Automation of plate inoculation and reading reduces process time in the clinical microbiology laboratory, compared to a manual workflow

We evaluated the benefits of automation for clinical microbiology laboratory workflow and compared the time required for culture inoculation and plate reading between manual processes and BD Kiestra (but Kiestra) automation. Kiestra automation consists of different modules, each of which facilitate different tasks and work together from specimen inoculation to results reporting. We tracked the steps and measured the associated times-to-completion for both manual and automated workflows, including number of plate touches, hands-on-time (HOT), total technologist time (TTT), and walk-away time (WAT). Manual and automated processes both included 90 samples, and a total of 180 agar plates. There were three media quantity protocols used to mimic common laboratory samples: 1) 30 urine specimens on one biplate; 2) 30 urine specimens divided across two plates; and 3) 30 blood or wound specimens divided across three plates; all cultures incubated for a minimum of 18 hours prior to reading or imaging. Automation reduced HOT by 85% and overall plate touches by 88%. Automated inoculation (through implementation of the BD InoqulA module) resulted in a 100% reduction in plate touches and created 81.5-minutes of WAT. Automation of culture reading (through implementation of the BD ReadA incubation/plate imaging module and BD Synapsys Informatics) reduced HOT by 53%. Overall, laboratory automation resulted in shorter TTT and created WAT when compared to manual processes. Automation can facilitate increased processing capacity, more efficient use of the labor force, and reduced time to results in the clinical microbiology laboratory.


INTRODUCTION 24
Infectious disease diagnosis often relies on the identification of bacterial pathogens in clinical 25 specimens, such as urine, blood, and wound. 1, 2 Traditionally, this involves manual inoculation of 26 clinical specimens onto culture plates, which are then incubated for a period of time (typically 27 ≥12 hours) to recover suspected bacterial pathogens. Plate reading to identify significant 28 bacterial growth has traditionally been accomplished by experienced medical laboratory 29 scientists. Even being a lengthy and labor-intensive process, the demand for clinical 30 microbiology laboratory services has increased in recent years and the need for laboratory 31 automation has become more urgent due to staffing challenges. [3][4][5][6][7][8][9] In addition, manual processes 32 are susceptible to specimen contamination, labeling error, and variability in culture quality. 10, 11 33 To address these and other common problems encountered in the clinical laboratory, automation 34 can be employed to streamline and standardize laboratory workflows. Several studies have 35 shown improved laboratory workflow efficiency, 3-5, 10, 12-16 result reproducibility, 4, 5, 13, 15-19 and 36 decreased incidence of errors 10, 13, 18-21 following the implementation of automation-related 37 processes. 38 39 incubation, plate sorting, and high-resolution plate imaging 24 by providing standardized digital 46 image acquisition through the BD Synapsys Informatics Solution (but Synapsys) (version 3.5). 47 Several studies have shown the benefits of implementing Kiestra automation systems in clinical 48 microbiology laboratories, including shortening workflow time and optimizing result quality. 13 This report describes a step-wise utilization of automated inoculation, incubation, and imaging 52 modules, in a simulated clinical laboratory environment, in order to streamline workflow 53 processes. Here, we recorded each workflow step (plate setup, inoculation, and reading) and 54 timing for plates that were processed manually, compared to the same number and types of plates 55 that were processed via Kiestra automated modules: InoqulA (BD product number: 446973) for 56 inoculation, ReadA (BD product number: 446948) for incubation and imaging, and Synapsys 57 (BD product number: 444158) for digital image reading. We selected six quantifiable parameters 58 to evaluate the efficiency for both manual and automated approaches: number of plate touches, 59 hands-on time (HOT), setup time, cleanup time, total technologist time (TTT), and walk-away 60 time (WAT). These metrics were chosen in order to determine whether automation can reduce 61 Manual and automated processes both used the proposed mimic specimens, therefore a total of 81 180 plates were testing in each method (manual and automated) for this comparison. The 82

MATERIALS AND METHODS 66
Chocolate plates were incubated in CO2 (5.0%) while the bi-plates, blood agar, and MacConkey 83 II plates were incubated in ambient air (ambient O2/N2) at 35-37°C. 84 85 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) based on the pre-programmed incubation protocol. The built-in ReadA camera (using embedded 117 Optis imaging technology) captured images of plates at predefined time points. The medical 118 laboratory scientist then reviewed and analyzed digital images and determined the growth 119 outcome from the culture plates using Synapsys software and determined whether to discard the 120 plates or to order further workup ( Figure 1C (Table 1). 142 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

