Secondary databases in equine research

Knowledge on disease occurrence in the Swedish equine population is lacking. Secondary data (data not produced primarily for research) including medical information offer potential to investigate disease occurrence in populations without primary data collection. This thesis explored the potential use of two nation-wide secondary equine databases for research on diseases in the Swedish horse population. The data quality in one insurance database and one database from a national equine clinic network was evaluated. For diagnostic information, the agreement in insurance data was 84% whereas the completeness (proportion of problems in the clinical records recorded in the database) and correctness (proportion of recorded disease events in the database truly occurring) of clinic data was 91% and 92%, respectively. Logistic regression showed that agreement/correctness was significantly associated with type of visit (clinic data and veterinary care claims in insurance data), whether diagnostic codes were present in the clinical record and affected system (clinic data). To present disease occurrence in the respective study populations, disease was presented as incidence rates (insurance data) and proportional morbidities (both databases). For insurance data, the most commonly affected system was joints, followed by whole body, skin and digestive system. The most common specific diagnosis was fetlock inflammation. For clinic data 22% of all visits were health visits, and for problems visits, the most commonly affected body system was joints, followed by whole body, respiratory and skeleton system. For both databases, disease occurrence was highly related to demographic factors in the horse population. The data quality in both databases was found adequate for research purposes, with due consideration of variation in data quality among disease problems. Presentation of disease indices from the two databases provided useful information on disease occurrence in horses throughout Sweden. Importantly however, as many factors affect disease, results from other studies are not directly applicable to Sweden. Thus disease statistics need be obtained from the specific population of interest