Spinal fractures account for approximately 5% to 6% of all fractures (1), with the most common cause being high-energy trauma events. During the perioperative management of patients with spinal fractures, the risk of developing deep vein thrombosis (DVT) in the lower limbs shows a significant upward trend. DVT formation poses significant risks, potentially leading to delayed surgical treatment in the early stages, thereby prolonging the patient's hospital stay and significantly increasing medical costs. More seriously, if DVT is not managed promptly and effectively, it may progress to life-threatening pulmonary embolism (PE) (2). According to literature reports, the overall incidence rate of DVT following spinal surgery ranges from 0.2% to 31% (3). Based on the analysis of collected data, proactive prevention strategies have demonstrated significant clinical benefits in the prevention of DVT during the perioperative period in patients with spinal fractures. Although current orthopedic research on DVT in major surgeries has preliminarily identified some common risk factors (4), there remains a significant gap in research focused on specific risk indicators and targeted intervention measures during the perioperative period in patients with spinal fractures. First, there is a lack of risk indicators that account for dynamic changes during the perioperative period. Existing studies primarily focus on preoperative static factors (such as age, D-dimer levels, comorbidities, etc.), while neglecting the impact of perioperative dynamic indicators (such as surgical approach, intraoperative blood loss, postoperative bedrest duration, timing of anticoagulant initiation, etc.) on DVT occurrence (5); Second, there is no specific risk warning model established for the Chinese population. Existing DVT risk models (such as the Risk Assessment Profile) are primarily designed for Western trauma populations, and some key variables (such as post-discharge follow-up data) are difficult to obtain in real time in clinical practice, making them unsuitable for guiding preoperative decision-making (6). The incidence of DVT in Asian populations is significantly lower than in Western populations, and directly applying Western models may underestimate or overestimate actual risks (7). In particular, there is currently no multi-factor-defined, accurate, and targeted DVT risk warning model. This study employed a retrospective analysis method, systematically collected clinical case data from patients undergoing spinal fracture surgery. Using univariate screening and logistic multivariate regression models, we investigated the significant independent risk factors for DVT formation after spinal fracture surgery. Based on this, we constructed a DVT risk warning model based on logistic risk regression using standardized modeling formulas to achieve rapid and accurate screening of the risk of lower extremity DVT in patients undergoing spinal fracture surgery. The specific findings are reported as follows:

The results of this study revealed that the incidence of perioperative lower extremity deep vein thrombosis in patients with spinal fractures was 17.46%, which is an approximate or slightly lower incidence than the epidemiological data reported in the literature (0.2%–31%) (9, 10). However, this result still suggests that the potential risk of perioperative DVT in such patients needs to be given due attention in the clinic. Notably, perioperative hemodynamic and venous haemodynamic abnormalities in patients with thoracolumbar spine fractures caused by high-energy injuries are more pronounced because of the complex mechanism of trauma and the high demand for tissue repair, further exacerbating the formation of DVTs. In clinical therapeutic interventions for spinal fractures, establishing a risk warning model to quickly and accurately assess the risk of DVT and provide patients with individualized preventive and therapeutic measures is an urgent problem. We conducted a retrospective study of patients admitted for spinal fracture surgery in recent years to systematically assess a series of potential high-risk factors associated with postoperative DVT formation. Univariate and logistic multifactorial analyses were performed on all suspected high-risk factors closely related to perioperative DVT, based on which a risk warning system was constructed via logistic regression modeling to provide a more precise and effective strategy for preventing DVT in major orthopedic surgeries.

