Risk factors analysis and risk prediction model construction and validation of cognitive dysfunction after brain trauma
10.3760/cma.j.cn371468-20220713-00398
- VernacularTitle:脑创伤后认知功能障碍危险因素分析及风险预测模型构建与验证
- Author:
Xijun HAO
1
;
Ping LEI
;
Xiaobin MA
;
Changxiang CHEN
Author Information
1. 天津医科大学总医院老年医学科,天津 300052
- Keywords:
Brain trauma;
Cognitive impairment;
Risk factors;
Prediction model
- From:
Chinese Journal of Behavioral Medicine and Brain Science
2023;32(1):37-44
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To analyze the independent risk factors for the occurrence of post-traumatic cognitive dysfunction, construct a prediction model for the risk factors of post-traumatic cognitive dysfunction, and verify the effectiveness of the risk prediction model, so as to provide a clinical tool for early prediction of the risk of post-traumatic cognitive impairment.Methods:Part I: patients with brain trauma (training set with 556 subjects) who were hospitalized in 21 tertiary and secondary hospitals from Tangshan, Cangzhou and Chengde cities of Hebei province were retrospectively collected from February to May 2021 for Montreal cognitive assessment, and 33 influencing factors (general data, symptoms and signs, laboratory and imaging parameters) were obtained obtained through literature research.The patients were divided into case group and control group according to whether they had cognitive impairment or not, and univariate and multivariate analysis were used to screen independent risk factors.Part Ⅱ: a binary Logistic regression equation was used to construct a cognitive impairment prediction model, the visualization model of line graph is presented.Part Ⅲ: brain trauma patients (260 subjects of the validation set) hospitalized in the aforementioned 21 hospitals from August to October 2021 were collected as a prospective validation population for the prediction model of cognitive impairment, and the grouping basis of case group and control group was the same as before.And the risk factors between the two groups were compared.The receiver operating characteristic curve(ROC), calibration curve and clinical applicability of the model were drawn to evaluate the effectiveness of the model for internal and external verification of the model.Results:Binary Logistic regression analysis showed that the risk factors for post-traumatic cognitive dysfunction were basal ganglia injury, severe injury, amnesia experience after injury, frequent headache after injury, upper limb dysfunction after injury, age ≥ 60 years, and education level of elementary school or below.Visual nomograms showed that the experience of amnesia after injury, frequent headache after injury, upper limb dysfunction, and degree of injury among the symptom factors were the factors that contributed greatly to the risk of traumatic brain injury cognitive impairment in this model.Predictive model discrimination using area under curve(AUC) values of the area under the ROC curve showed that internal validation and external validation were 0.868 and 0.885 for R language analysis and 0.868 and 0.901 for SPSS analysis, respectively.The curve after model calibration almost coincided with the reference line, Hosmer-Lemeshow test P>0.05.The two decision curve analysis (DCA) curves drawn by the clinical applicability of the model were higher than the two extreme curves, predicting that traumatic brain injury patients with cognitive impairment could benefit from the predictive model, and there was a net benefit rate in the range of Pt about 0.1-0.8, when Pt reached about 0.1 until the approximate 1.0 composite evaluation model. Conclusion:Risk factors such as experience of amnesia after injury, frequent headache after injury, upper limb dysfunction, and degree of injury are predicting factors contributed to the risk of cognitive impairment in traumatic brain injury, and their prediction models have good predictive effect, high predictive accuracy and good clinical applicability, which can be applied in clinical diagnosis.