Construction and validation of a nomogram for predicting unfavorable prognosis at 6 months after moderate and severe traumatic brain injury
10.3760/cma.j.cn501098-20240423-00303
- VernacularTitle:中重型创伤性脑损伤患者伤后6个月预后不良预测列线图的构建与验证
- Author:
Hongqiao YANG
1
;
Zhaopeng ZHOU
;
Mei LIU
;
Changgeng DING
;
Wenwen CHE
;
Yuhai WANG
Author Information
1. 中国人民解放军联勤保障部队第九〇四医院神经外科，安徽医科大学无锡临床学院，无锡 214008
- Keywords:
Brain injuries;
Prognosis;
Model, statistical;
Nomograms
- From:
Chinese Journal of Trauma
2024;40(6):487-497
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a nomogram for predicting unfavorable prognosis at 6 months after moderate and severe traumatic brain injury (msTBI) and validate its predictive effectiveness.Methods:A retrospective cohort study was conducted to analyze the clinical data of 387 patients with msTBI who were admitted to 904th Hospital of the Joint Logistic Support Force of PLA from January 2020 to December 2022, including 265 males and 122 females, aged 6-97 years [58(47, 68)years]. According to the Glasgow outcome scale (GOS) score at 6 months after injury, the patients were divided into favorable prognosis group (GOS 4-5 points, n=201) and unfavorable prognosis group (GOS 1-3 points, n=186). The clinical characteristics, imaging manifestations, and laboratory test results of the two groups on admission were recorded. Univariate analysis was applied to evaluate the correlation between the aforementioned indicators and the unfavorable prognosis of the msTBI patients at 6 months after injury. Receiver operating characteristic (ROC) curves of single variable and the correlation heatmap among continuous variables were plotted. Lasso regression was used to select variables and multivariate Logistic regression analysis was used to determine independent predictive factors so as to construct Logistic regression equation and plot the nomogram. The internal verification was carried out by means of random and non-random split of data. In random split, the data were divided randomly with a ratio of 6∶4 into training group ( n=232) and verification group ( n=155). In non-random split, the patients admitted from January 2020 to December 2021 were assigned to the training group ( n=260), while those admitted from January 2022 to December 2022 to the verification group ( n=127). Area under the curve (AUC) was used to evaluate the predictive ability of the model in the training group and verification group, calibration curve and Hosmer-Lemeshow (H-L) test to evaluate its goodness of fit, and decision curve analysis (DCA) to evaluate its clinical applicability. The influence of inclusion of neutrophil-to-lymphocyte ratio (NLR) model on the warning effectiveness of poor prognosis was analyzed in comparison with the model without inclusion of NLR. Results:Univariate analysis showed that there was a certain correlation between age, length of hospital stay, Glasgow coma scale (GCS), American Society of Anesthesiologists Physical Status (ASA-PS) classification, Injury severity score (ISS), prehospital tracheal intubation, hypotension, hypoxia, pupillary responsiveness, midline shift, basilar cisterna status, traumatic subarachnoid hemorrhage (tSAH), D-Dimer, prothrombin time activity (PTA), glucose, hemoglobin, K +, Cl -, Ca 2+, HCO -, creatinine, albumin, lactic acid, platelet, lymphocyte, systemic immune-inflammation index (SII), NLR, lymphocyte-to-monocyte ratio (LMR) and unfavorable prognosis of msTBI patients at 6 months after injury ( P<0.05 or 0.01). The ROC curve of single variable showed that GCS (AUC=0.82), ISS (AUC=0.81), pupillary responsiveness (AUC=0.76), basal cistern status (AUC=0.73) and NLR (AUC=0.73) had good predictive validity. The results of the correlation heatmap showed that there was a significant correlation and collinearity among the continuous variables, while no collinearity was found between ISS and NLR. Fourteen potential predictors selected by Lasso regression were included in multivariate Logistic regression analysis and its results showed that age ( OR=0.86, 95% CI 1.38, 5.19), GCS 6-8 points ( OR=3.13, 95% CI 1.06, 9.27), GCS 3-5 points ( OR=12.36, 95% CI 2.81, 54.27), ISS ( OR=3.68, 95% CI 1.38, 9.80), pupillary responsiveness ( OR=2.45, 95% CI 0.85, 7.07), and NLR ( OR=2.62, 95% CI 1.52, 4.51) were identified as the independent risk factors for unfavorable prognosis of msTBI patients at 6 months after injury ( P<0.05 or 0.01). The multivariate Logistic regression equation was Logit [P/(1-P)]=0.066×"age"+ 1.474×"GCS 6-8"+2.357×"GCS 3-5"+0.066×"ISS"+0.965×"absence of pupillary light reflex"+0.194×"NLR"-10.704. In the internal verification of random split of data, the AUC value of the model was 0.93 (95% CI 0.89, 0.96) in the training group and 0.93 (95% CI 0.89, 0.97) in the verification group. In the internal verification of non-random split, the AUC value was 0.94 (95% CI 0.91, 0.97) in the training group and 0.93 (95% CI 0.89, 0.97) in the verification group. The calibration curve and H-L test showed that the model had good calibration ability ( P>0.5). The results of DCA showed that the application of the nomogram would increase the net benefit of the patients (risk threshold probability of 0.0-0.8). Compared with the conventional model (AUC=0.90), inclusion of NLR model (AUC=0.93) enhanced the warning effectiveness. Conclusions:Age, GCS, ISS, pupillary responsiveness and NLR are independent risk factors affecting unfavorable prognosis in msTBI patients at 6 months after injury, based on which the nomogram constructed can better predict the clinical outcome of msTBI patients.
