Optimizing Preoperative Venous Thromboembolism Risk Assessment in Elderly Hip Fracture Patients: A Refined Caprini Model Integrating D-Dimer and Injury-to-Admission Time

Background: The Caprini risk assessment model (RAM) is widely used to evaluate venous thromboembolism (VTE) risk across diverse patient populations. However, the original risk stratifications may require modification for hip fracture patients to improve predictive accuracy. This study aimed to optimize VTE prediction by refining the Caprini model and integrating additional predictive factors. Methods: This retrospective cohort study included 1114 elderly hip fracture patients screened at Peking Union Medical College Hospital between May 2012 and February 2023. A modified VTE prediction model was developed by integrating D-dimer levels and injury-to-admission time into the Caprini RAM. The sensitivity, specificity, and the area under the curve (AUC) of the model were determined to assess its predictive performance. Results: The revised Caprini model effectively stratified patients into three VTE risk levels on the basis of their scores: 14.0% for ≤9, 24.0% for 10–11, and 39.4% for ≥12. Both the Caprini score and risk level demonstrated better predictive ability than D-dimer did (AUCs: 0.606 and 0.614 vs 0.552). Among patients admitted within one day after injury, the D-dimer levels were significantly higher in VTE patients compared with non-VTE patients from day 3 onwards (p = .012). Incorporating the injury-to-admission time with the Caprini score further improved the AUC from 0.614 to 0.649. Among models maintaining the sensitivity and negative predictive value above 90%, this combination model demonstrated the best performance. In contrast, for patients with delayed admission (>1 day), D-dimer levels were significantly higher in those with VTE than in those without VTE (p < .001). Combining the Caprini score with the D-dimer level provided a more accurate prediction (AUC: 0.681), significantly outperforming the Caprini score alone (AUC: 0.552). When the sensitivity exceeded 95%, the D-dimer threshold model across the three Caprini risk levels outperformed the Caprini risk levels combined with the D-dimer level (AUC: 0.592 vs 0.571) and demonstrated a significant advantage at 100% sensitivity. Conclusions: Reclassifying the Caprini score into three risk levels may improve preoperative VTE stratification in elderly hip fracture patients. Integrating the injury-to-admission time and D-dimer level into the Caprini model was associated with improved predictive performance. Standardizing the timing of D-dimer sampling may reduce timing-related variability in VTE assessment. External multicentre validation is warranted.