Error management in large data sets: A critical component of data collection in veterinary medicine

Abstract

Listen

Translate article

ObjectiveTo assess the usefulness of a webpage incorporating a real-time error management system for the Confidential Enquiry into Perioperative Equine Fatalities 4 (CEPEF4). Study designAn observational, descriptive, retrospective, longitudinal study. AnimalsData sets from 48,075 general anaesthetics and 12,453 standing sedations. MethodsComparison of three consecutive phases of data collection [phase I: personal document file (PDF) by e-mail, manual revision; phase II: PDF by e-mail, automatic revision; phase III: web system, automatic correction] objectively (number of e-mails sent and received by the data managers related to data correction; number of detected errors by phase) and subjectively (16 questions in Likert format; one open-ended question). ResultsObjectively, the average number of e-mails sent per day was reduced from phase I (3.6) and II (3.9) to 1.7 in phase III. The ratio of e-mails per 100 cases received was reduced from phase I (5.8) and II (4.0) to 2.4 in phase III. Errors remained below 2000 during phase I, reaching a peak of 5430 in phase 2 owing to a strengthened error detection strategy. During phase III, errors were reduced from 3200 to 423 within 94 days. Subjectively, questions evaluating the PDF system indicated that users overall agree/strongly agree (54–84%) with its comfort level versus the (89–99%) of the web system. When compared, users found the web system more comfortable than the PDF system (86–95%). Conclusions and clinical relevanceThe transition to a real-time error management system in phase III led to considerably fewer e-mails sent and received plus fewer errors. Users also found the website preferable to the PDF system. Our findings highlight the importance of implementing an error management system to minimize data inaccuracies and to improve the overall efficiency and ease of use of large data set collection in veterinary medicine.