Using machine learning and big data for the prediction of venous thromboembolic events after spine surgery: A single-center retrospective analysis of multiple models on a cohort of 6869 patients.

Objective: Venous thromboembolic event (VTE) after spine surgery is a rare but potentially devastating complication. With the advent of machine learning, an opportunity exists for more accurate prediction of such events to aid in prevention and treatment. Methods: Seven models were screened using 108 database variables and 62 preoperative variables. These models included deep neural network (DNN), DNN with synthetic minority oversampling technique (SMOTE), logistic regression, ridge regression, lasso regression, simple linear regression, and gradient boosting classifier. Relevant metrics were compared between each model. The top four models were selected based on area under the receiver operator curve; these models included DNN with SMOTE, linear regression, lasso regression, and ridge regression. Separate random sampling of each model was performed 1000 additional independent times using a randomly generated training/testing distribution. Variable weights and magnitudes were analyzed after sampling. Results: Using all patient-related variables, DNN using SMOTE was the top-performing model in predicting postoperative VTE after spinal surgery (area under the curve [AUC] =0.904), followed by lasso regression (AUC = 0.894), ridge regression (AUC = 0.873), and linear regression (AUC = 0.864). When analyzing a subset of only preoperative variables, the top-performing models were lasso regression (AUC = 0.865) and DNN with SMOTE (AUC = 0.864), both of which outperform any currently published models. Main model contributions relied heavily on variables associated with history of thromboembolic events, length of surgical/anesthetic time, and use of postoperative chemoprophylaxis. Conclusions: The current study provides promise toward machine learning methods geared toward predicting postoperative complications after spine surgery. Further study is needed in order to best quantify and model real-world risk for such events.

Classifying vertebral artery anatomy abnormality in children with skeletal dysplasia.

PURPOSE: Skeletal dysplasia (SKD) have predictably abnormal occipitocervical skeletal anatomy, but a similar understanding of their vertebral artery anatomy is not known. Knowledge and classification of vertebral artery anatomy in SKD patients is important for safe surgical planning. We aimed to determine if predictably abnormal vertebral artery anatomy exists in pediatric SKD. METHODS: We performed a retrospective review of CTAs of the neck for pediatric patients at a single institution from 2006 to 2018. CTAs in SKD and controls were reviewed independently in blinded fashion by two radiologists who classified dominance, vessel curvature at C2, direction at C3, and presence of fenestration and intersegmental artery. RESULTS: 14 skeletal dysplasia patients were compared to 32 controls. The path of the vertebral artery at C2 foramen was no different between the cohorts or by side, right (p = 0.43) or left (p = 0.13), nor for medial or lateral exiting direction from C3 foramen on right (p = 0.82) or left (p = 0.60). Dominance was most commonly neutral in both groups (71% in SKD and 63% in controls). There were no fenestrated nor first intersegmental arteries in our cohort. CONCLUSION: No systematic differences were detected between SKD and control patients with respect to vertebral artery anatomy. Nonetheless, surgically relevant variability was observed in both groups. Paying particular attention to the direction of exit at C3 and curvature at C2 with respect to the foramen and vessel dominance are important and easily classifiable abnormalities that both surgeons and radiologists can use to communicate and employ in pre-operative planning. LEVEL OF EVIDENCE: III.

Relationship between crossover sign and anterior center-edge angle.

PURPOSE: The aim of this study was to evaluate the relationship between the anterior center-edge angle (ACEA) and lateral center-edge angle (LCEA) and crossover ratio. METHODS: Consecutive patients presenting for evaluation of hip pain were reviewed. The following measurements were recorded and analyzed: Crossover ratio, LCEA, ACEA, and alpha-angle. RESULTS: 68 patients met inclusion criteria. The only statistically significant radiographic measurement when stratified by gender was alpha angle (P < 0.001). There was moderate correlation between crossover ratio and ACEA and LCEA with coefficients of -0.48 and -0.48, respectively. CONCLUSION: A correlation exists between crossover ratio and ACEA and LCEA.

Venous thromboembolism events following spinal fractures: A single center experience.

