Patients with tibial plateau fractures face inherent inconvenience in the diagnosis of deep vein thrombosis (DVT) relying on lower limb venous ultrasound. We grapple with the challenge that the demand for lower limb venous ultrasound in fracture patients may arise at any time, potentially necessitating repetitive examinations. Moreover, the training and logistical aspects of ultrasound professionals pose challenges[12]. These challenges render us unable to implement timely preventive measures for certain patients. Consequently, a need for objective and reliable predictive modeling arises, avoiding arbitrary interventions. A predictive model, formed by integrating various risk factors and patient-specific features using a random forest algorithm, enables clinical practitioners to early identify individuals with a higher preoperative deep vein thrombosis (DVT) risk during the treatment process. This prospective prediction enhances the accuracy and timeliness of intervention strategies, ultimately improving the prognosis of preoperatively DVT-susceptible individuals with tibial plateau fractures. This study employed the random forest algorithm. According to existing research, the random forest regression is less prone to overfitting compared to traditional regression methods, ensuring robust predictions[13]. The algorithm demonstrates advantages in handling data with missing values and complex relationships. The ensemble learning approach reduces the risk of bias in individual decision trees, and out-of-bag (OOB) data provides internal validation. Based on previous research findings, the reported incidence of preoperative deep vein thrombosis (DVT) in tibial plateau fracture patients ranges from 17.3–36.43%[14, 15]. Additionally, Zhang et al.'s study reported a DVT incidence of 42.0% in patients with traumatic knee periarticular fractures[16]. This study employed the random forest algorithm to discern several pivotal clinical variables, including the elderly, age, FDP, D-dimer, RBC, RDW-CV, ALB, AGR, hypoproteinemia, other extremity fractures, spinal injury, and time from injury to admission.
This study encompassed adults of all age groups, highlighting the susceptibility of elderly patients to DVT after trauma or surgery. Elderly patients experience restricted lower limb activity, poor venous valve function, diminished vascular elasticity, often coupled with underlying health conditions. Moreover, the post-fracture hypercoagulable state further promotes the formation of DVT[17]. In a multicenter retrospective study conducted by Zhang et al.[16], age emerged as an independent risk factor influencing the formation of DVT in patients with traumatic fractures, with individuals aged 50 and above exhibiting an increased susceptibility to DVT. Additionally, Li et al.[18] conducted a prospective cohort study on patients with tibial plateau fractures, investigating the incidence and location of lower limb DVT. The study revealed that age (≥ 41 years) independently correlated as a significant risk factor, markedly increasing the likelihood of DVT occurrence. In a study on preoperative deep vein thrombosis (DVT) risk factors for lower limb fractures, Hu et al.'s meta-analysis revealed that the incidence of preoperative DVT in patients aged ≥ 50 years and < 50 years was 24.9% and 23.1%, respectively[19]. Subsequently, a chi-square test was conducted to assess the differences. The test results indicated a statistically significant difference between age groups, with higher preoperative DVT prevalence in elderly patients with lower limb fractures compared to younger patients. Chang et al.'s study also demonstrates that in patients with lower limb fractures, the incidence of deep vein thrombosis (DVT) reaches its peak when the age exceeds 60, with a probability of 6.04%[20]. In summary, these studies reflect an association between advanced age (especially 60 years and above) and increased susceptibility to DVT formation in lower limb fractures, aligning with the findings of our study.
We aspire to enhance the predictive capability of deep vein thrombosis (DVT) formation through readily accessible and cost-effective blood test indicators following patient admission and before surgery. The blood test indicator data incorporated in this study underwent univariate and multivariate analyses, along with AUC-based prioritization. These blood test indicators were derived from multiple examinations conducted within 48 hours preceding the patient's surgery, either by averaging results from multiple tests or directly inputting single test outcomes.
