Development and validation of an autoverification system (AVS) for tests studied at the Clinical Biochemistry Laboratory of Bezmialem Foundation University Faculty of Medicine Hospital

Introduction and Aim: Medical laboratories play a crucial role in diagnosis and treatment in modern medicine. The number of tests studied in medical laboratories has increased over the years and this situation has significantly augmented workload on laboratory specialist. Autoverifaction system (AVS) is a software system designed to approve test results based on predefined rules. The use of AVS can reduce workload on the laboratory specialist. In our study, we aimed to develop an AVS for selected tests studied in the Clinical Biochemistry Laboratory of Bezmialem Foundation University Faculty of Medicine Hospital and to validate AVS' performance. Methods: The tests selected for our study were selected from those frequently studied in our laboratory. AVS was implemented as a middleware system and integrated with the Laboratory Information Management System. AVS rules and algorithms were created and applied on the software system. The rules were initially validated on virtual patient data and it was shown that rules worked correctly. Next, they were validated using real patient test results. Finally, the agreement between AVS and manual verification was evaluated. Results: 724692 test results between November 1, 2023 and November 30, 2023 were evaluated in our study. Complete blood count panel tests accounted for the majority of these results, while coagulation parameters test results were the least frequent. Autoverification rates were between 55.6 - 94.6% for biochemistry tests, 39.6 - 95.5% for immunochemistry tests, 38.4 - 90.9% for coagulation parameter tests, and 80.2% and 49.6% for complete blood count and urine panel tests, respectively. For the comparison of manual approval and autoverification, the highest agreement was found to be between AVS and the assesments made by assistant doctor. Discussion: AVS algorithm and rules were created for the tests frequently studied in our laboratory. With AVS, we aimed to reduce workload, to detect possible errors early and to standardize the verification process of our laboratory. We believe that the autoverification rates we obtained in our study are sufficient for this initial phase. Further and multidisciplinary studies are requred to increase the performance of these systems.