2004. The stud farms were visited regularly to collect growth, exercise conditions, and feeding practices data from the end of the gestation to the weaning of the foal. The carpus, the front and hind digits, the tarsus and the stifle of the foals were radiographed at approximately six months of age. Each foal was classified according to the absence (no RF) or presence (>or=1 RF) of DOD and to the number of RF. A zero-inflated Poisson model was used to simultaneously determine risk factors for presence/absence and for the number of RF of the disease. The convenience sample consisted of 378 foals, 53.1% of which did not have any RF. The mean number of RF per foal was 1.25 and the variance was 2.80. Comparing to the Poisson, negative binomial, and zero-inflated negative binomial regressions, the zero-inflated Poisson model was found to be best suited for these data. Of the 295 foals with complete data, three risk factors were significantly associated with a higher risk of being affected by DOD: high wither height increase, large group size in pasture and large pasture size. Seven risk factors were associated with the number of RF: wither height at 30 days of age, age of the mare, breed, regularity of exercise, Ca/P level in the mare and foal rations, group size in pasture, and the type and frequency of handling. This study shows that risk factors vary for the presence and extent of DOD, and that the use of a count model achieves a more detailed determination of risk factors for the presence of DOD and for the number of RF of DOD compared to the more conventional dichotomy presence/absence of the disease.