Automation in clinical microbiology has become essential to address persistent staffing shortages and increasing specimen volumes. Automated technologies improve efficiency compared to manual methods across laboratories of varying size and patient acuity (1). Following implementation of automation, artificial intelligence (AI) algorithms offer the next step in optimizing workflows.

Copan’s PhenoMATRIX (PM) software (Copan Diagnostics, Murrieta, CA 92652, USA) uses AI algorithms to continuously assess and sort culture plates based on software interpretations of bacterial growth from WASPLab-generated digital plate images. For urine cultures, the software categorizes plates demonstrating no growth, no significant growth (<10 colonies/plate), mixed growth (>2 colony types), and/or specific colony colorations on chromogenic media. Prior studies have shown the accuracy of these algorithms across multiple specimen types and media (2–6). Beyond accuracy, AI has the potential to accelerate result reporting through continuous culture assessment.

This study evaluated time to result reporting (TTRR) using PM in a dual-center Canadian study of urine cultures. Additionally, PhenoMATRIX PLUS (PM+) expands on PM by enabling automatic result release based on lab-defined criteria without staff involvement, as evaluated at one site. This represents the first dual-site evaluation of PM AI-driven algorithms and their impact on TTRR in North America.

This study was conducted at two Canadian diagnostic laboratories with differing volumes, patient populations, and geographic settings. The first site, Queen Elizabeth II Health Sciences Centre (QEII), is a large tertiary care hospital located in Halifax and provides 24/7 diagnostic testing on behalf of the Nova Scotia Health Authority for other laboratories across the province. The laboratory also serves as the reference laboratory for the Canadian Atlantic region, processing over 65,000 urine specimens annually. The QEII is staffed with medical lab technologists from 7 a.m. to 11 p.m. and with a medical laboratory assistant from 11 p.m. to 7 a.m. The second site, Dynacare Laboratories in Brampton, Ontario, is a high-volume community laboratory operating 24/7 and processing an annual average of approximately 650,000 urine specimens collected across the province.

The study was conducted over a 3-year period from May 2022 to May 2025, with discrete data collection timepoints at each site before and after PM and PM+ implementation. At QEII, TTRR data were collected in June 2023 (PrePM), September 2024 (PostPM), and May 2025 (PostPM+). At Dynacare, data were collected in May 2022 (PrePM) and May 2023 (Post-PM).

For Dynacare, PrePM and PostPM TTRR were defined as the time from specimen placement on the WASPLab (Copan Diagnostics) to final approval and release of the culture report in the laboratory information system (LIS). At QEII, TTRR was calculated from specimen receipt in the laboratory to result release in the LIS. At both laboratories, TTRR included both organism identification (ID) and/or antimicrobial susceptibility test (AST) reporting when performed.

Midstream urine specimens were plated and incubated using WASPLab automation and imaged at approximately 16 h of incubation. At both Dynacare and QEII, organism identification was performed using Vitek MS, and antimicrobial susceptibility testing (AST) was conducted by MALDI-TOF using Vitek 2 MS (MS) in accordance with CLSI guidelines. At Dynacare, Escherichia coli exhibiting characteristic pink colonies on Oxoid Brilliance UTI Clarity Agar was reported without MS confirmation. QEII utilized chromogenic Orientation Agar for primary culture and MS for culture identification. E. coli exhibiting characteristic colonies on Orientation Agar were reported without MS confirmation.

Workflow changes occurred at Dynacare following PM implementation to facilitate review and release of PM-generated results, which were available earlier than the previously manually interpreted cultures PrePM. Urine culture screening expanded from evening-only hours (16:00–24:00) PrePM to combined day and evening shifts (08:00–24:00) PostPM. In addition, one full-time equivalent (FTE) was reallocated from the urine bench to another area following PM implementation due to efficiency gains. These changes were considered in the interpretation of study outcomes.

At QEII, urine cultures were screened on the day shift (07:00–16:00) for all TTRR timepoints measured (PrePM, PostPM, and PM+). Urine cultures with a final result of no growth or normal urogenital flora of less than 10 colonies were automatically released and reported via PM+.

