Explainable Machine Learning Models for Rapid Risk Stratification in the Emergency Department: A Multicenter Study.

BACKGROUND: Risk stratification of patients presenting to the emergency department (ED) is important for appropriate triage. Diagnostic laboratory tests are an essential part of the workup and risk stratification of these patients. Using machine learning, the prognostic power and clinical value of these tests can be amplified greatly. In this study, we applied machine learning to develop an accurate and explainable clinical decision support tool model that predicts the likelihood of 31-day mortality in ED patients (the RISKINDEX). This tool was developed and evaluated in four Dutch hospitals. METHODS: Machine learning models included patient characteristics and available laboratory data collected within the first 2 h after ED presentation, and were trained using 5 years of data from consecutive ED patients from the Maastricht University Medical Center (Maastricht), Meander Medical Center (Amersfoort), and Zuyderland Medical Center (Sittard and Heerlen). A sixth year of data was used to evaluate the models using area under the receiver-operating-characteristic curve (AUROC) and calibration curves. The Shapley additive explanations (SHAP) algorithm was used to obtain explainable machine learning models. RESULTS: The present study included 266 327 patients with 7.1 million laboratory results available. Models show high diagnostic performance with AUROCs of 0.94, 0.98, 0.88, and 0.90 for Maastricht, Amersfoort, Sittard and Heerlen, respectively. The SHAP algorithm was utilized to visualize patient characteristics and laboratory data patterns that underlie individual RISKINDEX predictions. CONCLUSIONS: Our clinical decision support tool has excellent diagnostic performance in predicting 31-day mortality in ED patients. Follow-up studies will assess whether implementation of these algorithms can improve clinically relevant end points.

Defining trimester-specific reference intervals for carbohydrate deficient transferrin in pregnant women.

OBJECTIVES: Extensive consumption of alcohol during pregnancy can lead to severe complications for the unborn child. Carbohydrate-deficient transferrin (CDT) levels in serum have become a common biomarker for excessive alcohol intake. However, during pregnancy CDT levels can rise to levels above commonly used cut-off values, for reasons unrelated to alcohol intake. The aim of this study is to investigate the changes in CDT values during pregnancy and to determine accurate, trimester dependent reference intervals. METHODS: 439 serum samples of 147 healthy pregnant women were obtained for trimester 1, 2, 3, and post-partum and were analysed by high-performance liquid chromatography (HPLC) and an N-Latex immunonephelometric assay. New trimester-specific reference intervals were established. RESULTS: This study demonstrates there is a trimester-dependent increase of %CDT, as up to 39.4% of the population exceeded the previously established upper reference limit of 1.7%. In our study the estimated upper reference limit for %DST/%CDT were 1.55%, 1.96%, 2.05% and 1.35% for trimester 1, 2, 3 and post-partum for the HPLC-method and 2.02%, 2.19%, 2.19% and 1.96% for the N-Latex immunoassay. CONCLUSIONS: We demonstrate that CDT levels rise during pregnancy. The magnitude of the increase is method-dependent and needs to be taken into account. We have established method- and trimester-specific reference intervals to prevent false-positive results in alcohol abuse screening tests during pregnancy.

ISO 15189 is a sufficient instrument to guarantee high-quality manufacture of laboratory developed tests for in-house-use conform requirements of the European In-Vitro-Diagnostics Regulation.

The EU In-Vitro Diagnostic Device Regulation (IVDR) aims for transparent risk-and purpose-based validation of diagnostic devices, traceability of results to uniquely identified devices, and post-market surveillance. The IVDR regulates design, manufacture and putting into use of devices, but not medical services using these devices. In the absence of suitable commercial devices, the laboratory can resort to laboratory-developed tests (LDT) for in-house use. Documentary obligations (IVDR Art 5.5), the performance and safety specifications of ANNEX I, and development and manufacture under an ISO 15189-equivalent quality system apply. LDTs serve specific clinical needs, often for low volume niche applications, or correspond to the translational phase of new tests and treatments, often extremely relevant for patient care. As some commercial tests may disappear with the IVDR roll-out, many will require urgent LDT replacement. The workload will also depend on which modifications to commercial tests turns them into an LDT, and on how national legislators and competent authorities (CA) will handle new competences and responsibilities. We discuss appropriate interpretation of ISO 15189 to cover IVDR requirements. Selected cases illustrate LDT implementation covering medical needs with commensurate management of risk emanating from intended use and/or design of devices. Unintended collateral damage of the IVDR comprises loss of non-profitable niche applications, increases of costs and wasted resources, and migration of innovative research to more cost-efficient environments. Taking into account local specifics, the legislative framework should reduce the burden on and associated opportunity costs for the health care system, by making diligent use of existing frameworks.

