Year: 2026
Type: article
Study design: Observational Study
Abstract: Background Venous thromboembolism (VTE), although uncommon in pediatric populations, occurs more frequently in hospitalized children. Robust, prospectively derived risk prediction models for symptomatic VTE remain limited in this setting. Aims To identify independent risk factors for VTE and to develop a predictive model for symptomatic thrombosis in hospitalized pediatric patients. Methods This prospective multicenter cohort study (September 2018–January 2022) evaluated 2,722 pediatric patients (birth–18 years) across 5,040 hospitalizations, identifying 39 imaging-confirmed symptomatic VTE cases (1.43%). A multivariable logistic regression model was built using forward selection to incorporate clinical and procedural predictors. Adjusted odds ratios (95% CI) and model discrimination (out-of-bag C-index) were validated using 50,000 BCa bootstrap resamples. Results Among 5,040 pediatric hospitalizations, 39 patients developed VTE (0.77% incidence; median age 5.0 years, 45.9% female). Compared to non-VTE controls, patients with VTE had significantly higher rates of ICU admission (79.5% vs. 26.6%), central venous catheterization (69.2% vs. 29.8%), bed immobilization (76.9% vs. 37.4%), cardiovascular disease (17.9% vs. 3.0%), and neurological disease (12.8% vs. 3.2%; all p < 0.01). Furthermore, VTE was associated with a markedly prolonged median hospital stay (29.5 vs. 5.0 days, p < 0.001). Increasing age was associated with higher risk (aOR 1.07 per year; 95% CI, 1.03–1.12), as were intensive care unit admission (aOR 1.91; 95% CI, 1.42–2.66) and longer length of hospitalization (aOR 1.01 per day; 95% CI, 1.01–1.02). Procedural factors showed the strongest associations with thrombosis, including 2≥ catheter insertion attempts (aOR 6.47; 95% CI, 2.81–14.49), femoral venous catheter (aOR 28.22; 95% CI, 7.23–90.36), and cervicothoracic venous catheter (aOR 4.71; 95% CI, 2.84–7.94). The final model demonstrated strong discriminative performance, with a C-index of 0.90 (95% CI, 0.83–0.97), sensitivity of 0.85, specificity of 0.78, and a negative predictive value of 0.99. Risk factors for symptomatic pediatric thrombosis were evaluated using multivariable logistic regression (forward selection with 50,000 BCa bootstrap resamples) alongside ten machine learning (ML) algorithms (70/30 train–test split, 10-fold cross-validation, and SHAP explainability analysis). Both statistical modeling and ML approaches independently converged on the exact same clinical and procedural predictors, confirming the biological and clinical validity of the identified risk drivers. Discussion Aligning with prior literature, the use of femoral access and insertion attempts are likely mediated by mechanical endothelial disruption, flow turbulence, catheter mechanical distress, and emergency access placement. Targeted protocols aiming to minimize catheter manipulation and avoid high-risk sites like the femoral vein are essential for VTE prevention. Conclusion Symptomatic VTE in hospitalized children is strongly influenced by modifiable procedural and clinical factors. This predictive model shows high discriminative performance and may support individualized risk stratification and targeted thromboprophylaxis strategies in pediatric inpatient care – after external validation and clinical impact assessment. Approved by Ethics Committee of UNICAMP (CAAE: 87364418.2.1001.5404). Supported by FAPESP (#2016/14172-6).
Authors: Stephany Cares HUBER, Mohamad Al BANNOUD, Silmara Aparecida de Lima MONTALVÃO, Sandra Martins SILVA, Sumara Zuanazi Pinto RIGATTO, Ricardo Onofre Benez VILELA, Arlindo Nascimento Lemos JUNIOR, Mônica Pinheiro de Almeida VERÍSSIMO, Tiago Dias MARTINS, Gabriel Yoshiaki OTTAIANO, Rubens Maciel FILHO, Walter AGENO, Leonardo Rodrigues BRANDÃO, Joyce Maria ANNICHINO-BIZZACCHI
Language: English
Subfield: Internal Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7223129611