Multisite validation of a venous thrombosis risk model in critically ill children through the CHAT Consortium

Critically ill children have a high risk of hospital associated venous thromboembolism (HA-VTE). Developing a validated risk assessment model (RAM) to identify children who may benefit from thromboprophylaxis is essential. We aimed to prospectively validate the Children's Healthcare Advancements in Thrombosis (CHAT) intensive care unit (ICU) VTE RAM containing five clinically significant variables: CVC, immobility, congenital heart disease, autoimmune or inflammatory conditions and hospital stay of ≥3 days in a multicenter cohort study. Randomly selected patients aged 0-21 years admitted to a pediatric ICU (PICU) at 32 institutions were monitored via medical record review for HA-VTE. Discrimination was assessed using the area under the receiver operating characteristic (AUROC) curve. Calibration was assessed using calibration plots. Complete case and imputed analyses were performed and model risk scores were generated along with post-test probability. The RAM was validated in 4,674 patients with an AUROC of 0.71 [95% CI:0.64-0.78], calibration slope of 1.0 (95% CI:0.87-1.14) and intercept of 1.81x10-5 (95% CI: -5.40x10-3-5.44x10- 3). The AUROC for the imputed model was 0.69 (95% CI:0.68-0.70) with the calibration slope of 1.03 (95% CI: 0.85-1.21) and intercept of 1.22x10-3 (95% CI: -7.80x10-3-5.36x10-3). Calculated risk scores were 1 or 2 for each variable in the RAM with a total risk score ranging from 0 to 6. The estimated probability for developing HA-VTE ranged from 1% to 17.4% depending on the total score. In conclusion, the CHAT-ICU RAM has good discriminatory validity, is well-calibrated and reliably identifies children in the PICU at high and low risk of HA-VTE.