Nov 22, 2021
With the exponential growth of data-driven technologies, laboratories will likely face challenges when incorporating artificial intelligence (AI) and data integration into their enterprise. To implement AI and become part of a connected health care network, the laboratory community will need education on the technology and guidance on implementing appropriate changes at technical, content, and organizational levels.
AI and interconnected data: a trend in health care
With the growing value of laboratory medicine in informing clinical decisions, AI can improve diagnostics through more accurate detection of pathology, better laboratory workflows, and improved decision support, leading to accelerated productivity and ultimately enabling better patient outcomes. A recent study with key stakeholders in laboratory medicine showed that AI is currently used in 15.6 percent of organizations, while 66.4 percent felt they might use it in the future. Most had an unsure attitude about what they would need to adopt AI in the diagnostics space.
Uncovering the full potential of digital medicine and AI implementation requires an interconnected data infrastructure with accurate, readily available, and contextualized data. Digital health depends on interoperability—the ability to electronically share health-related data between two or more organizations. In many cases, rather than sophisticated analytics or complex AI algorithms, simply making the right information available to the right person at the right time can significantly improve patient care.
This article provides actionable tips on how to make the necessary technical, content, and organizational changes to successfully implement AI and interoperability in your clinical laboratory.
International standard terminologies enable unambiguous data exchange
"Uncovering the full potential of digital medicine and AI implementation requires an interconnected data infrastructure with accurate, readily available, and contextualized data."
Even within the same country, laboratory analyses usually have different names. For example, vitamin D is also called 25-OH-vitamin D, vitamin D3, calciol, calcidiol, 25-hydroxycholecalciferol, or cholecalciferol and is usually reported in at least two different units, ug/L and mmol/L. To harness the power of AI, laboratories must consistently adopt international standard terminologies and structure reports. This removes any confusion around analyte names and enables computers to easily identify, process, and exchange data, contributing to correct interpretation from all stakeholders. Code systems, such as Logical Observation Identifiers Names and Codes (LOINC), Nomenclature for Properties and Units (NPU), Unified Code for Units of Measure (UCUM), and SNOMED-CT, can facilitate the consistent application of terminology, units, and reporting across laboratory testing providers.
Traceability of measurements for comparable results
In laboratory medicine, the between-method variability is a source of uncertainty that can affect patient safety and clinical outcomes. Measurement traceability aims to reduce the between-method variability, making results comparable regardless of the principle of measurement, the method, the actual measurement procedure (test kit), and the laboratory that carried out the analysis. This is essential for sharing information in an interconnected health care system.
The International Vocabulary of Metrology defines traceability as a property of the measurement result or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons, all with stated uncertainties. Traceable measurement results allow for the establishment of generally accepted and usable reference intervals rather than method-specific reference ranges. ISO 15189 already asks laboratories to verify and document the traceability of measurements for their analysis. Standardizing laboratory analyses with traceable methods and reference materials produces an overall higher quality of results and enables health care professionals and computerized systems to transfer knowledge and interpret data from multiple laboratories more accurately.
Interpretive comments as a crucial part of electronic laboratory reports
Other than measurement results, interpretative comments are an essential part of clinical laboratory services. Analytical limitations (e.g., less than ideal sample conditions), the impact of clinical treatments, measurement uncertainty, or interferences inherent in an assay may affect the interpretation of a result. Although no electronic standard has been developed to store and transmit this information, interpretive comments are crucial to correctly interpreting laboratory results, whether by clinical professionals or AI, and should be integrated into the electronic report.
Moreover, most laboratories also include comments regarding critical values or other medical decision limits, such as diagnostic cut-offs. While not considered interpretative comments, they may help remind or inform the recipients of the report of such diagnostic cut-offs. However, these non-interpretive comments are not standardized and vary between laboratories. Since these cut-offs are widely used in laboratory reports, consensus standards and guidelines regarding both diagnostic cut-offs and the way data is structured and presented are needed.
Taken one step further, comments can summarize measurements and explain a patient's likely condition or point directly to a diagnosis, sometimes combined with recommendations for follow-up. These comments should be patient-focused and answer the question raised by the requesting clinician. With little or no clinical details typically available on the request form, individualizing this kind of comment is often difficult. With an interconnected data infrastructure, patient information—including insights from electronic health records, medical imaging systems, and other sources—becomes more accessible to laboratory staff, who can then tailor comments based on the clinical context and a requesting clinician’s needs.
Finally, in an interconnected health care system, administrative comments that communicate organizational directions and notes to the recipient of the laboratory report become of little importance as they are not associated with diagnostic decision making. Laboratories could consider using other technologies to communicate these messages, such as email or text messages.
Message-based information allows the decision support system to react appropriately
Ideally, electronic reports should organize results into smaller (and more strategic) messages. For example, a clinical decision support system warning of acute kidney injury may only use creatinine measurements and not the whole report. Such a function requires a report with information organized into single measurements, with any interpretive comments or uncertainty description attached to the respective measurement electronically. With all the relevant information, a decision support system can compile useful data and incorporate it into comprehensive reports to inform clinical decision making.
Get all stakeholders on board
Engaging different stakeholders and subject matter experts is essential to successful health care interoperability. When conducting external quality assessments, organizations should increasingly focus on incorporating international standards, whereas when conducting internal quality control programs, laboratories should consider the traceability of values assigned to calibrators and control materials.
"Engaging different stakeholders and subject matter experts is essential to successful health care interoperability. "
The appropriate use of laboratory monitoring and diagnostic testing is crucial for medical services, so laboratories need to maintain a close relationship with the treating physician. More than samples and measurement requests, laboratories should also receive clinical information. Easy electronic information exchange maintains high efficiency and enhances the value of the results.
For providing high-quality interpretative comments, laboratories need to maintain and develop the necessary expertise, including improving IT skills to extract and process the relevant clinical data and electronically attach interpretive comments. With fast and easy data sharing, clinicians and AI programs will have to cope with massive data traffic, increasing the risk of misinterpretation of laboratory results. This, in turn, increases the importance of interpretative comments to facilitate clinicians’ and AI-systems’ decision-making processes.
On the horizon
Challenges to implementing AI in the clinical laboratory include non-categorized, non-standardized, and incomplete data in electronic reports. Overcoming these challenges means putting prerequisites in place for exchanging, combining, and analyzing data in machine-readable formats. Besides improving patient care, this could enable AI developments, enhance public health surveillance processes, and accelerate scientific progress.