Doctors Identify Hemorrhage Better during Chart Review when Assisted by Artificial Intelligence

OBJECTIVES

This study evaluated if medical doctors could identify more hemorrhage events during chart review in a clinical setting when assisted by an artificial intelligence (AI) model and medical doctors' perception of using the AI model.

METHODS

To develop the AI model, sentences from 900 electronic health records were labeled as positive or negative for hemorrhage and categorized into one of 12 anatomical locations. The AI model was evaluated on a test cohort consisting of 566 admissions. Using eye-tracking technology, we investigated medical doctors' reading workflow during manual chart review. Moreover, we performed a clinical use study where medical doctors read two admissions with and without AI assistance to evaluate performance when using and perception of using the AI model.

RESULTS

The AI model had a sensitivity of 93.7% and a specificity of 98.1% on the test cohort. In the use studies, we found that medical doctors missed more than 33% of relevant sentences when doing chart review without AI assistance. Hemorrhage events described in paragraphs were more often overlooked compared with bullet-pointed hemorrhage mentions. With AI-assisted chart review, medical doctors identified 48 and 49 percentage points more hemorrhage events than without assistance in two admissions, and they were generally positive toward using the AI model as a supporting tool.

CONCLUSION

Medical doctors identified more hemorrhage events with AI-assisted chart review and they were generally positive toward using the AI model.

Laboratory Tests and Outcome for Patients with Coronavirus Disease 2019: A Systematic Review and Meta-Analysis

BACKGROUND

Severe acute respiratory syndrome coronavirus 2 causes coronavirus disease 2019 (COVID-19) and poses substantial challenges for healthcare systems. With a vastly expanding number of publications on COVID-19, clinicians need evidence synthesis to produce guidance for handling patients with COVID-19. In this systematic review and meta-analysis, we examine which routine laboratory tests are associated with severe COVID-19 disease.

CONTENT

PubMed (Medline), Scopus, and Web of Science were searched until March 22, 2020, for studies on COVID-19. Eligible studies were original articles reporting on laboratory tests and outcome of patients with COVID-19. Data were synthesized, and we conducted random-effects meta-analysis, and determined mean difference (MD) and standard mean difference at the biomarker level for disease severity. Risk of bias and applicability concerns were evaluated using the Quality Assessment of Diagnostic Accuracy Studies-2.

SUMMARY

45 studies were included, of which 21 publications were used for the meta-analysis. Studies were heterogeneous but had low risk of bias and applicability concern in terms of patient selection and reference standard. Severe disease was associated with higher white blood cell count (MD, 1.28 ×109/L), neutrophil count (MD, 1.49 ×109/L), C-reactive protein (MD, 49.2 mg/L), lactate dehydrogenase (MD, 196 U/L), D-dimer (standardized MD, 0.58), and aspartate aminotransferase (MD, 8.5 U/L); all p < 0.001. Furthermore, low lymphocyte count (MD -0.32 × 109/L), platelet count (MD -22.4 × 109/L), and hemoglobin (MD, -4.1 g/L); all p < 0.001 were also associated with severe disease. In conclusion, several routine laboratory tests are associated with disease severity in COVID-19.

The role of platelets in bleeding in patients with thrombocytopenia and hematological disease

This review evaluates the role of platelets in bleeding risk among patients with hematological disease and thrombocytopenia. Platelets are pivotal in primary hemostasis, and possess non-hemostatic properties involved in angiogenesis, tissue repair, inflammation and metastatis. Also, platelets safeguard vascular integrity in inflamed vessels. Overall, bleeding risk depends on the underlying disease, and patients with cancer and platelet count <6-10 × 109/L have a markedly increased bleeding risk, while the platelet count does not correlate with bleeding risk at higher platelet counts. Other factors might affect platelet properties and thus bleeding risk, for example, drugs, low hematocrit, coagulation system impairments or transfusion of dysfunctional donor platelets. For patients with leukemia and immune thrombocytopenia, reduced platelet activation, platelet aggregation, or thrombopoiesis, reflected by the reduced presence of reticulated platelets, are associated with bleeding phenotype. However, mechanistic insight into the cause of reduced platelet function in different thrombocytopenic conditions is sparse, except for some inherited platelet disorders. Promising tools for platelet function studies in thrombocytopenia are flow cytometry and biomarker studies on platelet constituents. An important message from this current paper is that bleeding risk assessment must be tailored to specific patient populations and cannot be applied broadly to all patients with thrombocytopenia.