Evaluation of the Atellica® UAS 800: a new member of the automated urine sediment analyzer family.

BACKGROUND: In 2017 the Atellica® UAS 800 urine sediment analyzer was introduced by Siemens Healthineers. We investigated its applicability in the standardization and automation of the laboratory urinalysis workflow, including the prediction of urine culture outcome and glomerular pathology. METHODS: We evaluated the performance characteristics of the Atellica® UAS 800 and its correlation with the iQ200 (Beckman Coulter). In addition, we studied the agreement between Atellica® UAS 800 and CLINITEK Novus® and determined the predictive value of bacteria and leukocyte counts for urine culture outcome. Furthermore, we investigated the ability of Atellica® UAS 800 to identify pathological casts and dysmorphic erythrocytes in comparison to manual microscopy. RESULTS: Erythrocyte and leukocyte analyses indicated high intra- and inter-run precisions and good correlations with the iQ200. We found that the Atellica® UAS 800 detects bacteria with higher sensitivity than the iQ200. The Atellica® UAS 800 and CLINITEK Novus® showed a high degree of conformity. We determined seven combinations of clinical cut-off values of bacteria and leukocytes for predicting urine culture outcome with sensitivity, specificity, and negative predictive values of 95%, 52%, and 93%, respectively. Using the Atellica® UAS 800, hyaline casts, erythrocyte casts, leukocyte casts, and dysmorphic erythrocytes were correctly recognized in 76%, 22%, 2%, and 39% of the samples, respectively. CONCLUSIONS: The Atellica® UAS 800 is a robust, fast, and user-friendly analyzer, which accurately quantifies erythrocytes, leukocytes, bacteria and squamous epithelial cells, and may be utilized for predicting positive urine cultures. The detection of clinically important pathological casts and dysmorphic erythrocytes proved insufficient.

Evaluation of carbohydrate-deficient transferrin measurements on the V8 capillary electrophoresis system and comparison with the IFCC approved HPLC reference method and N-Latex immunonephelometric assay.

OBJECTIVES: Carbohydrate-deficient transferrin (CDT) measurements are commonly used for the identification and follow-up of individuals suspected of chronic alcohol abuse. This study describes the analytical characteristics of the CDT assay on the Helena Biosciences V8 electrophoresis analyzer and compares its diagnostic performance to the International Federation of Clinical Chemistry and Laboratory Medicine approved high performance liquid chromatography (HPLC) method and the N-Latex CDT immunonephelometric assay. METHODS: The analytical performance of the V8 assay, including the linearity and the imprecision, was studied at two separate locations. Method comparison analysis was performed by studying the correlation, bias and agreement between the V8, HPLC and the N-Latex assays in 231 patient samples. RESULTS: The total imprecision ranged between 5.1 and 24.3% and was ≤13.1% for samples with concentrations above the clinical cut-off value (≥1.62%). The method comparisons revealed excellent correlations with r2≥0.97 for all comparisons. Measurements on the V8 showed a bias of -0.83 (-22.24%) and -0.40 (-12.26%) with the HPLC and N-Latex assays, respectively. The assays showed excellent agreements (Kappa scores ≥ 0.8) in classifying subjects with elevated CDT values. Receiver operating characteristic (ROC)-curve analysis, using the HPLC classification as reference, revealed areas under the ROC-curves of 0.981 (95% CI, 0.97-0.99) and 0.996 (0.99-1.00) for the N-Latex and V8 assays, respectively. CONCLUSIONS: CDT measurements on the V8 assay are highly correlated with both the HPLC and the N-Latex assay and show excellent agreement in classifying subjects with elevated CDT values. Overall, the V8 CDT analysis is a robust, reliable and effective method to measure CDT concentrations in serum samples.

Current state of the morphological assessment of urinary erythrocytes in The Netherlands: a nation-wide questionnaire.

