Targeting Extended Thromboprophylaxis after Surgery Through Multidimensional Dynamic Pulmonary Embolism Risk Assessment: Development and Temporal Validation in 4.8 Million Operations
Ramsey M Dallal, Jeffrey Riggio, Michael Li, Luis H Eraso, Benjamin Moran, Geno J MerliBackground:
Extended thromboprophylaxis reduces postoperative venous thromboembolism but is inconsistently used because pulmonary embolism (PE) risk varies across operations and may change after discharge. We developed and temporally validated a two-stage calculator to estimate PE risk at discharge and update risk after unplanned readmission or reoperation.
Study Design:
ACS-NSQIP adult operations from 2020–2023 were used for development (n=3,864,605), with 2024 operations reserved for temporal validation (n=956,434). The primary outcome was 30-day postoperative PE. Penalized logistic regression modeled discharge risk using patient, surgical, and early postoperative variables; dynamic reassessment added timing of first unplanned readmission and reoperation. A prespecified 0.5% predicted-risk threshold was evaluated.
Results:
Among 4,856,597 operations, 17,045 patients developed PE (0.35%). In temporal validation, the discharge model demonstrated AUC 0.811 (95% CI, 0.801–0.818), calibration slope 1.024, and intercept 0.036. Dynamic reassessment improved discrimination to AUC 0.892 (95% CI, 0.887–0.897), with calibration slope 0.991 and intercept 0.030. Among 57,840 patients with an unplanned-return event, reassessment moved 24,320 patients above the 0.5% treatment-consideration threshold despite being below threshold at discharge. This newly flagged group had 711 observed PEs, a 2.92% PE rate, and 1 PE within 30 days for every 34.2 newly flagged patients under usual care. Sensitivity analyses indicated that unplanned-return occurrence, rather than exact postoperative day, was the durable risk signal.
Conclusions:
Across adult surgery, dynamic PE reassessment after unplanned readmission or reoperation identifies an actionable high-risk postoperative state not captured at discharge. This surgeon-facing calculator can make extended-prophylaxis reconsideration reproducible when prevention remains possible.