Development and performance evaluation of clinical decision support system algorithms for clinical biochemistry laboratory

Manual verification of clinical chemistry laboratory analysis results is time-consuming and tedious, and varies among laboratory staff. The aim of this study was to create a set of autoverification (AV) rules for complete blood count, clinical chemistry and immunochemistry analyses and to evaluate the performance of the rules. The study included 69 tests belonging to complete blood count, clinical chemistry and immunochemistry panels. After the AV rules were applied, the AV behaviour of the system was compared with expert users before the clinical decision support system (CDSS) was activated. After activation of the CDSS, the change in laboratory turn around time (TAT) was evaluated and AV pass rates were monitored. With the prepared algorithms, autovalidation rates of 79.5% and 74.9% for clinical chemistry and immunochemistry on a test basis, and 24.5% and 57% on a sample basis, respectively, were obtained. Since leukocyte subgroups in the complete blood countrequired holistic evaluation, they were subject to approval on a tube basis and AV was limited to 17.3% on a sample basis. When the AV behaviour of the system we set up with two expert users was compared, no autoverified results were found among the results that were not directly approved by the experts. For the autoverified tests, a decrease of 4-9% in TAT was observed after the system was fully activated, despite an increase of approximately 10% in the average laboratory test volume. In our laboratory, the implementation of the AV system accelerated the test validation process and therefore the TSS, despite the increase in the volume of tests finalised during the setup process. With the reduction of manual procedures, more time was saved, allowing the verification of critical results and increased test requests to be responded.