Background The morphological assessment of urinary erythrocytes (uRBC) is a convenient screening tool for the differentiation of nephrological (dysmorphic) and urological (isomorphic) causes of hematuria. Considering the morphological heterogeneity, this analysis is often perceived as difficult. There is no clear (inter)national consensus and there is a lack of external quality assessment programs. To gain insight into the heterogeneity within and between laboratories, we scrutinized the current state of this analysis in Dutch medical laboratories. Methods The laboratories, affiliated with the Dutch Foundation for Quality Assessment in Medical Laboratories, were invited to participate in a web-based survey, consisting of two questionnaires. The first one provided information about the institution and laboratory organization, and the second explored the variability in the morphological analysis of uRBC on the basis of categorization of 160 uRBC images. Statistical analysis was premised on binomial significance testing and principal component analysis. Results Nearly one third of the Dutch medical laboratories (65/191) with 167 staff members participated in the survey. Most of these laboratories (83%) were an integral part of secondary care. The statistical analysis of the evaluations of the participants in comparison to the consensus (three experts from two different medical laboratories) suggested a great degree of heterogeneity in the agreement. Nearly half of the participants consciously disagreed with the consensus, whereas one fifth demonstrated a random relationship with it. Conclusions In Dutch medical laboratories, results from morphological analysis of uRBC are heterogeneous, which point out the necessity for standardization and harmonization.

The end of the laboratory developed test as we know it? Recommendations from a national multidisciplinary taskforce of laboratory specialists on the interpretation of the IVDR and its complications.

The in vitro diagnostic medical devices regulation (IVDR) will take effect in May 2022. This regulation has a large impact on both the manufacturers of in vitro diagnostic medical devices (IVD) and clinical laboratories. For clinical laboratories, the IVDR poses restrictions on the use of laboratory developed tests (LDTs). To provide a uniform interpretation of the IVDR for colleagues in clinical practice, the IVDR Task Force was created by the scientific societies of laboratory specialties in the Netherlands. A guidance document with explanations and interpretations of relevant passages of the IVDR was drafted to help laboratories prepare for the impact of this new legislation. Feedback from interested parties and stakeholders was collected and used to further improve the document. Here we would like to present our approach to our European colleagues and inform them about the impact of the IVDR and, importantly we would like to present potentially useful approaches to fulfill the requirements of the IVDR for LDTs.

Rapidly rule out acute myocardial infarction by combining copeptin and heart-type fatty acid-binding protein with cardiac troponin.

BACKGROUND: The rapid exclusion of acute myocardial infarction in patients with chest pain can reduce the length of hospital admission, prevent unnecessary diagnostic work-up and reduce the burden on our health-care systems. The combined use of biomarkers that are associated with different pathophysiological aspects of acute myocardial infarction could improve the early diagnostic assessment of patients presenting with chest pain. METHODS: We measured cardiac troponin I, copeptin and heart-type fatty acid-binding protein concentrations in 584 patients who presented to the emergency department with acute chest pain. The diagnostic performances for the diagnosis of acute myocardial infarction and NSTEMI were calculated for the individual markers and their combinations. Separate calculations were made for patients presenting to the emergency department <3 h, 3-6 h and 6-12 h after chest pain onset. RESULTS: For ruling out acute myocardial infarction, the net predictive values (95% CI) of cardiac troponin I, copeptin and heart-type fatty acid-binding protein were 90.4% (87.3-92.9), 84% (79.8-87.6) and 87% (83.5-90), respectively. Combining the three biomarkers resulted in a net predictive value of 95.8% (92.8-97.8). The improvement was most pronounced in the early presenters (<3 h) where the combined net predictive value was 92.9% (87.3-96.5) compared to 84.6% (79.4-88.9) for cardiac troponin I alone. The area under the receiver operating characteristic for the triple biomarker combination increased significantly (P < 0.05) compared to that of cardiac troponin I alone (0.880 [0.833-0.928] vs. 0.840 [0.781-0.898], respectively). CONCLUSIONS: Combining copeptin, heart-type fatty acid-binding protein and cardiac troponin I measurements improves the diagnostic performance in patients presenting with chest pain. Importantly, in patients who present early (<3 h) after chest pain onset, the combination improves the diagnostic performance compared to the standard cardiac troponin I measurement alone.

Cardiac troponin in ischemic cardiomyocytes: intracellular decrease before onset of cell death.