Fibrin degradation products (FDP), arising from the breakdown of fibrin after dissolution, serve as indicators reflecting the endogenous fibrinolysis activity within the body[21, 22]. Elevated FDP levels following thrombotic events hinder the physiological characteristics of both fibrin and platelets. Yang et al.[23] conducted a risk factor analysis for deep vein thrombosis (DVT) in patients with lower limb fractures within the acute compartment syndrome (ACS) population. The study identified FDP as an independent risk factor for DVT in patients with ACS and concomitant lower limb fractures. D-dimer, a prominent subtype of soluble fibrin degradation products, typically reflects the systemic coagulation and fibrinolytic activation status within the body, indicating potential thrombus formation. In this study, D-dimer was identified as a crucial variable, aligning with findings by Hu et al.[24], who discovered that D-dimer levels were an independent risk factor for DVT. Drawing from Yang et al.'s research on preoperative deep vein thrombosis in patients with femoral shaft fractures, both study results and multivariate analyses consistently highlight D-dimer levels as an independent and significantly meaningful risk factor for DVT occurrence, distinguishing between the DVT and non-DVT groups[25]. Fascinatingly, Zhang et al.[16] conducted a multicenter retrospective study examining the formation of deep venous thrombosis (DVT) in trauma patients with fractures upon admission. Their research underscores the significance of plasma D-dimer levels upon admission as a crucial hematological parameter for predicting the occurrence of DVT in fractures below and around the knee (with sensitivity and specificity of 0.861 and 0.678, respectively). Furthermore, they elucidate that the value of D-dimer levels in patients with fractures above the knee is lower compared to fractures below and around the knee. Zhu et al.[26] conducted a study on the preoperative formation of deep vein thrombosis (DVT) in patients with tibial plateau fractures, proposing that 86.4% of DVT occurrences happen within the first 7 days after the fracture. Elevated levels of d-dimer (> 0.5 mg/L) were identified as a correlated factor in the formation of preoperative DVT.
Red Blood Cell (RBC) count serves as a quantification of the erythrocytes within a sample. This count, along with Hemoglobin concentration (Hb), Hematocrit (Hct), and related parameters, proves instrumental in the diagnosis and categorization of anemia[27]. An insufficiency of an adequate number of red blood cells, erythrocytes, or a decrease in the total hemoglobin content results in a diminished oxygen transport to tissues[28]. Red Cell Distribution Width (RDW-CV) is employed to gauge variations in the size of red blood cells. An elevation in RDW anticipates disruption in the homeostasis of red blood cells, which can occur during erythrocyte generation or their lifespan[29]. In light of the previously established significant association between elevated Red Cell Distribution Width (RDW) and acute pulmonary embolism, Cay N and colleagues undertook a study examining the relationship between RDW elevation and the formation of acute deep vein thrombosis (DVT)[30]. The research findings indicate that when RDW surpasses 14.9%, the specificity for predicting proximal DVT is 73%, with a sensitivity of 85%. The risk of incidence elevates by a factor of 12. Furthermore, the risk of DVT occurrence continues to increase with the continued augmentation of RDW. Further studies have reported a significant association between elevated Red Cell Distribution Width (RDW-CV) and an increased overall mortality rate in non-anemic patients with hip fractures[31]. Xiong et al.[32] identified in a total joint arthroplasty (TJA) patient cohort that both RBC ≤ 3.92*10^9 /L and RDW-CV ≥ 13.2% independently contribute to the formation of preoperative deep vein thrombosis (DVT) in TJA. Additionally, Zhang et al.[33] reported in a cohort study on perioperative DVT formation in patients with bone trauma that preoperative anemia (hemoglobin threshold 125 g/L) serves as a valuable screening indicator for predicting perioperative DVT.
Albumin (ALB) possesses diverse functionalities, serving as a reflective indicator of the body's nutritional status, osmotic pressure maintenance, and substance transportation. In clinical practice, blood examination for albumin also provides insights into the functioning of the liver and kidneys. Reduced levels of albumin may suggest underlying conditions such as hepatic or renal diseases, malnutrition, or chronic ailments. When albumin decreases below 35 g/L, it leads to hypoalbuminemia, causing fatigue, weakness, and increased susceptibility to infections. The Albumin-Globulin Ratio (AGR) stands as a valuable metric in balancing albumin and globulin levels in the blood. A decreased AGR may indicate liver dysfunction, inflammation, or malnutrition, while an elevated ratio may signal dehydration or immune system dysregulation. As albumin levels decline, AGR inversion can occur. AGR inversion is associated with chronic inflammation, autoimmune diseases, and certain types of cancers. Ma et al.[34] conducted a multifactor logistic regression analysis on preoperative deep venous thrombosis (DVT) occurrence in patients with spinal fractures. Their findings revealed that serum albumin (ALB) levels below 35 g/L serve as an independent and significant risk factor. Similarly, Tan and colleagues focused on preoperative patients with closed patellar fractures, discovering a substantial correlation between low albumin levels (< 35 g/L) and a higher incidence of preoperative DVT[35]. Furthermore, they identified low albumin as an independent risk factor in the analysis of DVT occurrence. In a retrospective study on patients undergoing total joint arthroplasty (TJA), Xiong et al.[36] emerged as trailblazers, being the first to establish that serum albumin levels < 37.2 g/L serve as an independent predictive factor for preoperative occurrence of deep vein thrombosis (DVT) in patients. This retrospective inquiry also delineated an escalating probability of preoperative DVT detection as albumin levels decreased, coupled with a concurrent increase in the likelihood of perioperative blood transfusion.
