Clinical laboratories around the world see the potential to improve patient care by test standardization at the preanalytical, analytical, and post-analytical stages. Reference intervals (RIs) are a fundamental post-analytical tool used by healthcare professionals to interpret laboratory test results. Ideally, RIs represent the range of values present in a healthy, nondiseased population of the corresponding laboratory. Results that fall outside of the RI might be interpreted as abnormal and indicate the need for additional medical follow-up.
Unfortunately when reporting test rules, many labs continue to depend on outdated, inappropriately determined, and/or incomplete RIs. In addition, many patients visit more than one laboratory during their clinical follow-up for testing of the same analyte. Interpreting results for the same analyte processed by different labs can be confusing to clinicians and patients, leading to medical errors. The clinical community broadly recognizes these problems, thereby fueling interest in developing common RIs for use over large geographic areas, even when the included labs operate more than one type of analytical platform.
Baseline Diversity
In the western Canadian province of Alberta, clinical laboratory testing is offered at more than 110 sites, including two high-volume community laboratories, and many rapid response urban hospital laboratories and rural hospital laboratories. These labs perform testing on analytical platforms from several manufacturers, such as Roche Diagnostics, Beckman Coulter, Ortho Clinical Diagnostics, and Siemens Healthineers. A survey of currently employed RIs across Alberta revealed significant discrepancies in RI boundaries, as well as in age and sex partitions. This led our team of Alberta clinical biochemists to begin developing common RIs for frequently ordered laboratory tests.
The team employed a three-step process to harmonize RIs across multiple laboratories and platforms. First, we used an a posteriori indirect sampling approach to mine de-identified patient results available in the laboratory information systems of the various sites. The large number of tests performed in the province at the two high-volume laboratories (>40 million/year) enabled the team to develop the adult RIs solely from test values measured in community patients. When necessary, we developed pediatric partitions from a combination of community, inpatient, or outpatient results, with additional data filtering used when possible to better ensure data came from a nondiseased population. We then organized this data according to age and sex partitions determined by group consensus and supported by clinical relevance, literature, and, if necessary, additional data analysis. The data sets ranged from 3,400 to 25,000 individual test results, depending on partition and analyte.
Second, we determined the statistical RI through Bhattacharya analysis of the partitioned data. This statistical tool requires large data sets (N>3,000) and excludes outliers by sampling the central 95% of a normal distribution. In addition, Bhattacharya analysis can help determine if additional populations—such as unidentified age- or sex-specific populations or diseased populations—are within the employed data sets, thereby requiring additional partitions.
Our third step involved taking each Bhattacharya-derived RI and building the clinically relevant RI. We did so by assessing the clinical significance of the statistically determined RI and through group consensus. As part of this last step, we could create a small buffer within the RI to account for minor analytical differences between platforms and to increase clinical convenience for using the RI. We followed the philosophy of employing easy-to-remember numbers, if doing so would not compromise patient care.
Analyte-Specific Challenges
Despite our efforts, we could not fully harmonize RIs for two of the analytes reviewed: calcium and potassium. In the case of calcium, we found significant analytical bias between methodologies, and with potassium, we encountered preanalytical bias due to differences in sample type.
Clinical biochemistry laboratories in Alberta have now reached consensus on common RIs for 12 tests, including albumin, alkaline phosphatase, calcium, chloride, creatinine, glucose, phosphorous, potassium, sodium, total bilirubin, total carbon dioxide, and total protein. These common RIs will not only benefit the small rural hospitals that often have minimal clinical support but also provide significant improvement to the care of all patients within Alberta, who often have test results from multiple laboratories.
Allison A. Venner, PhD, FCACB, is a clinical biochemist and provincial point-of-care testing medical lead at Alberta Public Laboratories and a clinical assistant professor in the Department of Pathology and Lab Medicine at the University of Calgary. +Email: allison.venner@albertapubliclabs.ca
Jessica L. Gifford, PhD, FCACB, is a clinical biochemist at Alberta Public Laboratories and a clinical assistant professor in the Department of Pathology and Lab Medicine at the University of Calgary. +Email: jessica.gifford@albertapubliclabs.ca