Derivation and Validation of a Machine Learning Algorithm for Predicting Venous Thromboembolism in Injured Children
Stephanie C Papillon1, Christopher P Pennell1, Sahal A Master1
1St. Christopher's Hospital for Children, Department of Pediatric General Thoracic, and Minimally Invasive Surgery, Philadelphia, PA 19134, USA.
Insights
Machine learning models accurately predict venous thromboembolism (VTE) risk in pediatric trauma patients. This tool can help guide VTE prophylaxis decisions in injured children.
Area of Science:
- Pediatric Trauma Care
- Medical Informatics
- Clinical Prediction Models
Background:
- Venous thromboembolism (VTE) is a significant cause of morbidity in pediatric trauma patients.
- Accurate risk stratification is crucial for appropriate VTE prophylaxis in this population.
Purpose of the Study:
- To develop and validate a machine learning-based risk prediction model for VTE in injured children.
- To identify key predictors of VTE in pediatric trauma patients.
Main Methods:
- Utilized the Trauma Quality Improvement Program (TQIP) database (2017-2019) including 383,814 patients ≤18 years.
- Identified 15 predictors including intubation, oxygen requirement, spinal/pelvic/long bone fractures, major surgery, age, transfusion, ICP monitor, and Glasgow Coma Scale score.
- Trained and tested machine learning algorithms on split data subsets (training: 251,409; testing: 118,175).
Main Results:
- All developed models significantly outperformed the baseline VTE rate (0.15%) in the testing subset.
- Predicted VTE rates were low (0.01-0.02%), with 88.4-89.4% of patients classified as low risk.
- High-risk prediction models also outperformed baseline, with no significant difference in performance among the three models tested.
Conclusions:
- A validated predictive model effectively differentiates injured children at risk for VTE.
- The model demonstrates high discrimination and can inform clinical decisions regarding VTE prophylaxis.
Background:
Venous thromboembolism (VTE) causes significant morbidity in pediatric trauma patients. We applied machine learning algorithms to the Trauma Quality Improvement Program (TQIP) database to develop and validate a risk prediction model for VTE in injured children.
Methods:
Patients ≤18 years were identified from TQIP (2017-2019, n = 383,814). Those administered VTE prophylaxis ≤24 h and missing the outcome (VTE) were removed (n = 347,576). Feature selection identified 15 predictors: intubation, need for supplemental oxygen, spinal injury, pelvic fractures, multiple long bone fractures, major surgery (neurosurgery, thoracic, orthopedic, vascular), age, transfusion requirement, intracranial pressure monitor or external ventricular drain placement, and low Glasgow Coma Scale score. Data was split into training (n = 251,409) and testing (n = 118,175) subsets. Machine learning algorithms were trained, tested, and compared.
Results:
Low-risk prediction: For the testing subset, all models outperformed the baseline rate of VTE (0.15%) with a predicted rate of 0.01-0.02% (p < 2.2e-16). 88.4-89.4% of patients were classified as low risk by the models.
High-Risk Prediction:
All models outperformed baseline with a predicted rate of VTE ranging from 1.13 to 1.32% (p < 2.2e-16). The performance of the 3 models was not significantly different.
Conclusion:
We developed a predictive model that differentiates injured children for development of VTE with high discrimination and can guide prophylaxis use.
Level Of Evidence:
Prognostic, Level II.
Type Of Study:
Retrospective, Cross-sectional.
Related Concept Videos
01:29Venous Thrombosis III: Interprofessional Care
01:20Venous Thrombosis II: Clinical Manifestations and Diagnostic Studies
01:30Venous Thrombosis IV: Nursing Management
01:30Venous Thrombosis I: Introduction