Patients with tibial plateau fractures, compounded by additional fractures in other limbs and spinal injuries, contribute to the heightened complexity of clinical treatment and care[37]. Individuals experiencing concomitant fractures in other limbs may encounter challenges in physical activities, exercise, and overall functionality. For patients with spinal cord injuries, surgical intervention is often probable, with the primary objective being the restoration of spinal stability and neural control regions. Instances where patients undergo secondary surgeries during hospitalization necessitate elevated preoperative preparation. The fixation of multiple sites in the lower limbs, when affected by fractures, induces alterations in hemodynamics. Trauma, encompassing fractures and severe injuries, amplifies the stimulation of blood stasis, entry of inflammatory substances into the bloodstream, and activation of endothelial cells during limb fixation[38, 39]. Ultimately, this cascade leads to coagulation activation and the formation of venous thrombosis[40]. Within the initial 24 hours post-trauma, the blood assumes a heightened coagulative state, persisting for a duration extending from 10 to 30 days following the traumatic incident[40, 41]. Dicks et al. [42] provide a comprehensive overview of risk factors associated with venous thrombosis, highlighting fractures in different lower limb regions as pivotal contributors to venous thromboembolism. However, variations in thrombotic incidence are evident across distinct anatomical sites, with the highest risks observed in hip fractures, tibial plateau fractures, and tibial shaft fractures, corresponding to rates of 16.6%, 16.3%, and 13.3%, respectively. Interestingly, the article remains silent on the impact of combining fractures in multiple limb regions on the occurrence of deep vein thrombosis (DVT). In a rare study investigating the risk factors for venous thromboembolism in patients with tibial plateau fractures, Wang et al.[5] identified concomitant spinal or limb injuries as clinically significant contributors to venous thromboembolism. Notably, tibial plateau fracture patients with associated spinal injuries demonstrated a more pronounced association.
In the establishment of a random forest prediction model for deep vein thrombosis (DVT) in tibial plateau fractures, the time elapsed from injury to hospital admission emerges as a pivotal risk factor influencing the occurrence of deep vein thrombosis[43, 44]. The impact of tibial plateau fractures, coupled with prolonged time from injury to admission, escalates the risk of DVT, underscoring the imperative for timely medical intervention[45]. Yang et al. [5] specifically elucidated the correlation between the extension from injury to surgery and the increased incidence of preoperative deep vein thrombosis (DVT). This manifests as a 12% rise in the corresponding preoperative DVT occurrence rate for each day of delay. Yang et al. further advocate for the surgical timing as the sole controllable risk factor within this study, emphasizing the need for clinical practitioners to accord it due consideration. In a study conducted by Takaomi et al.[46], the results of both unrefined and adjusted models, following the analysis through logistic regression, indicate a significant correlation between the prolonged period known as "the time from injury to surgery" (TFITS) and a higher incidence rate of preoperative deep vein thrombosis (DVT). The TFITS intervals of 0–2 days and > 7 days correspond to preoperative DVT incidence rates of 8.3% and 36.1%, respectively. Controlling for other influencing factors, for lower limb fracture patients, each additional day in TFITS is associated with a 9.3% increase in the preoperative DVT occurrence rate.
This study is not without several limitations. Firstly, being a retrospective study, it inherently possesses the unavoidable drawbacks commonly associated with retrospective research. Secondly, while we conducted a multivariate analysis, it was not feasible to encompass all potential risk factors for deep venous thrombosis (DVT) formation. Thirdly, we attempted to derive averages from blood test indicators taken within a 48-hour interval, yet the process of blood collection and analysis remains susceptible to numerous uncontrollable factors. Lastly, the patients admitted to our hospital with tibial plateau fractures are generally complex cases, often comprising individuals referred from lower-tier medical facilities. The incidence rate of DVT may potentially be higher than the overall average, and the number of patients in our tibial plateau fracture cohort is relatively limited, rendering the patient cohort insufficient.