Platelet aggregation is not altered among men with diabetes mellitus

AIMS

Platelets are pivotal in arterial thrombosis, and platelet hyperresponsiveness may contribute to the increased incidence of cardiovascular events in diabetes mellitus. Consequently, we hypothesized that increased in vitro platelet aggregation responses exist in men with diabetes mellitus.

METHODS

The Danish Cardiovascular Screening Trial (DANCAVAS) is a community-based cardiovascular screening trial including men aged 65-74 years. Platelet aggregation was tested using 96-well light transmission aggregometry with thrombin receptor-activating peptide (TRAP), adenosine diphosphate, collagen type 1, arachidonic acid and protease-activated receptor-4 in three concentrations. Further, cardiovascular risk factors and coronary artery calcification (CAC), estimated by CT scans and ankle-brachial index, were obtained.

RESULTS

Included were 720 men aged 65-74 years, 110 with diabetes mellitus. Overall, there was no difference in platelet aggregation among men with versus without diabetes mellitus when adjusting for or excluding platelet inhibitor treatment and men with established cardiovascular disease (CVD). This was true for all agonists, e.g., 10 µM TRAP-induced platelet aggregation of median 69% (IQR 53-75) versus 70% (IQR 60-76) in men with versus without diabetes mellitus. Platelet aggregation did not correlate with HbA_1c or CAC. Men with diabetes mellitus displayed higher CAC, median 257 Agatston units (IQR 74-1141) versus median 111 Agatston units (IQR 6-420) in the remaining individuals, p < 0.0001.

CONCLUSIONS

Among outpatients with diabetes mellitus, but no CVD and no platelet inhibitor treatment, neither are platelets hyperresponsive in diabetes mellitus, nor is platelet aggregation associated with glycemic status or with the degree of coronary atherosclerosis.

TRIAL REGISTRATION

ISRCTN12157806.

Use of clinical data and acceleration profiles to validate pneumatic transportation systems

Background

Modern pneumatic transportation systems (PTSs) are widely used in hospitals for rapid blood sample transportation. The use of PTS may affect sample integrity. Impact on sample integrity in relation to hemolysis and platelet assays was investigated and also, we wish to outline a process-based and outcome-based validation model for this preanalytical component.

Methods

The effect of PTS was evaluated by drawing duplicate blood samples from healthy volunteers, one sent by PTS and the other transported manually to the core laboratory. Markers of hemolysis (potassium, lactate dehydrogenase [LD] and hemolysis index [HI]) and platelet function and activation were assessed. Historic laboratory test results of hemolysis markers measured before and after implementation of PTS were compared. Furthermore, acceleration profiles during PTS and manual transportation were obtained from a mini g logger in a sample tube.

Results

Hand-carried samples experienced a maximum peak acceleration of 5 g, while peaks at almost 15 g were observed for PTS. No differences were detected in results of potassium, LD, platelet function and activation between PTS and manual transport. Using past laboratory data, differences in potassium and LD significantly differed before and after PTS installation for all three lines evaluated. However, these estimated differences were not clinically significant.

Conclusions

In this study, we found no evidence of PTS-induced hemolysis or impact on platelet function or activation assays. Further, we did not find any clinically significant changes indicating an acceleration-dependent impact on blood sample quality. Quality assurance of PTS can be performed by surveilling outcome markers such as HI, potassium and LD.

The effects of storage on platelet function in different blood products

OBJECTIVES

Reduced platelet (PLT) function during storage has been shown for buffy-coat-derived platelet concentrates (BCP) and apheresis platelet units (AP), while for whole blood (WB) it has not been well studied. The aim of this study was to investigate PLT function in these blood products throughout storage using a novel flow cytometric assay.

METHODS

Flow cytometric measurement of agonist-induced platelet aggregation, CD62P expression and PAC-1 binding during storage in BCP, AP (1-9 days at 20°C) and WB (1-21 days at 2-6°C).