OBJECTIVE: Venous thromboembolic events (VTE), including deep venous thrombosis (DVT) and pulmonary embolism (PE), are a major cause of readmission, morbidity, and mortality after spine surgery. Patients with spinal fractures are particularly at an increased risk for VTE. The objective of this study is to understand VTE risk factors in this patient population and to examine current institutional practices. PATIENTS AND METHODS: We retrospectively examined records from 195 consecutive patients with spinal fractures who underwent spinal stabilization surgeries- amongst a cohort of 6869 patients who underwent spinal surgery. We collected data on patient demographics, surgery, hospital course, and 30-day rates of VTE, readmission, reoperation. Multivariable logistic regression was used to identify independent predictors of each outcome. RESULTS: Among 195 patients undergoing surgery for spinal fractures, 9.2% experienced a VTE, compared to 2.3% among all other spine patients (OR 4.466, p < 0.0001). 48.7% spine fracture patients received chemoprophylactic anticoagulation, compared to 35.7% of all other spine patients (OR 2.657, p < 0.0001). Within 30 days of surgery, estimated blood loss (EBL) was associated with VTE (OR 1.001, p = 0.0415) and DVT (OR 1.001, p = 0.049), and comorbid cardiac disease burden showed a trend toward significance in predicting both VTE (OR 1.890, p = 0.0956) and DVT (OR 4.228, p = 0.0549). Number of levels in surgery predicted PE within 30 days of surgery (OR 1.573, p = 0.0107). CONCLUSIONS: Compared to all other patients undergoing spine surgery, patients with spinal fractures are more likely to receive chemoprophylactic anticoagulation, but nevertheless have a higher rate of VTE events. EBL and comorbid disease burden predict VTE events in patients with spine fractures.

Epidemiology and national trends in prevalence and surgical management of metastatic spinal disease.

Surgical treatment for spinal metastasis has benefited from improvements in surgical techniques. However, the trends in treatment and outcomes for spinal metastasis surgery have not been well-established in a pediatric population. Patients <20 years old with metastatic spinal tumors undergoing spinal surgery were identified in the KID database. Trends for spinal metastases treatment and patient outcomes were analyzed using weight-adjusted ANOVAs. 333 patients were identified in the KID database. The top five primary diagnoses were metastatic brain/spinal cord tumor (19.8%), metastatic nervous system tumor (15.9%), metastatic bone cancer (13.2%), spinal cord tumor (4.2%), and tumor of ventricles (3.0%). There was an increased incidence of spinal metastasis diagnoses from 2003 to 2012 (88.5-117.9 per 100,000; p < 0.001) and an increased trend in the incidence of surgical treatment for spinal metastasis from 2003 to 2012 (p = 0.014). The average age was 10.19 ±â€¯6.33 years old and 38.4% were female. The average length of stay was 17.34 ±â€¯24.36 days. Average CCI increased over time (2003: 7.87 ±â€¯1.40, 2012: 8.44 ±â€¯1.39; p = 0.006). The most common surgeries were excision of spinal cord/meninges lesions (69.1%) and decompression of spinal canal (38.1%). Length of hospital stay and in-hospital mortality did not change over time (17.34-18.04 days, p = 0.337; 1.6%-2.9%, p = 0.801). 10.5% of patients underwent a posterior fusion and 22.2% had at least one complication (nervous system, respiratory, dysphagia, infection). The overall complication rate remained stable over time (23.4%-21.8%, p = 0.952). Surgical treatment for spinal metastasis in the last decade has increased, though the complication rates, in-hospital mortality, and length of stay have remained stable.

The timing of venous thromboembolic events after spine surgery: a single-center experience with 6869 consecutive patients.

