Introduction

Venous thromboembolism (VTE) is one of the five most common blood vessel diseases [1], and is a common complication after surgery with an incidence rate ranging from 0.2 to 31.2% [2], including two stages of deep vein thrombosis (DVT) and pulmonary embolism (PE). DVT is more commonly seen in the lower extremities and often presents with swelling, pain, purplish-red skin coloration, elevated skin temperature, superficial vein dilation, and can even lead to disability in severe cases. The clinical presentation of PE depends on the extent and speed of vessel obstruction and the cardiopulmonary functional status. Mild cases may be asymptomatic, while severe cases may present with sudden onset of dyspnea, chest pain, hemoptysis, and even shock or death. Orthopedic patients are at high risk for VTE due to factors, such as surgical trauma, cast immobilization, long recovery time, and medication use. The incidence of VTE after spinal surgery ranges from 0.2 to 13.6% [3], and rates of pulmonary embolism range from 0.03 to 2.4% [4,5,6], and the mortality rate was as high as 0.34% [7]. It also prolongs the patient's hospitalization time and brings enormous economic and social burden, and has become an important public health problem. Therefore, the prevention of VTE after spinal surgery is of great significance in accelerating the recovery of spinal patients, reducing complications, and lowering the medical burden.

The nomogram is one of the most commonly used statistical methods in clinical prediction models. It not only has advantages such as simplicity, intuitiveness, and easy to operate, but also can visualize abstract and complex regression equations, making it more convenient to calculate the probability of risk factors. The nomogram has been proven to be more reliable than other systems and has therefore been recommended as a replacement or even a new standard [8, 9]. The nomogram can help clinicians to vividly show the patients their risk of develo** VTE, improving patient compliance. Additionally, it can guide doctors, nurses, and management staff to perform early diagnosis, early prevention, and early intervention of postoperative VTE, providing a theoretical basis for the perioperative management of spinal surgery patients [10]. Therefore, in this study, a clinical prediction model using nomogram was established to explore the independent risk factors for VTE after spinal surgery, and a risk assessment model was constructed to identify patients at risk of VTE after spinal surgery at an early stage, to manage these patients accurately and to allocate medical resources effectively.

Patients and methods

Study population and experimental design

We conducted a retrospective analysis of a patient cohort who underwent surgical treatment in the Department of Spinal Surgery at Qingdao University Affiliated Hospital from January 2015 to December 2020. Inclusion criteria: (1) Age ≥ 18 years; (2) Received spinal surgery treatment; (3) Postoperative hospital stay > 72 h; (4) Complete case information with sufficient data. Exclusion criteria: To be excluded if having any of the following: (1) Percutaneous vertebroplasty, nerve root blockade, microdiscectomy; (2) Combined pelvic or lower limb fractures; (3) Surgery within 3 months; (4) Preoperative diagnosis of lower limb DVT by lower limb vascular ultrasound; (5) PE diagnosed by CT pulmonary angiography before surgery; (6) History of thrombosis; (7) Comorbid with blood system diseases or long-term use of anticoagulants; (8) Missing or incomplete medical records. This study was approved by the institutional ethics committee, and informed consent was obtained from each patient.

Clinical outcomes and definitions

We defined the occurrence of lower limb DVT as the clinical outcome. Venous ultrasound is the gold standard for determining whether lower limb DVT has formed [11]. According to the protocol for lower limb vascular ultrasound examination, the filling state of the blood flow in each vein lumen and the presence or absence of thrombosis were determined [12]. If a thrombus was present, its location was recorded. Patients who did not undergo lower limb venous ultrasound examination during hospitalization were excluded from the analysis in this study.

Selection of predictors

Demographic characteristics included gender, age, body mass index (BMI), hypertension, diabetes, previous surgery (pre_surgery), bed > 3 days, smoke, drinking, cancer, and trauma. Laboratory examinations included white blood cell count (WBC), platelet count (PLT), hemoglobin (Hb), triglyceride (TG), serum total cholesterol (TC), D-dimer, prothrombin time (PT), thrombin time (TT), and activated partial thromboplastin time (APTT). Surgical indicators included location (cervical, thoracolumbar), duration, and blood loss. The above information was obtained from the electronic medical record.

Statistical analysis

A total of 2754 postoperative spinal patients was randomly divided into training and validation groups with a ratio of 7:3 using SPSS 22.0 software (SPSS Inc., Chicago, Illinois, USA). Continuous variables were described as mean ± standard deviation, while categorical variables were described as proportions. Statistical differences between means and proportions were confirmed using Student's t-test (for continuous variables) and Chi-Square Test (for categorical variables).