RESULTS

PLT-aggregation capacity decreased from day 1 to day 7 for almost all product-agonist combinations (P = .004 to P = .029) with aggregation capacity of WB being similar to that of AP and BCP. WB aggregation capacity remained relatively unchanged from day 7 to day 21. For all blood products, the fraction of agonist-induced CD62P-expression remained high and the fraction of PAC-1 binding decreased during storage. WB PLTs underwent only small changes in CD62P expression and PAC-1 binding from day 7 to day 21.

CONCLUSION

This study found PLT aggregation in WB stored at 4°C to be as least as good as for BCP and AP stored at 20°C. WB retained significant PLT-aggregation capacity to day 21.

Measurement of platelet aggregation, independently of patient platelet count: a flow-cytometric approach

Essentials Platelet function may influence bleeding risk in thrombocytopenia, but useful tests are needed. A flow cytometric platelet aggregation test independent of the patient platelet count was made. Platelet aggregation was reduced in thrombocytopenic patients with hematological cancer. High platelet aggregation ruled out bleeding tendency in thrombocytopenic patients.

SUMMARY

Background Methods for testing platelet aggregation in thrombocytopenia are lacking. Objective To establish a flow-cytometric test of in vitro platelet aggregation independently of the patient's platelet count, and examine the association of aggregation with a bleeding history in thrombocytopenic patients. Patients/methods We established a flow-cytometric assay of platelet aggregation, and measured samples from healthy individuals preincubated with antiplatelet drugs, and samples from two patients with inherited platelet disorders. Then, we included 19 healthy individuals and 20 patients with platelet counts of ≤ 50 × 109 L-1 , diagnosed with acute myeloid leukemia or myelodysplastic syndrome. We measured platelet aggregation and platelet activation by platelet surface expression of activated glycoprotein IIb-IIIa, P-selectin and CD63 after addition of agonists: collagen-related peptide, thrombin receptor-activating peptide (TRAP), and ADP. Results The platelet aggregation assay showed a low intraserial coefficient of variation of ≤ 3%. Similar results were obtained for platelet-rich plasma and isolated platelets at platelet counts of > 10 × 109 L-1 ; otherwise, platelet isolation was required. The platelet aggregation percentage decreased with increasing antiplatelet drug concentration. Platelet aggregation in patients was reduced as compared with healthy individuals: 42% (interquartile range [IQR] 27-58) versus 66% (IQR 60-67) for TRAP; 41% (IQR 25-48) versus 70% (IQR 69-72) for collagen-related peptide; and 44% (IQR 30-53) versus 65% (IQR 46-72) for ADP. Platelet activation after stimulation was reduced in patients and correlated with platelet aggregation (e.g. r = 0.78-0.81 when stimulated with collagen-related peptide). Platelet aggregation had a negative predictive value of 100% for a bleeding tendency among patients. Conclusion The established platelet aggregation assay was applicable for thrombocytopenic patients, and improved the identification of bleeding risk.

The effect of centrifugation speed and time on pre-analytical platelet activation

BACKGROUND

The results of laboratory analyses are affected by pre-analytical variables, and in particular can platelets be activated by shear handling stress and secrete granular substances. We therefore evaluated the effect of centrifugation speed and time on pre-analytical platelet activation.

METHODS

Citrate- and EDTA-anticoagulated blood from healthy volunteers were centrifuged at 80-10,000 g for 5-15 min to prepare plasma and platelet-rich plasma. Pre-analytical platelet activation was assessed by flow cytometric measurement of platelet P-selectin (CD62p) expression. Blood cell counts, mean platelet volume (MPV), immature platelet fraction (IPF), and platelet distribution width (PDW) were measured. Platelet aggregation in platelet-rich plasma induced by arachidonic acid (AA), ADP or thrombin receptor activator peptide-6 (TRAP) was tested by 96-well aggregometry.

RESULTS

The median percentage of platelets expressing P-selectin in citrate- and EDTA-plasma centrifuged at 2000 g for 10 min were 43% [interquartile range (IQR), 38%-53%] and 56% (IQR, 31%-78%), respectively (p=0.82). Platelet-rich plasma prepared at 100-250 g for 10 min had significantly lower platelet P-selectin expression (11%-15%), p<0.001. Platelet count in plasma samples decreased with increasing speed but platelets were only completely removed if plasma was re-centrifuged. In platelet-rich plasma, increasing centrifugation speed significantly increased platelet yield but decreased contamination from other blood cells, platelet composition was altered as platelet parameters (MPV, IPF, and PDW) was lowered. Platelet aggregation was not affected by the centrifugation speed platelet-rich plasma was prepared.

