Analysis of High-Risk Factors and Construction of a Nomogram Predictive Model for Deep Venous Thrombosis in Pelvic and Acetabular Fracture Patients Treated Conservatively.

Abstract

This study aims to develop a predictive nomogram model to assist physicians in making evidence-based decisions and potentially reduce the incidence of deep venous thrombosis (DVT). We conducted a retrospective study, including patients admitted to the hospital from January 2014 to January 2022 with a closed, single pelvic or acetabular fracture. Comprehensive data were collected for each patient, encompassing demographics, injury characteristics, comorbidities, and results from laboratory tests and lower extremity ultrasounds. Potential risk factors were identified by univariate and multivariate logistic regression analyses. The predictive model was constructed and then internally validated. Calibration accuracy was assessed using a calibration slope and the Hosmer-Lemeshow goodness-of-fit test. The discrimination of the nomogram model was evaluated using the C-statistic. Out of 232 individuals who underwent conservative treatment, 57 (24.6%) were classified into the DVT group and 175 (75.4%) into the non-DVT group based on lower extremity ultrasound findings. Predominantly, patients were aged between 41 and 65 in both groups. Body mass index (BMI) comparison showed that 54.29% (95/175) of the non-DVT group fell within the healthy weight range, while 45.61% (26/57) in the DVT group were overweight. Notably, the proportion of obesity in the DVT group was more than double that in the non-DVT group, indicating a higher DVT risk with increasing BMI (P=0.0215). Lower red blood cell (RBC) counts were observed in DVT patients compared to non-DVT ones (P<0.001). A similar pattern emerged for D-dimer, a marker for blood clot formation and dissolution, with significant differences noted (P=0.029). Multivariable analysis identified age, BMI, associated organ injury (AOI), American Society of Anesthesiologists score, hemoglobin (HGB), RBC, and D-dimer as candidate predictors. Significant variables included age (OR, 3.04; 95% CI, 1.76-5.26; P<0.001), BMI (OR, 1.97; 95% CI, 1.22-3.18; P=0.006), AOI (OR, 2.05; 95% CI, 1.07-3.95; P=0.031), and HGB (HR, 0.59; 95% CI, 0.39-0.88; P=0.010). The discrimination was 0.787, with a corrected c-index of 0.753. Calibration plots and the Hosmer-Lemeshow test indicated a good fit (P=0.7729). Decision curve analysis revealed a superior net clinical benefit when the predicted probability threshold ranged from 0.05 to 0.95. We developed a nomogram predictive model, and it could act as a practical tool in clinical workflows to assist physicians in making favorable medical decisions, which potentially reduces the incidence of DVT in those patients with pelvic and acetabular fractures treated conservatively.