This dual-site study demonstrates that artificial intelligence-assisted culture interpretation improves TTRR in automated microbiology laboratories; however, the magnitude of improvement depends on workflows that enable timely result release. While PM facilitated earlier availability of interpretable culture results at both sites, reductions in TTRR were only realized when paired with processes that supported earlier review or automated reporting.

At the tertiary care site (QEII), implementation of PM alone did not reduce TTRR, reflecting delays between result availability and manual release. In contrast, integration of PM+, which enabled automated release of negative and non-significant cultures, resulted in a measurable reduction in TTRR (~1.3 h) and a shift toward earlier reporting timepoints. This finding highlights that AI-driven interpretation improves efficiency primarily when coupled with mechanisms that translate earlier result availability into earlier reporting.

At the high-volume community laboratory (Dynacare), where PM+ was not implemented, a larger reduction in TTRR (~5.3 h) was observed following PM implementation. This improvement was not solely attributable to AI interpretation, but rather to concurrent workflow modifications that enabled earlier result review, including expansion of urine culture screening from evening-only to combined day and evening shifts. These changes facilitated earlier action on PM-generated results and underscore the importance of aligning laboratory workflows with AI-enabled data availability.

In addition to improvements in TTRR, implementation of PM at Dynacare was associated with measurable productivity gains, including the reallocation of one full-time equivalent (FTE) technologist from the urine bench. When normalized to test volume, this represents a reduction in technologist time per specimen, reflecting increased efficiency in culture screening and result interpretation. These findings support the role of AI not only in accelerating reporting but also in optimizing labor utilization in high-throughput environments. In Halifax, the implementation of TLA and PostPM+ allowed the reallocation of two FTE positions. One FTE savings (reallocation) was achieved through a reduction in labor requirements with the introduction of the TLA, and a second FTE was reallocated from the urine screening to another area of the laboratory.

These observations are consistent with prior studies demonstrating that algorithm-assisted culture interpretation (by Culbreath et al.) accelerates reporting and reduces the proportion of cultures requiring extended review (1). The present study extends these findings by demonstrating that the impact of AI on TTRR is not inherent to the technology itself but is contingent on the integration of AI outputs into laboratory workflows.

The ongoing shortage of qualified medical laboratory technologists and increasing specimen volumes underscore the importance of technologies that optimize efficiency without compromising quality. Full laboratory automation has already been shown to improve culture processing and mitigate staffing challenges (7–9). The addition of AI software such as PhenoMATRIX enables continuous assessment of culture plates and prioritization of review. However, as demonstrated in this study, the benefits of AI are maximized when paired with workflow strategies that enable timely release of results, either through automated reporting (PM+) or optimized manual review processes.

The combined benefits of continuous incubation, automated imaging, AI-assisted interpretation (PM), and automated result release (PM+) represent a major advancement in clinical microbiology. Smart incubation has been shown to shorten incubation times while maintaining accuracy and enabling faster identification and antimicrobial susceptibility testing results (10). Future innovations, such as AI-assisted selection of isolates for automated identification and susceptibility testing, could enable end-to-end automation of uncomplicated cultures. This evolution will allow laboratory professionals to focus their expertise on clinically complex specimens while maintaining high throughput and timely patient results.

This study represents, to our knowledge, the first dual-site evaluation in North America assessing the impact of PM AI-driven algorithms on urine culture TTRR. Across both laboratory settings, AI integration into automated microbiology workflows improved the timeliness of result availability, while reductions in TTRR were achieved when supported by workflow changes that enabled earlier reporting. These findings highlight that successful implementation of AI in clinical microbiology depends not only on algorithm performance but also on thoughtful integration into laboratory operations to realize its full clinical and operational benefits.

Ross Davidson, Email: ross.davidson@nshealth.ca.

Patricia J. Simner, Mayo Clinic, Rochester, Minnesota, USA