CONCLUSIONS

Proportional to centrifugation speed, platelets in plasma and platelet-rich plasma were activated with centrifugation speed, cell content and composition changed while platelet aggregation was unaltered.

An overview of platelet indices and methods for evaluating platelet function in thrombocytopenic patients

Thrombocytopenia is associated with bleeding risk. However, in thrombocytopenic patients, platelet count does not correlate with bleeding risk and other factors are thus likely to contribute to this risk. This review presents currently available platelet-related markers available on automated haematology analysers and commonly used methods for testing platelet function. The test principles, advantages and disadvantages of each test are described. We also evaluate the current literature regarding the clinical utility of the test for prediction of bleeding in thrombocytopenia in haematological and oncological diseases. We find that several platelet-related markers are available, but information about the clinical utility in thrombocytopenia is limited. Studies support that mean platelet volume (MPV) can aid diagnosing the cause of thrombocytopenia and low MPV may be associated with bleeding in thrombocytopenia. Flow cytometry, platelet aggregometry and platelet secretion tests are used to diagnose specific platelet function defects. The flow cytometric activation marker P-selectin and surface coverage by the Cone-and-Plate[let] analyser predict bleeding in selected thrombocytopenic populations. To fully uncover the clinical utility of platelet-related tests, information about the prevalence of platelet function defects in thrombocytopenic conditions is required. Finally, knowledge of the performance in thrombocytopenic samples from patients is essential.

An ELISA for the quantitation of von Willebrand factor: osteoprotegerin complexes in plasma

BACKGROUND

Von Willebrand factor (VWF) is pivotal in arterial thrombosis, and osteoprotegerin (OPG) is besides being a bone protein also related to cardiovascular diseases. OPG can bind VWF, but the significance of this interaction is not known.

OBJECTIVES

The aim was to develop an assay for measurement of von Willebrand factor-osteoprotegerin complex (VWF:OPG) in human plasma. Furthermore, the significance of VWF:OPG complex as a marker of cardiovascular disease (CVD) was evaluated.

PATIENTS/METHODS

A sandwich ELISA for quantification of VWF:OPG was developed using a polyclonal rabbit anti-human VWF capturing antibody and a monoclonal anti-human OPG detecting antibody. Samples were quantified relative to a standard curve obtained from dilutions of a plasma pool from healthy individuals. The assay was evaluated in two groups of patients with CVD and two groups of asymptomatic individuals with and without documented coronary calcification (total n=118).

RESULTS AND CONCLUSIONS

The assay detected VWF:OPG complexes in human plasma, while no significant signal was observed when testing solutions containing VWF or recombinant OPG alone. Importantly, the ELISA assay was able to detect in vitro formed complexes between human VWF and recombinant OPG in a dose-dependent manner. There was a large inter-individual variation in plasma VWF:OPG levels, but we found no significant differences in the level of VWF:OPG complexes between the four groups. Thus, we conclude that increasing OPG plasma levels in atherosclerotic CVD are not derived from increased levels of complexed form of VWF and OPG, but are more likely due to increased amounts of OPG secreted into the circulation.

Does C-reactive protein independently predict mortality in adult community-acquired bacteremia patients with known sepsis severity?

We evaluated whether sepsis severity and C-reactive protein (CRP) level on admission prognostically corroborated or annulled each other in adult patients with incident community-acquired bacteremia (Funen, Denmark, 2000-2008). We used logistic regression and area under the receiver operating characteristic curve (AUC) to evaluate 30-day mortality in four models: (i) age, gender, comorbidity, bacteria, and ward. (ii) Model 1 and sepsis severity. (iii) Model 1 and CRP. (iv) Model 1, sepsis severity, and CRP. Altogether, 416 of 1999 patients died within 30 days. CRP independently predicted 30-day mortality [Model 4, odds ratio (95% CIs) for 100 mg/L: 1.16 (1.06-1.27)], but it did not contribute to the AUC (Model 2 vs Model 4: p = 0.31). In the 963 non-severe sepsis patients, CRP independently predicted 30-day mortality [Model 4: 1.42 (1.20-1.69)] and it increased the AUC (Model 2 vs Model 4: p = 0.06), thus CRP contributed as much as sepsis severity to prognosis.