Diving into Data Science: A Clinical Laboratory Update

Résumé fourni par la source

“Data is like garbage. You’d better know what you are going to do with it before you collect it.” — Mark Twain Mark Twain’s words are perhaps controversial; it is undeniable, however, that data, and what one does with it, has fundamentally changed science and medicine. Clinical laboratories are no exception, and in fact have had a leading role in developing health data sources as well as health data science and analytics. At many laboratories and hospitals, laboratory information systems (LIS) predated robust electronic health record (EHR) systems by years or even decades, establishing them as the first “data warehouses” and bringing with them robust computational methods for quality control, reference range estimation, and clinical decision support. Moreover, the analytical roots of clinical chemistry and laboratory medicine predate even the earliest LIS, emphasizing the analytical mindset deeply rooted in laboratories and clinical pathology. Nonetheless, laboratory data science today is at an inflection point. The size and depth of data available to laboratories, as well as the addition of new data-intensive testing modalities such as genome sequencing or mass spectrometry, is pushing the limits of data storage and the analytic capabilities of laboratories. In parallel, the breakneck advancement of artificial intelligence and machine learning (AI/ML) technologies is creating new opportunities for laboratory data analysis. Finally, as laboratories seek to incorporate laboratory testing and interpretation as an integral element of diagnostic odysseys, integration within the broader healthcare data science field and operationalization of new analytic approaches presents interoperability and regulatory challenges. In this special issue, several articles focus on data pipelines, laboratory-based ML, and quality assurance, including an article by Ammer and colleagues highlighting the use of a R-based package, and a review by Spies and colleagues surveying data-driven anomaly detection methods. The infrastructure needed for data science in the laboratory is addressed in several articles and reviews by Cotten, Forsman, Krumm, McClintock, and Mooney. Future challenges for the field are also highlighted, including opinions and reflections on data science education (Kadauke), reproducibility (Mathias and Master), and interoperability (Chang), and a review of fair and equitable AI/ML practices (Zaydman). Taken as a whole, this special issue highlights many remarkable recent advances of laboratory data science; however, it would be remiss of the editors to not recognize 4 challenges the field faces: The operationalization of data pipelines and algorithms in clinical environments still faces substantial barriers, including: access to, and standardization of, data sources, the ability to leverage necessary technologies (e.g., cloud environments, workflow orchestration, open-source software), improved connectivity within informatics systems, and training of people with the right skills to develop, evaluate, and maintain these applications in production within healthcare/lab IT infrastructure. Second, the field must continue to emphasize the importance of “reproducibility” of data analyses. Such efforts will reduce errors, improve the success and safety of data analyses when operationalized, and support increased regulation of software and algorithms when needed. The field must place continued emphasis on data diversity, equity, and inclusiveness. Increasingly, it is recognized that population-specific data (e.g., reference intervals) improve patient care; conversely, we must be mindful of the (mis-)use of AI/ML methods that do not account for population-specific differences, or worse, actively misrepresent racial, ethnic, or gender-based features. Finally, the field must prioritize education and training of data science and related skills, across all positions and roles within our laboratories. We believe that the methods and tooling for data science, data reproducibility principles, and the fundamentals of AI/ML methods should be taught alongside other “core” skills for laboratory medicine trainees, medical technologists, and other members of the laboratory. We hope you enjoy this special issue of JALM. Author Contributions:All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. Authors’ Disclosures or Potential Conflicts of Interest:Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:Employment or Leadership: N. Krumm, L.A.L. Bazydlo, D.R. Bunch, S. Haymond, and D.T. Holmes, guest editors, The Journal of Applied Laboratory Medicine, AACC. D.T. Holmes, AACC and MSACL; S. Haymond, AACC. Consultant or Advisory Role: None declared. Stock Ownership: None declared. Honoraria: D.R. Bunch, AACC and MSACL; D.T. Holmes, AACC and MSACL; S. Haymond, AACC and Korean Society Laboratory Medicine. Research Funding: D.T. Holmes, SCIEX—loaned instrumentation. Expert Testimony: None declared. Patents: None declared. Other Remuneration: D.T. Holmes, support for attending meetings and/or travel from AACC and MSACL; S. Haymond, support for attending meetings and/or travel from AACC, Korean Society Laboratory Medicine, and MSACL.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.