In the present study, six independent risk factors were identified for predictive model development of DVT in TPF patients, which included combined femoral fractures, age ≥ 40 years, BMI ≥ 24 kg/m2, ISS, iCa-min, and FIB at admission, and these findings were in accordance with previous reports in general.

The risk of DVT for patients with fracture varies with different fracture location, and the patients with pelvic and femoral fractures usually undertook higher risk of DVT than those with fractures in other parts of the body [11]. The three predominant independent risk factors for DVT were slow blood flow, hypercoagulation and endovascular injury [12], which would explain the increased risk of DVT for patients with combined femoral fractures in the present study. The femur is the longest and thickest tubular bone in human body with high strength and toughness, so the energy of violence causing femoral fractures is usually higher than that causing fractures in other parts of the body. In addition, the intensive zone of blood vessels locates at the distal end of femur, and is easily involved in the endothelial injury of blood vessels when traumas occur. Meanwhile, the venous blood stasis caused by post-traumatic sickbed and immobilization, venous reflux influenced by the edema related to tissues injury and inflammation, and the imbalance between blood coagulation and fibrinolysis systems caused by trauma, synergistically contribute to the post-trauma hypercoagulable state to be prone to DVT [13, 14].

The present study indicated that ISS at admission was an independent risk factor for DVT in TPF patients, which is commonly adopted to evaluate the severity of traumatic patients in clinical practice [15]. The higher ISS is, the more severe a trauma is, and the more likely severe internal environment and coagulation disorder appear [16]. As a result, a large number of inflammatory factors would release into the blood, extensive vessel endothelial injury would occur, and subsequently the progression of DVT would be promoted [17].

The change with age in human body, such as decreased vascular wall elasticity and venous valve function, increased endothelial damage, blood viscosity and procoagulant substances, muscle relaxation and decreased pumping of lower extremity muscles, will increase the risk of DVT after fractures in the elderly [18, 19]. Several studies indicated that older age is an independent risk factor for DVT [20, 21]. However, it was still controversial in the threshold of age for increased DVT risk. It was reported that age > 40 years was an independent risk factor of DVT [22], while another study indicated that patients with fractures over the age of 30 had a higher risk of DVT [23]. This study revealed that age ≥ 40 years was an independent risk factor for DVT in TPF patients.

The patients with overweight and obesity who usually suffer from hyperlipidemia and atherosclerosis remain in a chronic inflammatory state for a long time, and are prone to be in a hypercoagulable state after trauma [19, 24]. For these patients, a decrease in venous valve function and hemodynamic abnormalities may be induced by less physical activity, and lower extremity venous blood flow stasis may be exacerbated by higher intra-abdominal pressure [25,26,27]. Hence, all these factors will increase the risk of post-traumatic DVT in such patients. Currently, several studies had addressed that increased BMI was an important risk factor for DVT [28, 29]. The Chinese guidelines consider BMI ranging from 24.0 to 28 kg/m2 as overweight, and BMI no less than 28.0 kg/m2 as obesity [30]. Therefore, in this study, the research subjects were divided into two groups based on BMI of 24.0 kg/m2, and the results indicated that BMI no less than 24.0 kg/m2 was an independent risk factor for DVT in TPF patients, OR = 1.546, 95% CI 1.101–2.171, P = 0.012. Although people all over the world belong to Homo sapiens, there are still some subtle differences in physical constitutions among people in different regions. Therefore, it is necessary to adopt different BMI thresholds based on the actual situation in different countries.