To determine the risk factors for VTE, we initially conducted univariate logistic regression analysis on training group to perform a preliminary screening of factors; subsequently, LASSO regression was also used to reduce high-dimensional data and identify the best predictive features and variables for VTE after spinal surgery [13], Variables with P values less than 0.05 were included in the multivariate logistic regression model to screen for independent risk factors. Finally, the “rms” and “regplot” packages from R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria) were used to visualize the results of the logistic regression analysis, creating a dynamic interactive nomogram; the accuracy, stability, discriminative ability and calibration of the model were evaluated by consistency index (C-index), receiver operating characteristic (ROC) curve, Hosmer–Lemeshow goodness-of-fit test and calibration curve for training group and verification group [14]. The clinical usefulness of the VTE nomogram was determined by quantifying the net benefit at different threshold probabilities in the spinal surgery cohort [15]. Except for individual labeling, all the definitions were statistically significant (P < 0.05).

Results

Clinical characteristics and univariate correlations with thrombosis-related risk factors

We collected data from 2754 patients who received surgical treatment for spinal disorders at Qingdao University Affiliated Hospital from January 2015 to December 2020. Among them, 144 patients (7.4%) developed DVT, including 43 cases in the left lower extremity, 49 cases in the right lower extremity, 52 cases in both lower extremities, and 3 cases of PE, all of which were accompanied by DVT in the lower extremities. Among the VTE patients, the highest number of VTE cases occurred after thoracolumbar surgery, accounting for 106 cases (73.6%), while cervical spine surgery accounted for 38 cases (26.4%). The average operation duration for VTE patients was 189.31 ± 90.461 min, and the amount of bleeding was 459.44 ± 462.723 mL (Table 1).

Table 1 Patient characteristics and univariate correlations with thrombosis-related risk factors
Full size table

Risk factors associated with VTE

In the univariate logistic analysis, there were significant differences (P < 0.05) between the two groups of patients in terms of age, hypertension, diabetes, pre_surgery, bed, smoking, drinking, cancer, trauma, WBC, Hb, TC, D_dimer, PT, APTT, location, duration, and bloodloss(Fig. 1A). LASSO regression was used to further eliminate overfitting and identified 17 risk factors (Fig. 2). Subsequently, these 17 indicators were included in the multivariate logistic analysis, which showed that age, hypertension, bed, drinking, trauma, TC, D_dimer, location, duration, and bloodloss were independent risk factors for postoperative VTE in spinal surgery (Fig. 1B).

Fig. 1
figure 1

The forest plot shows the results of univariate and multivariate analyses of VTE after spinal surgery. Notes: A In the univariate logistic analysis, 18 risk factors were presented. B In the multivariate logistic regression analysis, 10 independent risk factors for VTE were further screened out. BMI body mass index, Pre_surgery previous surgery, WBC white blood cell count, PLT platelet, Hb hemoglobin, TG triglyceride, TC total cholesterol, PT prothrombin time, TT thrombin time, APTT activated partial thromboplastin time

Full size image
Fig. 2
figure 2

The least absolute shrinkage and selection operator (LASSO) method for selecting postoperative deep vein thrombosis progression risk factors. Notes: A LASSO model was cross-validated using the minimum criterion, with dashed plumb lines drawn at the optimal values (9 factors). B The 18 feature LASSO coefficient profiles for logarithmic (lambda) sequences are constructed

Full size image

Establishment of a dynamic interactive nomogram

A dynamic interactive nomogram was established based on the 10 independent risk factors obtained from multiple logistic regression, which predicts the risk of postoperative VTE in spine surgery. As shown in Fig. 3, The 89th patient in the training group was selected as the subject for the present dynamic interactive predictive model. This patient was 80 years old, had a history of hypertension and bed rest > 3 days, no history of drinking or trauma, TC < 5.17, and D_Dimer > 3000 μg/L. underwent surgery in the thoraclumbar spine, with a surgical duration of 125 min and bloodloss of 1000 mL, and developed VTE after surgery. Moreover, the patient experienced VTE after the spinal surgery. According to this predictive model, the total score for this patient was 8.0, corresponding to a probability of 0.878 for postoperative VTE, which indicated an extremely high risk of VTE for this patient, consistent with the patient's outcome.

