A Risk Assessment Model for Predicting Perioperative Venous Thromboembolism in Patients Receiving Surgery under Anesthesia Care.

Summary

Using 319,134 surgical cases, a PSI-12–aligned perioperative VTE model achieved AUCs of 0.87 (development), 0.84 (internal temporal), and 0.76 (external validation), outperforming Caprini and Rogers scores. It predicted VTE effectively both preoperatively and postoperatively.

Key Findings

• Model achieved AUC 0.87 (development), 0.84 (internal temporal validation), and 0.76 (external validation) across diverse surgical populations.
• Outperformed Caprini (AUC 0.66) and Rogers (AUC 0.51) risk models.
• Predicted VTE effectively both before surgery (AUC 0.91) and after surgery (AUC 0.84).

Clinical Implications

Hospitals could adopt this model to better target pharmacologic/mechanical prophylaxis and surveillance from admission through discharge, potentially reducing VTE events and costs.

Limitations

• Retrospective registry-based design with reliance on ICD coding and imaging orders
• External validation limited to specific centers; generalizability to other health systems requires testing

Future Directions

Prospective implementation trials with decision support integration, calibration in diverse health systems, and impact evaluation on VTE incidence and bleeding.