Canine epilepsy is one of the most common neurological conditions in dogs but the actual incidence of the disease remains unknown. A Swedish animal insurance database has previously been shown useful for the study of disease occurrence in companion animals. The dogs insured by this company represent a unique population for epidemiological studies, because they are representative of the general dog population in Sweden and are followed throughout their life allowing studies of disease incidence to be performed. The database covers 50% of all insured dogs (in the year 2012) which represents 40% of the national dog population. Most commonly, dogs are covered by both veterinary care insurance and life insurance. Previous studies have shown that the general data quality is good, but the validity of a specific diagnosis should be examined carefully before using the database for incidence calculations. The aim of the current study was therefore to validate the information contained in the insurance database regarding canine epilepsy. The validation focused on the positive predictive value and the data-transfer from the veterinary practice records to the insurance database. The positive predictive value was defined as the proportion of recorded cases that actually had the disease in question. The quality of the data-transfer was assessed by comparing the diagnostic codes in practice records to the codes in the insurance database. The positive predictive value of the diagnostic codes for canine epilepsy (combining "epileptic convulsions" and "idiopathic epilepsy") in the insurance database was validated in a cross-sectional study where insurance claims for canine epilepsy were compared to diagnostic information in practice records. A random sample of dogs with a reimbursed insurance claim during 2006 was included in the study sample (n=235). Practice records were requested by mail from attending veterinarians. Two independent examiners scrutinized all the records. All 235 dogs were coded correctly in the database as they really had suffered seizures with or without convulsions, and the quality of the data-transfer was therefore excellent. In total, 167 dogs (71%) were classified as cases of canine epilepsy according to pre-defined criteria, and the positive predictive value was therefore considered relatively high. Based on these results, it was concluded that the data regarding canine epilepsy in the insurance database can be used for further population studies.