CURB-65 score for predicting deep vein thrombosis in severe community-acquired pneumonia: a cohort study

Abstract

Patients with severe community-acquired pneumonia (SCAP) are at significantly increased risk for venous thromboembolism (VTE). Traditional VTE risk assessment tools like the Caprini and Padua prediction scores are often underutilized in acute care settings due to their complexity. The CURB-65 score, a routine severity assessment tool for pneumonia, may offer a simplified and integrated approach for VTE risk stratification, but its predictive value for deep vein thrombosis (DVT) in SCAP patients requires validation. We retrospectively enrolled 766 adults with SCAP admitted to Beijing Chao-Yang Hospital between January 1, 2015 and August 31, 2024. Baseline characteristics, CURB-65, Caprini, and Padua prediction scores, VTE prophylaxis details, and the occurrence of ultrasound confirmed in-hospital DVT were collected. Multivariable logistic regression was employed to identify independent risk factors for DVT. We further performed a sensitivity analysis restricted to proximal DVT. The predictive performance of the scores was compared using the area under the receiver operating characteristic curve (AUC). Patients were stratified into low-risk (CURB-65 < 3) and high-risk (CURB-65 ≥ 3) groups for further analysis, including Kaplan-Meier survival analysis to compare cumulative DVT incidence. A separate multivariable analysis was conducted in the subgroup of patients who received VTE prophylaxis. The overall incidence of DVT was 30.3% (232/766), with the clinically actionable proximal DVT accounting for 3.0% (23/766) and distal DVT comprising 27.3% (209/766). Multivariable analysis identified a higher CURB-65 score as an independent predictor of DVT (adjusted Odds Ratio [OR] = 1.936, 95% Confidence Interval [CI]: 1.503–2.495, P < 0.001). The sensitivity analysis limited to proximal DVT demonstrated that CURB-65 remained an independent predictor (adjusted OR = 1.759, 95%CI:1.431–2.162, P < 0.001). The predictive performance of the CURB-65 score (AUC = 0.717, 95% CI: 0.684–0.749) was statistically higher than that of the Caprini prediction score (AUC = 0.648; raw P = 0.002 for these two curves) and comparable to the Padua prediction score (AUC = 0.680; raw P = 0.052 for these two curves). The high-risk group (CURB-65 ≥ 3, n = 391) had a significantly higher DVT incidence (47.3%) compared to the low-risk group (12.5%, P < 0.001), with Kaplan-Meier curves confirming a higher cumulative incidence (Log-rank P < 0.001). In the prophylaxis subgroup ( n = 626), the CURB-65 score remained an independent risk factor for DVT (adjusted OR = 1.953, 95% CI: 1.488–2.562, P < 0.001), and the overall DVT incidence was 30.8% (193/626). The CURB-65 score shows moderate performance in predicting DVT risk among SCAP patients, and retains stable predictive capacity for clinically critical proximal DVT, with discriminative capacity comparable to dedicated VTE assessment tools, though all three scoring systems deliver only fair-to-moderate discrimination. Routine calculation at admission may provide auxiliary clues to preliminarily screen patients with elevated residual DVT risk and facilitate risk-stratified monitoring and individualized thromboprophylaxis. Patients with CURB-65 ≥ 3 remain at substantial residual DVT risk despite standard prophylaxis, supporting the need for future research to optimize preventive strategies in this high-risk population. In addition to pneumonia severity assessment, the CURB-65 score may serve as a convenient auxiliary marker for DVT risk stratification.