CLN - Bench Matters

Harnessing the Power of Big Data Analytics to Achieve Reference Interval Harmonization in Clinical Laboratories

Bench Matters: June 2020

Mary Kathryn Bohn, BS, PhD candidate and Khosrow Adeli, PhD, FCACB, DABCC

As clinical laboratorians, we know that the test results and reference intervals (RIs) reported to patient medical records should reflect the preanalytical, analytical, and postanalytical conditions at any given laboratory. However, most clinicians and patients are not aware that test results can vary depending on the method used. Moreover, laboratories operating within the same healthcare system, and even using the same analytical platform, might report different RIs for a given test. This lack of harmonization significantly increases the risk of inappropriate as well as inconsistent test result interpretation, potentially leading to erroneous or missed diagnoses and unnecessary interventions. In recent years, more patients having direct access to their medical data combined with tighter integration of healthcare networks has underscored why harmonization matters. As laboratory professionals, we play a leading role in advocating for and achieving harmonized patient care.

Harmonization in laboratory medicine involves the total testing process, from collecting samples to reporting and interpreting results. However, past harmonization efforts have mainly been limited to achieving method standardization in the analytical phase of testing.

Gains From Assay Standardization

Result comparability across healthcare centers depends on standardizing laboratory measurements and tracing them to common reference materials. Without standardization, harmonization of RIs would not be feasible or appropriate. Recent method standardization efforts have been successful for many analytes, including cholesterol, creatinine, glucose, hemoglobin A1c, and sodium. For these standardized assays, RI harmonization across laboratories is very possible and arguably critical to clinical service. RI harmonization is also possible for nonstandardized assays that demonstrate good concordance between analytical platforms.

Unfortunately, despite these improvements, different RIs continue to be used across laboratories. The delay in developing and implementing harmonized RIs lies mostly in the challenge of recruiting a large representative healthy population to establish RIs. This critical gap limits our ability to deliver uniform laboratory service across healthcare networks and urgently needs to be addressed.

A Treasure Trove of Data

As challenging as the journey to RI harmonizations has been, the big data era in which we now find ourselves poses new opportunities to achieve harmonization in laboratory medicine.

The principle of using extensive laboratory datasets to assist in clinical service is not new, but novel applications for using this treasure trove of information keep unfolding as new software and statistical programs are developed. In the context of harmonization, outpatient data extracted from the laboratory information systems (LIS) of multiple clinical laboratories can be extremely useful in assessing inter-laboratory differences and establishing harmonized RIs. For example, data for a given assay and time period can be extracted from the LIS of several laboratories, reflecting their unique preanalytical, analytical, and population demographics. RIs can then be established based on outpatient data for each center as well as all centers combined (i.e. harmonized) and compared against each other to determine whether RI harmonization is feasible.

A New Method

Many approaches for establishing RIs based on outpatient data have been described in the literature. Older graphical models such as the Hoffman and Bhattacharya methods have often been questioned due to their inherent subjectivity, while newer methods such as the Arzideh method reported by the German Society for Clinical Chemistry and Laboratory Medicine show greater promise (Clin Chem Lab Med 2007;45:1043–57). Specifically, this new automated approach statistically isolates the healthy population in an outpatient dataset to derive accurate and robust RIs, demonstrating remarkable comparability to health-associated data. It is also uniquely suited to harmonization, eliminating the need to recruit a large healthy population and allowing for robust assessment of large datasets representative of multiple analytical platforms and geographic regions.

In our experience, we applied the Arzideh method to large outpatient datasets (up to 14 million results per test) extracted from community reference laboratory centers across Canada that use different analytical platforms for common laboratory tests (i.e. electrolytes, hepatic enzymes, and renal markers). When we applied this method to each center separately and to all centers combined, we observed only minimal differences in estimated reference limits. These findings highlight the use of big data in and the overall feasibility of harmonizing RIs in clinical laboratories for certain tests. We plan to expand this study to other analytes as well as verify established harmonized RIs using healthy adult samples.

As manufacturers’ platforms become more standardized and novel data-driven tools emerge, the prospect of RI harmonization in laboratory medicine appears much closer. We urge laboratories to collaborate and consider using big data analytic tools like the Arzideh method to assess the feasibility of RI harmonization in their regions. Of course, for some assays, different methods generate vastly different results and harmonization is not feasible. In these cases, method-specific RIs should be considered and implemented across centers that use the same analytical platform/method. Conversely, if methods are traceable to a common reference standard, the feasibility of RI harmonization is higher and should be considered a priority for laboratories globally. Only then will harmonized patient care at the level of clinical laboratories be possible and ultimately result in enhanced patient safety and a higher quality of healthcare for all.

Mary Kathryn Bohn, BS, is a PhD candidate with the CALIPER Project at The Hospital for Sick Children and the University of Toronto in Ontario, Canada. +Email: [email protected]

Khosrow Adeli, PhD, FCACB, DABCC, is head and professor of clinical biochemistry and pediatric laboratory medicine and senior scientist in molecular medicine at The Hospital for Sick Children and University of Toronto in Ontario, Canada. He also serves as president-elect of the International Federation of Clinical Chemistry and Laboratory Medicine. +Email: [email protected]