143
This study compared the processing times between manual and automated processes for 90 144 specimens on 180 plates for each workflow process. Specimen inoculation using InoqulA 145 resulted in the automation of 17 steps and elimination of six steps, leading to a reduction in 146 number of plate touches from five per plate and one per specimen (with manual processes) to 147 zero ( Table 2). Once the technologist prepared the InoqulA, there was no more direct contact 148 with the plates during the inoculation process. The technologist next handled the agar plates 149 during their transfer from the InoqulA output stacker to the ReadA destacker. Moving inoculated 150 plates to the incubator is the same process and approximate timing with automated and manual 151 methods, so stack touches/moves were not counted for this step in the total plate touches for 152 either workflow. 153 154 For incubation and media plate reading, the ReadA system either automated or eliminated six 155 steps and resulted in two fewer touches per media plate (Table 3). Retrieving plates intended for 156 manual reading and navigating to the correct worklist in Synapsys are considered equivalent. 157 Therefore, setup and cleanup time differences for reading cultures via the manual method and 158 automation are negligible. 159 160 InoqulA workflow had longer setup times in this study, compared to the manual process, due to 161 inoculation being performed on two different days. This is attributed to the pre-production status 162 of the instrument and is not expected with the intended workflow of the instrument. The HOT 163 over this time period with the InoqulA workflow was shorter, resulting in a net time reduction 164 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 18, 2022. ; https://doi.org/10.1101/2022.03.16.22272483 doi: medRxiv preprint with automation. For all plates combined, total HOT and TTT for the manual process was 79.0 165 minutes and 96.0 minutes, respectively; for the automation process the times were 0 and 20.7 166 minutes, respectively (Table 4). In addition, the automated process yielded 81.5 minutes of WAT 167 that were not observed with the manual process. The TTT involving all plates, combined, was 168 42.0 minutes for the manual process and 22.5 minutes for the automated process (Table 4). 169 170 For a total of 180 plates, using automation for inoculation reduced TTT by 75.3 minutes, and 171 automation reduced reading time by 19.5 minutes. The TTT for the whole workflow was 138 172 minutes if using manual processes and 43.2 minutes by automation. The total differences 173 indicate that combined automated inoculation and reading processes reduced total plate touches 174 by 88%, largely due to the automation of inoculation steps (accounting for 990 of the 1,350 plate 175 touches saved) (Table 4). Automation reduced the total HOT by 85%, or 98.5 minutes. Although 176 there were slight increases in total setup and cleanup times (3.6 minutes combined), automation 177 reduced the TTT by 69%, or 94.8 minutes (75.3 minutes for inoculation and 19.5 minutes for 178 reading). Automation also generated 81.5 minutes of WAT, which was not present in the manual 179 method ( Figure 2). 180 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. and 20%, respectively) than those same times associated with the manual workflow due to a 195 technical event that required inoculation to occur over two days instead of one. Thus, setup and 196 cleanup occurred twice for the automation workflow in this study compared to manual workflow, 197 which only had one setup and cleanup period. This increase in time associated with the 198 automation workflow is largely artificial and is not expected to apply as part of routine use in the 199 laboratory. Plate touches during inoculation were reduced 100% (990 touches) with automation 200 (Table 4). This is consistent with previous studies that show a significant savings in labor time 201 and better colony recovery by automated sample preparation and plate inoculation in the 202 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 18, 2022. ; https://doi.org/10.1101/2022.03.16.22272483 doi: medRxiv preprint alone to automate the inoculation or incubation and imaging process, respectively, can 225 sufficiently save time and even generate WAT for a medical laboratory scientists to work on 226 other valuable tasks. Additionally, the modular design can provide options for selecting some 227 components and not others based on laboratory preference to adjust scalable automation. Overall, 228 modular Kiestra solutions can help position microbiology laboratories to achieve more accurate 229 and efficient testing. 230

231
Limitations 232 As the system was in a controlled setting, any hardware/software issues that arose were not 233 counted in "real-time;" therefore, any service down time as a result of mechanical or software 234 issues were not reflected in the results, here. Since the study was conducted in a non-blinded 235 fashion, it was not possible to eliminate outcomes bias associated with these results. 236 Additionally, the scope of this study did not include culture workup (such as identification of 237 AST testing) as part of the process and did not provide actual clinical microbiology results as the 238 endpoint comparison. Therefore, this study did not capture any potential differences in diagnostic 239 performance between automated versus manual processes. Real-world studies set in clinical 240 laboratories might benefit from an end-to-end approach in comparing manual and automated 241 processes. The reproducibility of results from this study will depend on the actual protocols used 242 in the real-world clinical microbiology laboratories and other variables including medical 243 laboratory scientist expertise involving skills that could increase or decrease manual process 244 times compared to those reported here. 245 246 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Conclusions 247
Compared to manual processes, Kiestra lab automation drives reductions in total process time 248 through automated inoculation and digital reading technology. These improvements allow for 249 increased capacity for specimen processing and staffing in microbiology laboratories. 250 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  shown. Setup and cleanup times associated with manual and automated processes were similar 381 for plate reading. Manual processes involved approximately twice as much HOT, compared to 382 automated processes, with not WAT for either workflow approach. The difference in TTT 383 between manual and automated processes largely reflect the difference in Hot between the two 384 approaches. (C) Total time associated with workflow processes for plate inoculation and reading, 385 combined, for manual and automated processes. In total, automation process required about one-386 third of TTT compared to manual processes. WAT generated nearly doubled that for TTT with 387 respect to automated processes whereas manual processes involved no WAT. 388 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)