Skip to main content
ISTH 2023 Main banner
Back

    Oral Communication

    Diagnostics and OMICs

    Progress in Hemostasis Testing: Implications for Future Practice

    OC 77.1 - Development, validation, and implementation of a decision support tool for the screening of mild bleeding disorders

    Wednesday, June 28, 2023
    10:15 – 10:30 ET
    Room: 518
    Session Available As: On-Demand

      Presenting Author(s)

    • Henning Nilius, MD, MSc (he/him/his)

      PhD Student & Resident
      Department of Clinical Chemistry, Inselspital, Bern University Hospital, and University of Bern, Bern, Switzerland
      Bern, Bern, Switzerland

      • Co-Author(s)

    • JK

      Jonas Kaufmann

      MD
      Departement of Clinical Chemistry, Inselspital Bern University Hospital
      Bern, Bern, Switzerland

    • MA

      Marcel Adler

      Senior Physician
      Departement of Hematology and Central Hematology Laboratory, Inselspital University Hospital Bern & CHUV Lausanne University Hospital
      Bern, Bern, Switzerland

    • AW

      Anna Wieland

      MD
      Department of Hematology and Central Hematology Laboratory, Bern University Hospital, University of Bern, Switzerland
      Bern, Bern, Switzerland

    • LA

      Lorenzo Alberio, MD (he/him/his)

      Chief Physician, Associated Professor
      Division and Central Laboratory of Hematology, Lausanne University Hospital (CHUV) and University of Lausanne, Lausanne, Switzerland
      Lausanne, Vaud, Switzerland

    • GE

      Gabor Erdös

      Senior Physician
      Department of Anaesthesiology and Pain Therapy, Inselspital Bern University Hospital
      Bern, Bern, Switzerland

    • TS

      Thomas C. Sauter

      Senior Physician
      Department of Emergency Medicine, Inselspital Bern University Hospita
      3010, Bern, Switzerland

    • AK

      Andreas Koster

      Senior Physician
      Institute of Anesthesiology and Pain Therapy, Heart and Diabetes Center North Rhine-Westphalia, Ruhr University Bochum
      Bad Oeynhausen, Nordrhein-Westfalen, Germany

    • Jerrold H. Levy, MD (he/him/his)

      Professor
      Department of Anesthesiology, Critical Care, and Surgery, Duke University School of Medicine
      Durham, North Carolina, United States

    • Michael Nagler, MD, PhD, MSc

      Deputy Head
      Department of Clinical Chemistry, Inselspital, Bern University Hospital, and University of Bern, Bern, Switzerland
      Bern, Bern, Switzerland

    Background: Mild bleeding disorders (MBD) are the most common inherited bleeding disorders that often manifest with perioperative hemorrhages. However, practical screening tools are not available.

    Aims: We aimed to develop, externally validate and implement an easy-to-use, machine-learning-based decision support tool for the screening for MBD.

    Methods: Detailed clinical and laboratory data were collected in two independent prospective cohort studies including consecutive patients referred for a suspected MBD to specialized outpatient units (n = 555, training cohort; n = 217 validation cohort). The diagnostic workup was done following current guidelines and an expert panel established the final diagnosis. In the training cohort, the items of the ISTH-BAT were simplified by grouping levels with similar response patterns. Predictors were selected using the “Boruta” algorithm and focus group discussions. Multiple machine-learning algorithms were fitted to the data and five algorithms were further tuned. The best-performing model was externally validated in the validation cohort.

    Results: The following predictors were selected: (a) activated partial thromboplastin time, (b) PFA-200 closure time (epinephrine/collagen cartridge), (c) sex, and (d) a reduced bleeding history (surgery, tooth extraction, epistaxis, minor wounds, cutaneous bleeding, oral bleeding, postpartum hemorrhage, and menorrhagia). In the validation cohort, 87.5 % of patients with MBD were correctly identified (sensitivity; 95% confidence interval [CI]: 79.9, 93.0), and 54.3 % (specificity; 95% CI: 44.3, 64.0) of patients without MBD would have been correctly excluded from further work-up. The AUROC was 0.86 (95% CI: 0.81, 0.90), in contrast to 0.81 (ISTH-BAT, 95% CI: 0.75, 0.87), 0.73 (PFA-200, 95% CI: 0.65, 0.79), and 0.49 (INR, 95% CI: 0.42, 0.55). The decision support tool was implemented on an easy-to-use web application (https://toradi-hit.dbmr.unibe.ch/mbdcheck/).

    Conclusion(s): We were able to develop, externally validate, and implement an effective decision support tool for screening patients with MBD.

    Learning Objectives:

    • Upon completion, participant will be able ot reflect on the current state of screening tools for mild bleeding disorders.
    • Upon completion, participant will be able to summarize the usefullness of machine-learning models to integrate multi-dimensional data into accurate prediction models.
    • Upon completion, participant will be able to name the advantages of an easy-to-use implementation for transition into clinical use.