Construction and validation of prediction model for neurogenic shock in patients with traumatic cervical spinal cord injury
10.3760/cma.j.cn501098-20240228-00167
- VernacularTitle:创伤性颈脊髓损伤患者并发神经源性休克预测模型的构建及验证
- Author:
Zilin SONG
1
;
Liming HE
;
Qingqing LIU
;
Haoyu FENG
Author Information
1. 山西医科大学第三医院（山西白求恩医院，山西医学科学院，同济山西医院）骨科，太原 030032
- Keywords:
Spinal cord injuries;
Shock;
Nomograms
- From:
Chinese Journal of Trauma
2024;40(7):585-592
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a prediction model for neurogenic shock in patients with traumatic cervical spinal cord injury and validate its effectiveness.Methods:A retrospective case-control study was conducted on the clinical data of 381 patients with traumatic cervical spinal cord injury admitted to the Third Hospital of Shanxi Medical University from January 2017 to December 2022, including 311 males and 70 females, aged 12-86 years [(55.1±12.9)years]. A total of 121 patients (31.8%) were complicated with neurogenic shock. The patients were randomly divided into training set ( n=267) and validation set ( n=114) with a ratio of 7∶3. The training set was divided into neurogenic shock group ( n=81) and non-neurogenic shock group ( n=186) according to whether they were complicated with neurogenic shock. The general data, clinical data, laboratory indicators and imaging data of the patients were collected. Univariate analysis was used to determine differences in the aforementioned indicators between the neurogenic shock group and non-neurogenic shock group in the training set. Multivariate Logistic regression analysis was conducted to screen the predictors for neurogenic shock in patients with traumatic cervical spinal cord injury, and regression equation was constructed. A nomogram prediction model based on the regression equation was plotted with R programming language. Receiver operating characteristic (ROC) curves of the training set and validation set were plotted, when the area under the curve (AUC) was calculated to determine the discriminability of the model. The calibration of the model was assessed with calibration curves. The clinical applicability of the model was evaluated by the decision curve analysis (DCA). Results:The univariate analysis showed that there were statistically significant differences in the American Spinal Injury Association (ASIA) grade, tracheal intubation, serum albumin concentration within 24 hours on admission, intramedullary lesion length (IMLL), maximum spinal cord compression (MSCC), increased signal intensity (ISI), and highest damaged segment between the neurogenic shock group and non-neurogenic shock group in the training set ( P<0.05). The multivariate Logistic regression analysis revealed that AISA grade (grade C vs. grade A: OR=0.13, 95% CI 0.03, 0.59, P<0.01; grade D vs.grade A: OR=0.04, 95% CI 0.01, 0.28, P<0.01), serum albumin concentration within 24 hours on admission ( OR=0.75, 95% CI 0.65, 0.86, P<0.01), IMLL ( OR=2.71, 95% CI 1.68, 4.38, P<0.01), ISI (grade 2 vs.grade 0: OR=5.62, 95% CI 1.07, 29.48, P<0.05), and highest damaged segment ( OR=0.49, 95% CI 0.29, 0.83, P<0.01) were predictors for neurogenic shock in patients with traumatic cervical spinal cord injury. Based on the 5 forementioned variables, the regression equation was constructed as follows: Logit[ P/(1- P)]=10.99-1.06×＂AISA grade＂-0.29×＂serum albumin concentration within 24 hours on admission＂+1.04×＂IMLL＂+0.89×＂ISI＂-0.74×＂highest damaged segment＂. In the prediction model constructed based on the equation, the AUC values of the training set and validation set were 0.97 (95% CI 0.97, 0.99) and 0.95 (95% CI 0.91, 0.99). Calibration curves of the training set and validation set demonstrated the prediction curve roughly overlapped with the reference curve and the mean absolute errors of the two sets were 0.013 and 0.050. DCA results showed that the net benefit rate of patients was greater than 0 when the threshold probability ranged from 0% to 97% for the training set and from 0% to 100% for the validation set. Conclusion:The prediction model based on the AISA grade, serum albumin concentration within 24 hours on admission, IMLL, ISI, and highest damaged segment demonstrates good discriminability, calibration and clinical applicability in predicting neurogenic shock in patients with traumatic cervical spinal cord injury.
