Establishment and verification of a predictive model for hyponatremia after craniocerebral trauma / 国际外科学杂志

ABSTRACT

Objective:Based on clinical data, establish and verify the prediction model of hyponatremia after craniocerebral trauma.Methods:Through retrospective analysis of 226 patients with traumatic brain injury who were eligible for inclusion in Department of Neurosurgery, Qingpu Branch, Zhongshan Hospital, Fudan University from June 2019 to January 2021. After 6 months of follow-up, the patients were divided into the hyponatremia group ( n=81) and the normal group ( n=145) according to their blood sodium level. Various factors that may be related to the occurrence of hyponatremia were analyzed. Logistic multivariate regression was used to analyze the independent risk factors of patients with hyponatremia, the best cut-off point of was determined by the characteristic curve (ROC), and the nomogram prediction model was established and verified based on the independent risk factors of patients with hyponatremia. Results:Patients in the normal blood sodium group were(40.33±15.32) years old, 87 cases (60.00%) of cerebral contusion, 3 cases (2.07%) of intraventricular hemorrhage, 15 cases (10.34%) of cerebral hernia, 35 cases (24.14%) of GCS score <6 and 110 cases (75.86%) of GCS score ≥6; In the hyponatremia group, patients were (47.17±16.03) years old, with 65 cases (80.25%) of cerebral contusion, 7 cases (8.64%) of intraventricular hemorrhage, 19 cases (23.46%) of cerebral hernia, 44 cases (54.32%) of GCS score <6 and 37 cases (45.68%) of GCS score ≥6, the difference between the two groups was statistically significant ( P<0.05). Logistic multivariate regression analysis found that increasing age, ventricular hemorrhage, cerebral herniation, elevated ADH and GCS score <6 were all independent risk factors for patients with hyponatremia( OR=2.287, 2.531, 1.344, 1.387, 1.582; 95% CI 1.945-2.628, 1.338-3.723, 1.256-1.431, 1.013-1.761, 1.233-1.931; P<0.05). The GCS score predicts that the area under the ROC curve for patients with hyponatremia was 0.795, the cut-off point was 6 points, the sensitivity was 95.00%, and the specificity was 64.30%. A nomogram model was constructed to predict patients with hyponatremia based on independent risk factors, in which the C-index calculation result and ROC curve AUC were 0.798 (95% CI 0.751-0.842) and 0.815 (95% CI 0.767-0.877), respectively, indicating a good degree of discrimination; Calibration curve evaluation results showed that the model has good consistency. Conclusion:GCS score is an independent risk factor that affects whether patients with craniocerebral trauma are complicated by hyponatremia. It has a good predictive value for whether this type of patients is complicated by hyponatremia, which is conducive to early identification and intervention of patients.