The aim of this study was to evaluate the potential usefulness of the database maintained by the Swedish insurance company Agria for providing mortality statistics on Swedish horses. Mortality statistics (incidence rates and survival) were calculated, both crudely and stratified by sex, age, breed, breed group and diagnosis, for the horses with complete life insurance, which covers most health problems. The total mortality was 415 (95 per cent confidence interval [CI] 399 to 432) deaths per 10,000 horse-years at risk, and the diagnostic mortality, including only deaths with an assigned diagnosis, was 370 (95 per cent CI 355 to 386) deaths per 10,000 horse-years at risk. The diagnostic mortality of geldings was 459 (95 per cent CI 431 to 487), of mares 345 (95 per cent CI 322 to 365) and of stallions 214 (95 per cent CI 182 to 247) deaths per 10,000 horse-years at risk. The mortality rates increased with age and differed widely between breeds. Survival analysis showed that the median age at death of the horses enrolled before they were one year of age was 18·8 years. The most common cause of death or euthanasia was joint problems, which were responsible for 140 (95 per cent CI 130 to 149) deaths per 10,000 horse-years at risk. The results of multivariable models developed by using Poisson regression generally agreed well with the crude results.