AIM: Cardiac troponin I (cTnI) and T (cTnT) are the most important biomarkers in the diagnosis of acute myocardial infarction (AMI). Nevertheless, they can be elevated in the absence of AMI. It is unclear if such elevations represent irreversible cardiomyocyte-damage or leakage from viable cardiomyocytes. Our objective is to evaluate whether cTn is released from viable cardiomyocytes in response to ischemia and to identify differences in the release of cTn and its molecular forms. METHODS AND RESULTS: HL-1 cardiomyocytes (mouse) were subjected to ischemia (modeled by anoxia with glucose deprivation). The total contents and molecular forms of cTn were determined in culture media and cell lysates. Cell viability was assessed from the release of lactate dehydrogenase (LDH). Before the release of LDH, the intracellular cTn content in ischemic cells decreased significantly compared to control (52% for cTnI; 23% for cTnT) and was not matched by a cTn increase in the medium. cTnI decreased more rapidly than cTnT, resulting in an intracellular cTnT/cTnI ratio of 25.5 after 24 h of ischemia. Western blots revealed changes in the relative amounts of fragmented cTnI and cTnT in ischemic cells. CONCLUSIONS: HL-1 cardiomyocytes subjected to simulated ischemia released cTnI and cTnT only in combination with the release of LDH. We find no evidence of cTn release from viable cardiomyocytes, but did observe a significant decrease in cTn content, before the onset of cell death. Intracellular decrease of cTn in viable cardiomyocytes can have important consequences for the interpretation of cTn values in clinical practice.

A comprehensive review of upper reference limits reported for (high-)sensitivity cardiac troponin assays: the challenges that lie ahead.

Cardiac troponins (cTn) are the preferred markers for the diagnosis of acute myocardial infarction (AMI). The guidelines recommend the use of the 99th percentile upper reference concentration of a healthy population as the diagnostic cut-off for AMI. However, a broad range of upper reference limits is still employed, complicating the diagnosis of AMI. This overview is meant to assist laboratory specialists to define an appropriate cut-off value for the diagnosis of AMI. Therefore, we provide an overview of the analytical performance and upper reference limits of seven (high-)sensitivity cTn assays: Roche high-sensitivity cTnT and ADVIA Centaur, Stratus CS, Dimension Vista, Vitros ECi, Access and Architect cTnI assays. It is shown that none of the reference populations completely met the guidelines, including those in package inserts. Forty percent of the studies collected less than the advised minimum of 300 subjects. Many studies (50%) did not report their inclusion criteria, while lower 99th percentile limits were observed when more stringent selection criteria were applied. Higher troponin cut-offs were found in men and elderly subjects, suggesting sex- and age-specific cut-offs would be considered. Therefore, there is still need for a large, rigorously screened reference population to more accurately establish cTn upper reference limits.

Cardiac troponin T elevations, using highly sensitive assay, in recreational running depend on running distance.

BACKGROUND: Endurance exercise is frequently associated with cardiac troponin (cTn) concentrations, otherwise corresponding to minor myocardial infarction. However, research on the underlying mechanisms has been limited because of assay restraints in the low concentration range. METHODS AND RESULTS: Using the pre-commercial, highly sensitive hs-cTnT assay, cTnT concentrations were measured in samples from recreational runners obtained before and after running 5 km (trained, n = 43/untrained, n = 122), 15 km (n = 38), 21 km (n = 10), and 42 km (n = 85) (all trained). The percentage of runners with elevated cTnT concentrations after the run increased with running distance (0, 11, 13, 40, and 86%), in contrast to NT-proBNP (2, 7, 0, 0, 5). Median (IQR) cTnT post-run concentrations were 0.004 microg/L (0.003), 0.006 microg/L (0.008), 0.010 microg/L (0.006), 0.014 microg/L (0.019), and 0.030 microg/L (0.029), respectively. CONCLUSIONS: We found, using a novel hs-cTnT assay, the distance of recreational competitive running to be positively related to asymptomatic increases in cTnT post-run concentrations. In contrast, NT-proBNP showed no increase. In addition, the data indicated that a relatively short running distance of 5 km resulted in cTnT release of untrained participants, in contrast to trained participants, which underlines the necessity of sufficient training. Further effort is needed to clarify the significance of exercise-induced cardiac biomarker elevations.