Data analytics for error detection in clinical laboratories

Errors inevitably occur in the practice of laboratory medicine. A cornerstone of clinical laboratory quality management is the detection of erroneous results and the assessment of imprecision, bias, and other performance limitations of clinical test methods, particularly those affecting patient care. Errors can arise in each of what has been conventionally regarded as the three key phases of testing: pre-analytical, analytical, and post-analytical. In this review, both the standard concepts and methods of quantifying uncertainty and error are introduced in the context of clinical laboratory operations. Method validation and verification studies are presented as opportunities for preemptive and anticipatory error assessment-before tests are implemented for patient testing. Quality control monitoring is a key internal quality assurance strategy, whereas proficiency testing forms the basis of most external quality assurance initiatives. Data analytic approaches for error detection are reviewed, highlighting quantitative and statistical concepts on which they are based, and emerging machine learning and artificial intelligence algorithms are presented as contemporary tools currently under development for error detection in the clinical laboratory.