Fig. 3
figure 3

A dynamic interactive nomogram was established to predict postoperative VTE in spinal surgery. The corresponding score for each factor is based on the condition of the patient, which can be determined by making a vertical line upwards (e.g., a patient with hypertension will receive between 70 and 75 scores). Add all the scores to get the total score, then find the corresponding point on the total points axis and make a vertical line down to predict the risk of VTE after spinal surgery. *P < 0.05; ***P < 0.001

Full size image

Validation of the dynamic interactive nomogram

The C-index for the training and validation groups were 0.94 (95% CI: 0.9204–0.9596) and 0.955 (95% CI: 0.9354–0.9746) respectively, indicating high accuracy and stability of the prediction model. The ROC curves were plotted in the training and validation groups, and the area under the ROC curve (AUC) was calculated to determine the discrimination of the predictive model. The AUC for the training group was 0.940, and the AUC for the validation group was 0.942, indicating a high discrimination of the predictive model [17]. Additionally, the ROC curve in the training group showed a cutoff point of 0.085, indicating that patients with a calculated VTE risk probability greater than 0.085 in the predictive model should receive corresponding clinical interventions to reduce the risk of VTE after spinal surgery (Fig. 4A, B). The Hosmer–Lemeshow goodness-of-fit test (training group: Chi-square = 9.601285, P-value = 0.3837164; validation group: Chi-square = 6.942186, P-value = 0.6431388) and calibration curves demonstrated good calibration of the model in both the training and validation groups (Fig. 4C, D).

Fig. 4
figure 4

The AUC of training group (AUC = 0.940) (A) and validation group (AUC = 0.942) B showed that the model had a high discrimination ability. C, D The calibration curves for assessing the consistency between the predicted and the actual risk of postoperative VTE. Favorable consistencies between the predicted and the actual risk evaluation are presented

Full size image

Clinical application

The decision curve analysis(DCA) was carried out to evaluate the clinical implications of the prediction model. The DCA for the preoperative VTE progression nomogram is illustrated in Fig. 5. The DCA demonstrated that within the threshold probability range of 0.01–1, the DCA of the predictive model constructed in this study had a higher net benefit than the two ineffective curves, indicating that the use of the VTE nomogram in clinical practice to predict the risk of VTE after spinal surgery and take necessary preventive measures can effectively reduce the risk of postoperative VTE occurrence.

Fig. 5
figure 5

DCA for the preoperative VTE progression nomogram. Notes: The Y-axis indicates the net benefit. The solid red line indicates the risk of preoperative DVT progression nomogram. The thin solid line indicates the assumption that progression of preoperative DVT is assumed to have occurred in all patients. The thick solid line indicates the hypothesis that no patients had progression of preoperative DVT. The decision curve demonstrated that using this preoperative DVT progression nomogram in the current study to predict preoperative DVT progression risk adds more benefit than either the intervention-all-patients scheme or the intervention-none scheme

Full size image

Discussion

VTE is one of the most common complications in patients after spinal surgery. It presents with an insidious onset and can lead to severe complications such as stroke, pulmonary embolism, and even death, imposing a significant burden on patients and their families [16, 17]. Therefore, accurately assessing and predicting the risk of VTE in the early stages after spinal surgery is of crucial importance for prognosis. Thrombus formation is a complex process influenced by various factors, previous single-factor prediction models have failed to accurately predict the risk of thrombus formation and there is a scarcity of comprehensive multicenter studies and reliable clinical prediction models specifically focused on VTE after spinal surgery [18,19,

Conclusions

In conclusion, the independent risk factors for postoperative VTE in spinal surgery include age, hypertension, bed rest for more than 3 days, drinking, trauma, triglyceride, D-dimer, surgical location, operation duration, and blood loss, especially in patients with multiple risk factors, early intervention should be taken to prevent the occurrence of VTE. The VTE prediction model established by our team is simple and feasible, on the one hand, it can be used by Primary Healthcare and Medical Institution to educate and provide information to patients with spinal disorders who are undergoing conservative treatment, change daily lifestyle such as a low-cholesterol diet, limited alcohol consumption, effective blood pressure control, appropriate physical activity and so on. The patients take proactive measures for preventive care, thereby reducing the risk of potential complications. On the other hand, it can also encourage the spinal surgeons to improve their surgical skills, choose the most suitable surgical techniques, minimize surgical duration, and reduce blood loss et al., in order to achieve the purpose of passive prevention for patients undergoing spine surgery. By combining proactive and passive prevention strategies, intervention for high-risk factors of VTE throughout entire spinal disorder management can more effectively prevent the occurrence of VTE after spinal surgery.