Improving turnaround times with artificial intelligence in microbiology

This dual-center study evaluated the impact of artificial intelligence (AI) on urine culture turnaround times in Canadian diagnostic laboratories using microbiology laboratory automation. Data were collected before and after the implementation of PhenoMATRIX (PM), an AI-based software that provides continuous culture sorting and result interpretation support. In both a low-volume tertiary care hospital and a high-volume community laboratory, PM enabled earlier availability of interpretable results; however, reductions in time to result reporting (TTRR) were contingent on workflows for result release. At the tertiary care site, implementation of PM alone was associated with increased TTRR, reflecting delays between result availability and reporting. Integration of PM+ enabled automated release of negative results as they became available, resulting in a TTRR reduction of approximately 1.3 h. At the community laboratory, where PM+ was not implemented, a TTRR improvement of approximately 5.3 h was achieved by advancing manual screening workflows (08:00 vs 16:00), facilitating earlier review and release of PM-generated results. These findings indicate that AI-driven culture assessment reduces TTRR when coupled with processes that enable timely result release, either through automated reporting or optimized laboratory review workflows.IMPORTANCEAdvances in artificial intelligence (AI), coupled with the power of laboratory automation, is the next step in the evolution of the clinical microbiology laboratory. We demonstrated that bacterial culture assessment (urine cultures) by AI decreases the time required to release results to clinicians and reduces the hands-on time required by technologists to analyze and finalize laboratory analysis. This study was performed in two different laboratories that differed not only in geography but also in scope and scale. The same ultimate benefits were seen in both institutions demonstrating that this technology is widely applicable for most laboratories.