2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
04:30 p.m. - 06:00 p.m.
MCCNO - 217-219
Technology and Procedures, Bioinformatics, Treatment Considerations, Biological Therapies, Gene Therapy, Non-Biological Therapies, Anticoagulant Drugs, Coagulant Drugs, Workforce, Clinical Practice (Health Services and Quality), Clinical Guidelines, Health Economics, Patient Safety, Quality Improvement, Quality of Care, Systems-Based Hematology, Value, Diseases, Bleeding and Clotting, Hemophilia, Thromboembolism, Thrombotic Disorders, Research, Clinical Research, Clinical Trials, Epidemiology, Health Outcomes Research, Health Disparities Research, Study Population, Human, Adult, Pediatric
PAPER

1071 Independent external validation of a race-aware adjustment to the caprini risk assessment model for venous thromboembolism in Asian patients

Ryan C Sun, MD, Andrea McDowell, MS2, Subhash Kolar Rajanna, PhD3, Michael L Berbaum, PhD4, Margaret E Wright, PhD5, Jozef Madzo, PhD6, Chad Robichaux, MPH7, Ahmad Nassar, MBBS8, John Gerard Quigley, MD9
1 University of Illinois, Chicago, Chicago, Illinois, United States
2 Rush Medical College, Chicago, Illinois, United States
3 University of Illinois, Chicago, Chicago, Illinois, United States
4 University of Illinois, Chicago, Chicago, Illinois, United States
5 University of Illinois, Chicago, Chicago, Illinois, United States
6 The Wistar Institute, Philadelphia, Pennsylvania, United States
7 Emory University School of Medicine, Atlanta, Georgia, United States
8 Emory University School of Medicine, Atlanta, Georgia, United States
9 University of Illinois, Chicago, Chicago, Illinois, United States

Introduction

Hospital-associated venous thromboembolism (VTE) affects 350,000 US patients annually and remains the leading cause of preventable inpatient death. Risk assessment models (RAMs) such as the Caprini score are clinical decision tools to guide prophylaxis strategies. Although chemoprophylaxis significantly reduces VTE risk, use is associated with risk of major bleeding (PMID 42138924) and society guidelines emphasize a trade-off between thrombosis and bleeding risks when selecting prophylaxis strategies (PMID 22315261).

The validation cohort studies for the Caprini RAM comprised a predominantly White/Caucasian population, potentially driving VTE risk stratification errors when used in racially diverse patient populations. We previously reported findings suggesting systemic VTE risk misassignment in Black and Latino patients in both medical and surgical cohorts at a single academic medical center. Here, we investigate Caprini RAM VTE risk misclassification in Asian medical inpatients and assess the generalizability of our proposed recalibration method.

Methods

We conducted a two-center retrospective cohort study of adult medical inpatients at University of Illinois Hospital (n=56,617; 2010-2023) for model derivation, and Emory University Hospital (n=92,932; 2010-2024) for independent external model validation. The primary outcome was a new diagnosis of acute VTE, as defined by a diagnostic ICD-9/10 code. The main study variable was self-reported Asian race. We harmonized the collection process to independently extract Caprini RAM components at study sites.

We assessed race-aware Caprini (RA-Caprini) scores by stepwise point subtraction for Asian patients and evaluated VTE risk estimation by multivariable logistic regression with Asian race as a covariate. We determined optimal score subtraction for Asian patients exclusively within the derivation cohort, then applied the locked RA-Caprini score subtraction parameter to the external validation cohort without re-weighting or modification. We compared distinct logistic regression models predicting VTE using Caprini vs RA-Caprini RAMs at both sites for discrimination (AUC) and calibration (LOESS curves).

Results

In the derivation and validation cohorts, baseline characteristics were comparable: respectively, median age 54 (IQR 39–65) vs 68 (IQR 56–78), 51% vs 49% female, 3.1% (n=1,765) vs 3.4% (n=3,154) Asian. Respective full-cohort VTE incidences (including distal DVT) were 4.5% (n=2,544) vs 6.3% (n=5,903), and 2.9% (n=51) vs 4.9% (n=155) among Asian patients.

The Caprini RAM overestimated VTE risk in Asian patients for both the derivation and validation cohorts, with respective VTE OR 0.63 (95% CI 0.48–0.84, P<0.001) vs 0.74 (0.68–0.81, P<0.001). We found the optimized RA-Caprini score subtraction for Asian patients to be −2 points in the derivation cohort with a RA-Caprini VTE OR of 0.95 (0.82–1.10, P=0.7), replicated in the validation cohort with VTE OR 0.97 (0.89–1.05, P=0.7).

Implementing the Asian-specific RA-Caprini point subtraction improved VTE risk discrimination across the full cohort at both the validation site (AUC 0.621 vs 0.614) and the derivation site (AUC 0.701 vs 0.700) compared to the Caprini RAM. In the Asian patient subgroup, application of RA-Caprini markedly reduced baseline VTE risk overestimation in the validation cohort (calibration intercept 0.08 vs -0.40, slope 1.04 vs 0.96) and decreased mean absolute VTE risk estimation error to 1.0% (Eavg 0.010 vs 0.016, 37% reduction), mirroring the improvement seen in the derivation cohort (Eavg 0.010 vs 0.017, 41% reduction).

Notably, relative to the Caprini RAM, the RA-Caprini would downgrade 31% (n=555) of Asian patients to a risk level no longer mandating chemoprophylaxis per institutional policy in the derivation cohort.

Conclusions

The Caprini RAM demonstrates a significant VTE risk overestimation for Asian medical inpatients across two independent cohorts. Our RA-Caprini score recalibration enhanced VTE risk discrimination and calibration, demonstrating cross-site generalizability. Given a lower baseline risk of VTE, Asian patients face a disproportionate risk of bleeding with chemoprophylaxis (PMID 34144805). Adoption of the RA-Caprini RAM may mitigate this risk. While the multifactorial biological mechanisms driving VTE risk heterogeneity remain under active investigation by various groups including ourselves, our parsimonious intervention may improve patient outcomes as a stopgap measure.

Ryan C Sun, MD: Janssen Biotech - Honoraria