It is hard to imagine the sheer impact of recent budget cuts, reduced reimbursements, and severe staff shortages within the healthcare sector, especially when medical institutions are responsible for over 70% of decisions regarding a patient’s diagnosis, treatment, and discharge [1, 2]. Unfortunately, this is exactly the challenge that laboratory medicine, which encompasses clinical microbiology, has faced for the last couple of decades. In a traditional clinical microbiology laboratory, critical work consists of identifying pathogens from patient specimens and testing their reactions to a slew of antibiotics [3]. With these new constraints, many microbiology labs look to total laboratory automation (TLA) to sustain their high output levels. TLA and its associated technologies are not without their challenges, however, especially during the early implementation of these systems.
For many decades, clinical microbiology lagged in the transition to automated systems, instead opting to carry out the majority of its work manually—handling specimens, inoculating plates, incubating and reading results [4]. However, within the last 10 years, the development of new technologies has enabled the transition to increased use of automation. Today, total laboratory automation (TLA) workflows are the norm in clinical microbiology labs, automating the process from the moment when specimens are received to the moment results are reported to information systems [4]. Without a doubt, automation has been fruitful for the labs that have made this transition, with reported benefits including an increased number of specimen analyses, increased efficiency, lowered costs, lowered specimen turnaround times, and more accurate readings [2]. Another facet of automation has been the incorporation of machine learning (ML) systems. ML differs from traditional expert systems in that, rather than manually pre-programming a set number of rules, the system can synthesize its own rules simply using the provided data. For example, a traditional system might be programmed to recognize rod-shaped and round-shaped bacteria to distinguish between bacillus and cocci bacteria; an ML system would only need to be given microscopic images that are labeled as either bacillus or cocci to develop an algorithm or rule [2].
Despite being overwhelmingly supported by the field, total laboratory automation, especially ML, comes with its unique challenges. One of the more prominent limitations of ML is the transparency and interpretability of the intrinsic mechanisms behind these systems [2]. Often referred to as a “black box”, everything that goes on behind the scenes in one of these systems from input to output is unknown to the user, a microbiology technologist, or a healthcare professional [3]. Without the ability to investigate and validate the findings, or even understand the system's mechanism, these individuals may be hesitant to take on ML in their respective workspaces or laboratories [5]. Another challenge that comes with the rise of total laboratory automation is the concern for patient privacy and the implications of data sharing. With a greater need for data communication between personnel, satellite and central laboratories, and medical institutions, strong regulations surrounding data reporting and use of patient data are needed [3]. One final challenge that comes with TLA and ML is not with the systems themselves, but rather with the costs of their implementation. The high initial investment cost to incorporate these automated systems into a laboratory is often discouraging for stakeholders in laboratory medicine [6]. Without insight into the long-term benefits of cost reduction and laboratory workflow efficiency, these stakeholders likely will remain hesitant to take on automation.
Overall, the automation of systems and devices in clinical microbiology settings has revealed surprising successes and the potential to drastically improve clinical microbiology processes. With this sector of healthcare playing such a critical role in patient health, TLA must become the norm across the board. Nevertheless, it is important to appropriately address the challenges that come with automation, many of which discourage clinical microbiology stakeholders from making this important transition in their laboratories. With further research and increasing clinical applications of automated technology, current gaps in knowledge regarding how these systems work and the benefit that they bring to the table will likely be resolved.
References:
Badrick T. (2013). Evidence-based laboratory medicine. The Clinical Biochemist. Reviews, 34(2), 43–46.
Culbreath, K., Piwonka, H., Korver, J., & Noorbakhsh, M. (2021). Benefits Derived from Full Laboratory Automation in Microbiology: a Tale of Four Laboratories. Journal of clinical microbiology, 59(3), e01969-20. https://doi.org/10.1128/JCM.01969-20
Leo, S., Cherkaoui, A., Renzi, G., & Schrenzel, J. (2020). Mini Review: Clinical Routine Microbiology in the Era of Automation and Digital Health. Frontiers in Cellular and Infection Microbiology, 10, 582028. https://doi.org/10.3389/fcimb.2020.582028
Zimmermann S. (2021). Laboratory Automation in the Microbiology Laboratory: an Ongoing Journey, Not a Tale?. Journal of Clinical Microbiology, 59(3), e02592-20. https://doi.org/10.1128/JCM.02592-20
Peiffer-Smadja, N., et al.. (2020). Machine learning in the clinical microbiology laboratory: has the time come for routine practice?. Clinical Microbiology and Infection : The Official Publication of the European Society of Clinical Microbiology and Infectious Diseases, 26(10), 1300–1309. https://doi.org/10.1016/j.cmi.2020.02.006
Ketan Paranjape, et al., The Value of Artificial Intelligence in Laboratory Medicine: Current Opinions and Barriers to Implementation, American Journal of Clinical Pathology, Volume 155, Issue 6, June 2021, Pages 823–831, https://doi.org/10.1093/ajcp/aqaa170