Development and Validation of a Risk Assessment Model for Venous Thromboembolism in the Pediatric Intensive Care Unit

Introduction: Children requiring pediatric intensive care unit (PICU) management are at increased risk for venous thromboembolism (VTE) as well as bleeding complications due to critical illness. There is limited understanding regarding the subset of patients who may benefit from thromboprophylaxis during their PICU stay. The purpose of this study was to determine the risk factors for VTE in critically ill children and to develop and validate a risk assessment model (RAM) tailored for this patient population. Methods: Data from the Pediatric Health Information System (PHIS) was used for this study. Data for children (1 month - 21 years of age) directly admitted or transferred to the PICU within 5 days of admission were collected from PHIS from 2015-2023. Patients were excluded if they had a billing code for a neonatal ICU admission or missing sex information. Subsequent encounters for the same patient were also excluded. Data were collected regarding demographics, duration of hospitalization, reason for admission, acute organ dysfunction and underlying complex chronic condition (CCC). VTE was defined as presence of deep vein thrombosis (DVT), pulmonary embolism (PE) or both. For those with a VTE, data were collected regarding the type and location of VTE. Training (70%) and validation (30%) datasets were randomly generated using the 2015-2020 data. RAM testing was then performed on three consecutive annual datasets (2021, 2022, and 2023). For RAM derivation, 35 variables were initially chosen based on their relevance as VTE risk factors and data availability from PHIS. Predictors were screened using generalized logistic regression models. Predictors with p <0.05 were included in the final multivariable model. Least Absolute Shrinkage and Selection Operator (LASSO) was used to generate the final regression model. Discrimination was measured using receiver operating characteristic (ROC) curves and calibration plots were used to assess model accuracy. Results: We identified a total of 98,589 unique PICU managed patients. 56% of the study population were male and 44% were non-Hispanic White race. Of these, 5,060 (5.13%) experienced a VTE. Deep vein thrombosis was the most common VTE (n=3960, 78%) followed by pulmonary embolism (n=523, 10%). Median age at the time of hospitalization was 7.2 years (IQR: 1.0-14.8 years) for those with VTE compared to 3.6 years (IQR: 0.7-11.6 years) for those without a VTE (p <0.0001). Length of PICU stay was longer for those with a VTE (median of 11 days [IQR: 5-25] vs. 5 days [IQR: 3-8], p<0.01). Patients with VTE had 2.7 times higher odds of mortality in comparison to those without a VTE (p<0.001). Among the variables that were included in the final regression model: central venous line (CVL) (OR 6.52; 95% CI: 6.03-7.04); acute renal dysfunction (OR 3.95; 95% CI: 3.71-4.21); acute hemostatic derangement (OR 3.76; 95% CI: 3.47-4.08); sepsis (OR 3.26; 95% CI: 3.05-3.49); solid or hematopoietic cell transplantation (OR 3.09; 95% CI: 2.75-3.46); acute cardiac dysfunction (OR 2.80; 95% CI: 2.63-2.98); mechanical ventilation (OR 2.34; 95% CI: 2.20-2.50); recent surgery (OR 2.05; 95% CI: 1.93-2.18); chronic cardiac (OR 1.88; 95% CI: 1.78-1.99), GI (OR 1.56; 95% CI: 1.47-1.66), hematologic/immune (OR 2.20; 95% CI: 2.02-2.38), metabolic (OR 2.58; 95% CI: 2.41-2.74), and renal conditions (OR 2.20; 95% CI: 2.04-2.38); trauma (OR 1.51; 95% CI: 1.39-1.64) and age (OR 1.05; 95% CI: 1.04-1.05) were identified as independent risk factors for VTE. The ROC area under the curve (AUC) for the validation model was 0.78. Upon testing in the 3 independent annual cohorts, the RAM performed well with ROC AUCs of 0.80 (2021), 0.80 (2022), and 0.78 (2023). Conclusions: Using an administrative database, we successfully derived and validated a RAM specifically for PICU associated VTE. Among the several known risk factors, CVL was the most important predictor of VTE risk. While prospective validation is needed, this new RAM may improve VTE risk stratification and guide thromboprophylaxis use in children managed in the PICU.