This study found that iCa-min was an independent risk factor for DVT in TPF patients, OR = 0.009, 95% CI 0.003–0.033, which meant that the level of ionized calcium (iCa) might be inversely associated with DVT. The iCa, also called coagulation factor IV, is involved in almost all stages of the coagulation process and essential for the activation of thrombin and the conversion of prothrombin to thrombin [31,32,33]. In the endogenous coagulation pathway, iCa could assist in activating factor XI, together with activated factor VIII and factor IX, it activates factor X. In the exogenous coagulation pathway, iCa activates factor X together with factor III and factor VII. In the common pathway, together with activated factor V and factor X, it could convert FIB into fibrin monomers. In addition, iCa can assist in activating factor XIII and continue to assist factor XIII in converting soluble fibrin monomers into stable fibrin polymers. The hypocalcemia caused by the loss of blood components (including iCa) related to traumatic bleeding is associated with coagulation disorders, transfusion volume, and mortality [34]. When a large amount of blood is transfused for bleeding control, the hypocalcemia will be aggravated by the chelation between calcium and citrate which was adopted as anticoagulant in stock blood products [35], which might worsen trauma-induced coagulopathy (TIC) [36]. The main manifestation of TIC can be either hypocoagulable state with haemorrhage or hypercoagulable state with thromboembolism [37]. The hypercoagulable state of TIC might occur in the acute or late stages of trauma, and is triggered by complex mechanism such as stress-induced endothelial damage, tissue damage, inflammatory reactions and excessive release of procoagulant substances. In addition, TIC is often accompanied with high fibrinogen concentration, high platelet reactivity, decreased anticoagulant activity, and fibrinolysis inhibition in lab test [38]. Therefore, the hypercoagulable state of TIC might increase the risk of lower extremity DVT in TPF patients. However, the association among hypocalcemia, TIC and DVT is still to be further verified in the future.

Fibrinogen, also known as coagulation factor I, is a glycoprotein synthesized and secreted in liver cells. As a soluble fibrin precursor, FIB plays a significant role in the coagulation process, and its deficiency or dysfunction would lead to bleeding and thrombotic clinical events [39, 40]. FIB can promote platelet aggregation, as well as the growth, proliferation, and contraction of smooth muscle and endothelial cells, thereby increasing blood viscosity and peripheral resistance [41]. FIB also can accelerate thrombus formation by promoting collagen synthesis, chemotactic migration of monocytes and macrophages to the endometrium [42]. Several studies showed that FIB is associated with venous thrombosis in a concentration-dependent manner, and the early monitoring of FIB is a good predictor of DVT [43, 44]. This study found a positive association between FIB within the acute phase at admission and preoperative lower extremity DVT in TPF patients. Of course, further research is still needed to verify the impact of FIB levels on trauma prognosis.

As mentioned above, according to the six independent risk factors determined by Lasso and multivariate logistic regression analysis, we developed a nomogram for predicting the risk of DVT (DRNS model), in which each risk factor had a corresponding score, and the total score was obtained to predict the probability of DVT for TPF patients.

The DRNS model showed superior performance with C-index (0.748 and 0.920, respectively) in either training or validation set, indicating that the model had moderate discrimination in predicting the probability of DVT. The calibration of DRNS model was also well addressed by the strong consistency between actual and calibration curve in the training and validation sets, respectively, as well as the Hosmer-Lemeshow goodness-of-fit test.

According to the cut-off value of DRNS model determined by ROC curve, in the present study, the prospective population with TPF were divided into low- and high-risk groups of DVT progression. There was no statistical significance in the incidence of DVT progression observed between different administration frequencies (qd or q12h) of LMWH (P = 0.323) prophylactic treatment within low-risk group, whereas it was observed (P = 0.002) within high-risk group. Due to the concern of bleeding in clinical practice, we would like to recommend individual administration frequency of LMWH prophylactic treatment for TPF patients according to the risk, which is predicted by some tools such as DRNS model.

Pelvic fracture patients are a high-risk group for DVT, and the usual intervention measures for this group include LMWH anticoagulant therapy and CUS examination. However, anticoagulant therapy and CUS examination will incur certain medical expenses. The CUS examination process may aggravate the pain at the fracture site of the patients. There will be certain medical risks during the transportation when leaving EICU. And anticoagulant therapy may also increase the risk of bleeding in important organs of the patients. Therefore, the pros and cons should be carefully weighed before anticoagulant therapy and CUS examination are carried out for patients with pelvic fractures. Of course, whether it could reduce the risks and costs in the process of DVT examination and prevention, by using the DRNS model for quantitative analysis of the risk of DVT in the patients, prospective multicenter large sample studies are still needed.

There are several limitations in this study. First, it was a single-center study, and the DRNS model was developed in retrospective population. Second, although CUS has gradually replaced venous angiography and been widely used, this method is not the “golden standard” to detect DVT. Third, the sample size of the prospective cohort was small. Fourth, the controversy still existed in some risk factors for DVT which varied in the present and previous studies.