Validation of computerized Swedish horse insurance data against veterinary clinical records.

The study compared data from a large Swedish horse insurance database and veterinary clinical records to evaluate their agreement. The findings revealed a high level of concordance for demographic information, diagnostic information, and system information, with respective agreement percentages of over 94%, 84%, and 92%.

Detailed Explanation

This research involved the comparison of demographic, diagnosis and system information between a large insurance database and clinical records (CR) concerning horses in Sweden. The study sample consisted of 400 veterinary care and 140 life claims.

• Demographic data: The researchers found that there were cases where the CR had missing demographic data, from 2% missing horse names to 16% missing breed information. Excluding missing data, there was an over 94% agreement rate for demographic information.
• Diagnosis information: For veterinary care and life claims, the agreement rate for diagnoses was 85% and 83% respectively. Minor disagreements were 6% and 5%, and major disagreements were 9% and 12% respectively.

The researchers considered CR data as the gold standard for comparison. For the diagnostic systems evaluated, the sensitivity – the ability of the data to correctly identify positive instances – varied from 62% (skin conditions) to 89% (digestive conditions). However, specificity – the ability to correctly identify negative instances – was more than 96% for all systems. The positive predictive values, indicating the probability that subjects with a positive screening test truly have the condition, ranged from 86% (skin) to 97% (digestive).

• Factors affecting agreement: The research used logistic regression analysis to study factors associated with agreement for diagnosis. This included factors such as the type of visit (clinic/field), treating veterinarian or clinic, whether the clinical record was computerized or manual, processing clerk, status of claim (rejected or reimbursed), system diagnosis, and the presence of settlement or death certificate in the paper file. The results indicated that for veterinary care claims, the type of visit significantly affected agreement with clinic visits resulting in better agreement than field visits.

In conclusion, this research confirms the general reliability of insurance data, while highlighting areas that could potentially benefit from improvement, specifically concerning the quality of recorded information and understanding of factors that might affect data accuracy.