Construction and validation of a machine learning-based prediction model for venous thromboembolism in lung transplant recipients supported by ECMO - Scorecard - MDSpire

Construction and validation of a machine learning-based prediction model for venous thromboembolism in lung transplant recipients supported by ECMO

  • By

  • Yan Zhu

  • Fei Zeng

  • Mei-Juan Lan

  • Jiang-Shu-Yuan Liang

  • Ling-Yun Cai

  • Pei-Pei Gu

  • Lu-Yao Guo

  • June 4, 2026

  • 0 min

Share

Clinical Scorecard: Development and validation of a machine learning-driven model to predict venous thromboembolism in lung transplant patients on ECMO

At a Glance

CategoryDetail
ConditionVenous Thromboembolism (VTE) in lung transplant patients on ECMO
Key MechanismsMachine learning-driven risk prediction model
Target PopulationLung transplant recipients receiving ECMO
Care SettingSecond Affiliated Hospital of Zhejiang University

Key Highlights

  • Study included 189 lung transplant patients receiving ECMO.
  • Random Forest model achieved an AUC of 0.895.
  • Accuracy of the model was 89.7%, with sensitivity and specificity both around 89.5%.
  • Calibration curve showed strong agreement between predicted probabilities and observed outcomes.
  • Decision curve analysis indicated significant clinical utility.

Guideline-Based Recommendations

Diagnosis

  • Initial assessment for VTE using bedside color Doppler ultrasound (CUS) within 48 hours post-operation.

Management

  • Anticoagulation with unfractionated heparin, monitored by activated clotting time (ACT) targeting 160-180 seconds.

Monitoring & Follow-up

  • Regular monitoring of ACT and APTT for anticoagulation management.

Risks

  • Thrombotic events during ECMO support occur at a rate of nearly 36%.

Patient & Prescribing Data

Adults aged ≥18 years undergoing elective lung transplantation with ECMO support.

Anticoagulation regimen should be individualized based on patient characteristics and ECMO configuration.

Clinical Best Practices

  • Use machine learning models to enhance VTE risk prediction.
  • Implement routine anticoagulation monitoring during ECMO.

Related Resources & Content

Original Source(s)

Construction and validation of a machine learning-based prediction model for venous thromboembolism in lung transplant recipients supported by ECMO

  • by Yan Zhu, Fei Zeng, Mei-Juan Lan, Jiang-Shu-Yuan Liang, Ling-Yun Cai, Pei-Pei Gu, Lu-Yao Guo

  • June 4, 2026

Related Content

Association Between Edmonton Obesity Staging System Severity and 90-Day Postoperative Complications After Primary Metabolic and Bariatric Surgery: A Retrospective Cohort Study

Impact of internet-based multidisciplinary management model on continuity of care intervention for postoperative outcomes in patients with mixed hemorrhoids: a retrospective cohort study

Distinct prognostic value of aspartate-to-alanine aminotransferase ratio (AAR) in traumatic brain injury versus hemorrhagic stroke: a cohort study of 1,069 patients