OBJECTIVE Venous thromboembolic events (VTEs), including both deep venous thrombosis (DVT) and pulmonary embolism, are a major cause of morbidity and mortality after spine surgery. Prophylactic anticoagulation, or chemoprophylaxis, can prevent VTE. However, the timing of VTEs after spine surgery and the effect of chemoprophylaxis on VTE timing remain underinvestigated. METHODS The records of 6869 consecutive spine surgeries were retrospectively examined. Data on patient demographics, surgical variables, hospital course, and timing of VTEs were collected. Patients who received chemoprophylaxis were compared with those who did not. Appropriate regression models were used to examine selection for chemoprophylaxis and the timing of VTEs. RESULTS Age (OR 1.037, 95% CI 1.023-1.051; p < 0.001), longer surgery (OR 1.003, 95% CI 1.002-1.004; p < 0.001), history of DVT (OR 1.697, 95% CI 1.038-2.776; p = 0.035), and fusion surgery (OR 1.917, 95% CI 1.356-2.709; p < 0.001) predicted selection for chemoprophylaxis. Chemoprophylaxis patients experienced more VTEs (3.62% vs 2.03% of patients, respectively; p < 0.001), and also required longer hospital stays (5.0 days vs 1.0 days; HR 0.5107; p < 0.0001) and had a greater time to the occurrence of VTE (median 6.8 days vs 3.6 days; HR 0.6847; p = 0.0003). The cumulative incidence of VTEs correlated with the postoperative day in both groups (Spearman r = 0.9746, 95% CI 0.9457-0.9883, and p < 0.0001 for the chemoprophylaxis group; Spearman r = 0.9061, 95% CI 0.8065-0.9557, and p < 0.0001 for the nonchemoprophylaxis group), and the cumulative incidence of VTEs was higher in the nonchemoprophylaxis group throughout the 30-day postoperative period. Cumulative VTE incidence and postoperative day were linearly correlated in the first 2 postoperative weeks (R = 0.9396 and p < 0.0001 for the chemoprophylaxis group; R = 0.8190 and p = 0.0003 for the nonchemoprophylaxis group) and the remainder of the 30-day postoperative period (R = 0.9535 and p < 0.0001 for the chemoprophylaxis group; R = 0.6562 and p = 0.0058 for the nonchemoprophylaxis group), but the linear relationships differ between these 2 postoperative periods (p < 0.0001 for both groups). CONCLUSIONS Anticoagulation reduces the cumulative incidence of VTE after spine surgery. The cumulative incidence of VTEs rises linearly in the first 2 postoperative weeks and then plateaus. Surgeons should consider early initiation of chemoprophylaxis for patients undergoing spine surgery.

Predictors of Readmissions and Reoperations Related to Venous Thromboembolic Events After Spine Surgery: A Single-Institution Experience with 6869 Patients.

BACKGROUND: Readmission and reoperation are used as hospital and surgeon quality metrics. Venous thromboembolic (VTE) events, including deep vein thrombosis and pulmonary embolism (PE), are a major cause of readmission, morbidity, and mortality after spine surgery. Specific procedural, perioperative, and patient characteristics may be associated with these outcomes. METHODS: We retrospectively examined records from 6869 consecutive spine surgeries at our institution. We collected data on patient demographics, surgery, hospital course, and 30-day rates of VTE, readmission, reoperation, and epidural hematoma. Stepwise multivariable logistic regression was used to identify independent predictors of each outcome. RESULTS: Factors associated with VTE within 30 days of surgery include a history of VTE (odds ratio [OR] 3.92 [confidence interval 1.83-8.36], P < 0.001), estimated blood loss (EBL; OR 1.017 [1.005-1.029], P = 0.004), fracture (OR 5.42 [2.09-14.05], P = 0.001), history of PE (OR 4.04 [1.22-13.42], P = 0.023), and transfusion (OR 2.26 [1.07-4.77], P = 0.033). Factors associated with readmission were a history of PE (OR 3.27 [1.07-9.97], P = 0.038), PE (OR 8.07 [2.26-28.8], P = 0.001), transfusion (OR 2.54 [1.55-4.17], P < 0.001), comorbid disease burden (OR 1.35 [1.01-1.80], P = 0.041), and tumor surgery (OR 2.84 [1.32-6.10], P = 0.007). Factors associated with reoperation were EBL (OR 1.024 [1.006-1.042], P = 0.008), transfusion (OR 3.86 [1.38-10.79], P = 0.01), and PE (OR 6.05 [1.03-35.62], P = 0.046). Transfusion was associated with epidural hematoma within 30 days (OR 7.38 [1.37-39.83], P = 0.02). CONCLUSIONS: Transfusion and EBL are associated with numerous negative outcomes. Transfusion is an independent predictor of VTE, readmission, reoperation, and epidural hematoma requiring evacuation. Specific pathologies were associated with specific negative outcomes.

