An Assessment of the Quality of Practice-Generated Computerized Medical Records as a Data Source For Research Into Pregnancy Loss in Dairy Cows

Abstract

This paper discusses the suitability of medical records as a source of data for research into the problem of pregnancy loss in dairy cows. The process of converting the computerized veterinary practice records into analyzable data is described. Various measures of data quality are discussed including the percent of missing, incorrect and outlier values in the final data set. Finally, descriptive statistics are calculated to demonstrate the similarity between the population of dairy cows described by the medical records and the target population of commercial dairy cows. This paper demonstrates that the positive aspects of practice-generated computerized medical records (volume of data, similarity to the target population) compensate for the disadvantages of poor case definition and missing data. The analysis of these data to investigate pregnancy loss in dairy cattle is currently under way and will be reported elsewhere.