I spoke at an IFCC live webinar to explain our white paper on machine learing. Here are my slides. Abstract: During the webinar, my focus has been on how to evaluate machine learing models as a specialist in Laboratory Medicine. Interpretability methods are key to compare domain knowledge with the model’s inner workings. Standardization of analyses with tracability to reference methods is required for greater tranferability of machine learning models. References: Master, SR. et al.; ClinChem (2023); https://doi.org/10.1093/clinchem/hvad055 Slides How to evaluate machine learning models in Laboratory Medicine?
Covid has led to Euromedlab 2021 in Munich being rescheduled to 2022. These are my slides for a talk on PBRTQC. Abstract Patient-based real-time quality control (PBRTQC) has proven to detect clinically relevant measurement errors that were missed by traditional quality control procedures. However, PBRTQC is less well established. Computer simulations have uncovered the basic properties of PBRTQC and facilitate implementation into routine practice. PBRTQC works best for analyses with a large number of daily measurements and a small spread of results. Seasonal (or other extra-analytical) variations impede error detection. Winsorization of outlying measurements and Box-Cox transformation often lead to better performance.
In an increasingly interconnected health care system, laboratory medicine can facilitate diagnosis and treatment of patients effectively. This article describes necessary changes and points to potential challenges on a technical, content, and organizational level. As a technical precondition, electronic laboratory reports have to become machine-readable and interpretable. Terminologies 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 lead to the necessary semantic interoperability. Even if only single “atomized” results of the whole report are extracted, the necessary information for correct interpretation must be available. Therefore, interpretive comments, e.
Background: Patient-based real-time quality control (PBRTQC) avoids limitations of traditional quality control methods based on the measurement of stabilized control samples. However, PBRTQC needs to be adapted to the individual laboratories with parameters such as algorithm, truncation, block size, and control limit. Methods: In a computer simulation, biases were added to real patient results of 10 analytes with diverse properties. Different PBRTQC methods were assessed on their ability to detect these biases early. Results: The simulation based on 460 000 historical patient measurements for each analyte revealed several recommendations for PBRTQC. Control limit calculation with “percentiles of daily extremes” led to effective limits and allowed specification of the percentage of days with false alarms.
Background Terminologies facilitate data exchange and enable laboratories to assist in patient care even if complex treatment pathways involve multiple stakeholders. This paper examines the three common terminologies Nomenclature for Properties and Units (NPU), Logical Observation Identifiers Names and Codes (LOINC), and SNOMED Clinical Terms (SNOMED CT). Methods The potential of each terminology to encode five exemplary laboratory results is assessed. The terminologies are evaluated according to scope, correctness, formal representations, and ease of use. Results NPU is based on metrological concepts with strict rules regarding the coding of the measurand and the result value. Clinically equivalent results are regularly mapped to the same code but there is little support to differentiate results from non-standardized measurements.