Timing and risks of chemoprophylaxis after spinal surgery: a single-center experience with 6869 consecutive patients.

OBJECTIVE Venous thromboembolism (VTE) after spinal surgery is a major cause of morbidity, but chemoprophylactic anticoagulation can prevent it. However, there is variability in the timing and use of chemoprophylactic anticoagulation after spine surgery, particularly given surgeons' concerns for spinal epidural hematomas. The goal of this study was to provide insight into the safety, efficacy, and timing of anticoagulation therapy after spinal surgery. METHODS The authors retrospectively examined records from 6869 consecutive spinal surgeries performed in their departments at Northwestern University. Data on patient demographics, surgery, hospital course, timing of chemoprophylaxis, and complications, including deep venous thrombosis (DVT), pulmonary embolism (PE), and spinal epidural hematomas requiring evacuation, were collected. Data from the patients who received chemoprophylaxis (n = 1904) were compared with those of patients who did not (n = 4965). The timing of chemoprophylaxis, the rate of VTEs, and the incidence of spinal epidural hematomas were analyzed. RESULTS The chemoprophylaxis group had more risk factors, including greater age (59.70 vs 51.86 years, respectively; p < 0.001), longer surgery (278.59 vs 145.66 minutes, respectively; p < 0.001), higher estimated blood loss (995 vs 448 ml, respectively; p < 0.001), more comorbid diagnoses (2.69 vs 1.89, respectively; p < 0.001), history of VTE (5.8% vs 2.1%, respectively; p < 0.001), and a higher number were undergoing fusion surgery (46.1% vs 24.7%, respectively; p < 0.001). The prevalence of VTE was higher in the chemoprophylaxis group (3.62% vs 2.03%, respectively; p < 0.001). The median time to VTE occurrence was shorter in the nonchemoprophylaxis group (3.6 vs 6.8 days, respectively; p = 0.0003, log-rank test; hazard ratio 0.685 [0.505-0.926]), and the peak prevalence of VTE occurred in the first 3 postoperative days in the nonchemoprophylaxis group. The average time of initiation of chemoprophylaxis was 1.46 days after surgery. The rates of epidural hematoma were 0.20% (n = 4) in the chemoprophylaxis group and 0.18% (n = 9) in the nonchemoprophylaxis group (p = 0.622). CONCLUSIONS The risks of spinal epidural hematoma among patients who receive chemoprophylaxis and those who do not are low and equivalent. Administering anticoagulation therapy from 1 day before to 3 days after surgery is safe for patients at high risk for VTE.

Muscarinic cholinergic receptor antagonists in the VTA and RMTg have opposite effects on morphine-induced locomotion in mice.

The ventral tegmental area (VTA) and the rostromedial tegmental nucleus (RMTg) each contribute to opiate reward and each receive inputs from the laterodorsal tegmental and pedunculopontine tegmental nuclei, the two principle brainstem cholinergic cell groups. We compared the contributions of VTA or RMTg muscarinic cholinergic receptors to locomotion induced by morphine infusions into the same sites. VTA co-infusion of atropine completely blocked VTA morphine-induced locomotion providing additional support for the important role of VTA muscarinic cholinergic receptors in the stimulant effects of opiates. By contrast, RMTg co-infusion of atropine increased RMTg morphine-induced locomotion. Furthermore, RMTg co-infusion of the M3-selective antagonist 4-DAMP, but not the M4-selective antagonist Tropicamide, strongly increased RMTg morphine-induced locomotion. RMTg infusions of 4-DAMP, but not of Tropicamide, by themselves strongly increased drug-free locomotion. Muscarinic cholinergic receptors in the RMTg thus also contribute to the stimulant effects of morphine, but in a way opposite to those in VTA. We suggest that the net effect of endogenous cholinergic input to the RMTg on drug-free and on RMTg morphine-induced locomotion is inhibitory.