In this study, a history of perioperative blood transfusion, high levels of C-reactive protein (CRP), D-dimer concentrations >500 μg/L, a history of hypertension, age ≥60 years, and prolonged bed rest were identified by univariate and multivariate logistic regression analyses as independent risk factors for perioperative lower-extremity DVT in patients with spinal fractures. Previous studies have shown that perioperative blood transfusion during surgery is strongly associated with thrombosis (11, 12). A multicenter study conducted by a national scholar (13) confirmed that intraoperative and/or postoperative blood transfusion, as well as postoperative blood transfusion alone, were significantly associated with the development of DVT in patients with traumatic spinal fractures, and the results of the present study are consistent with this finding. The mechanisms by which perioperative blood transfusion increases the risk of DVT may involve a hypercoagulable state of the blood, vascular endothelial injury, and abnormalities in immune regulation (14). When red blood cells (RBCs) or plasma are transfused, coagulation factors in blood products may lead to increased erythrocyte aggregation and decreased cell membrane deformity, thereby triggering a hypercoagulable state in the blood (15). In addition, the transfusion process may trigger an inflammatory response, leading to vascular endothelial damage and hemodynamic alterations. It may affect the body's immunomodulatory function, increasing the risk of thrombosis. Findings have shown that C-reactive protein (CRP), an acute temporal response protein with low serum levels in healthy individuals, is significantly elevated to reflect the presence of a systemic sensitively, but not specific, inflammatory response state in the organism. Studies have shown (16) that the expression level of CRP is significantly correlated with the degree of activity of the inflammatory response, especially in the state of trauma or infection, and its dynamic changes can provide an objective basis for the quantitative assessment of the inflammatory state. Previous retrospective cohort studies have shown that the inflammatory response is one of the most important factors in thrombosis and can be regarded as a chronic, low-level sterile inflammatory process (17). Individuals with high C-reactive protein levels have a significantly increased risk of developing DVT during the perioperative period, suggesting that C-reactive protein (CRP) stimulates the production of tissue factor (TF) by monocytes, which directly triggers abnormalities in coagulation mechanisms and promotes platelet adhesion and aggregation, ultimately leading to thrombosis. In addition, elevated CRP levels are positively correlated with the degree of complement system activation, further exacerbating vascular endothelial damage and promoting thrombus formation and development (18–20). Therefore, continuous monitoring of C-reactive protein concentration is highly valuable for the early identification of patients at high risk of DVT, especially those undergoing different orthopedic surgery.

As a specific marker of the fibrinolytic process, an increase in D-dimer concentration directly reflects the presence of hypercoagulability and the subsequent enhancement of fibrinolytic activity in vivo, which is closely related to thrombosis and its induced dysregulation of the dynamic balance between coagulation and fibrinolysis (21). According to the available literature (22), activation of the fibrinolytic system is triggered when a thrombus is present in the body, a process in which cross-linked fibrin is broken down to produce derivatives such as D-dimer. Hui et al. (23) reported that the sensitivity and specificity for predicting DVT with an elevated D-dimer level on day 3 after spinal surgery were 72.7% and 76.5%, respectively, and the D-dimer cut-off level was 5.82 μg/ml. Joseph et al. (24) reported that surgery increased D-dimer levels in patients with spinal fractures combined with spinal cord injury. When D-dimer concentrations were ≥2 times the normal's upper limit, their DVT risk was 3.5 times greater than that of patients with typical D-dimer values. The present study revealed that D-dimer levels fluctuate and increase after surgery in patients with spinal fractures. Elevated D-dimer levels can reflect the status of thrombosis in the body. A D-dimer level >500 μg/L may indicate the presence of DVT, and the results of the present study are basically in agreement with those of the above studies. This study confirmed that there is a close relationship between a history of hypertension and the risk of DVT in patients undergoing spinal fracture surgery. Most scholars believe prolonged elevated blood pressure damages vascular endothelial cells, an important factor in developing DVT (25). Second, hypertension is often associated with increased red blood cell aggregation and increased plasma viscosity, coupled with decreased vascular elasticity and luminal narrowing, which trigger the body's coagulation cascade response, which in turn significantly elevates the risk of DVT (26). The present study revealed that age greater than or equal to sixty years and prolonged bed rest were independent risk factors for DVT in patients with spinal fractures and that these factors can be used together as indicators to assess the risk of DVT. Prolonged bed rest increases the risk of DVT after orthopedic surgery. Reduced activity after orthopedic surgery, decreased cardiac output, decreased pump function of the lower limb muscles, increased resistance to venous blood return in the lower limbs, and activation of the coagulation state are more likely to lead to DVT (27). Spinal fractures are unique compared with other fractures, especially when combined with spinal cord and nerve damage, sensory and motor dysfunction of the lower limbs, and prolonged bed rest is more common in these patients after surgery, which can further exacerbate venous blood flow stagnation and weaken muscle pump function, and therefore the incidence of DVT is even greater. Previous studies have established advanced age as a risk factor for DVT in patients undergoing major orthopedic surgery and noted that this risk tends to increase with age (28). This is mainly because changes in vascular fragility are positively correlated with age, and the older the patient is, the poorer the cardiac function, the greater the fibrin activity, the greater the platelet aggregation, and the more severe the intima-media damage and blood flow abnormalities in his/her vessels (29). In addition, advanced age usually coexists with other DVT risk factors (e.g., obesity, hypertension, and prolonged bed rest), and there is a synergistic effect between these risk factors, which can further increase the risk of DVT.

In this study, based on the results of a multivariate regression analysis model (OR values and their influence weights), an early warning model for early risk identification of perioperative DVT in patients undergoing spinal fracture surgery (incorporating a total of six independent risk factors) was further developed through data processing with logistic regression formulas. Using the Hosmer-Lemeshow test, we validated the predictive efficacy of this predictive model in terms of the number of perioperative DVT cases, and the results showed consistency between its predictive values and the actual number of observed cases. A good fit (χ² = 12.139, P = 0.807) suggests that the early warning ability of the model has some credibility and is clinically beneficial. This study further validated the logistic regression model internally and externally. The AUC, sensitivity, and specificity of the model for predicting the perioperative occurrence of DVT in patients who underwent spinal fracture surgery in the model group, as plotted by the subjects' operating characteristic (ROC) curves, were 0.75%, 82.75%, and 75.9%, respectively. The validation group's values were 0.75%, 82.75%, and 75.9%, respectively. Moreover, specificity was 0.81, 72.73%, and 84.72%, respectively, and the best diagnostic value was obtained according to the maximum principle of Jordon's index, with a p-value of 0.075, that is, when the predictive probability was ≥0.075, suggesting that perioperative patients are at risk of developing DVT. The above data revealed that the sensitivity and specificity were significantly more significant than the threshold of 70% in the constructed model and validation phase, which indicated that the developed model demonstrated a strong differentiation ability to identify risk factors and that the probability of appearing against false-negative and false-positive cases was low. These results suggested that the early warning model had good clinical predictive efficacy and applicability, which is beneficial for clinical operation and application.

In this study, unifactorial and multifactorial logistic regression analyses were used to identify and quantify risk variables that contribute significantly and independently to the development of DVT as a clinical event in patients undergoing spinal fracture surgery during postoperative recovery. The model considers a variety of factors that may influence the development of DVT, including a history of blood transfusion, medical history, age, coagulation, and inflammatory markers. The study's results revealed that the constructed early warning model had a high AUC in both the model and validation groups, and the evaluation indices, such as sensitivity and specificity, were within the acceptable range. This study suggested that the model demonstrated excellent predictive performance and accurately quantified the likelihood of patients undergoing spinal fracture surgery to develop lower-limb DVT, thus providing strong support for decision-making in clinical practice. Notably, although this study achieved its intended goals and produced valuable findings, its results are subject to several limitations that need to be fully considered when interpreting the research results are interpreted. For example, the sample sizes included in the retrospective analyses tend to be more limited, which may limit the ability of the constructed models to be generalized, that is, the efficacy of the model's performance in the face of unseen data. Notably, there is significant heterogeneity in treatment protocols and patient composition across different healthcare organizations. This implies that the existing models' scope of application and generalizability must be more fully explored and confirmed through subsequent studies. In the future, we will continue to monitor the research progress in this area and continuously improve and update such risk warning models.