Comparison of survival to hospital discharge predictions in elderly traumatic brain injury using CHID, CIT, and CART algorithms

Background: Elderly patients with a traumatic brain injury (TBI) are at increased risk for a worse outcome and a poor prognosis. Lack of clinical prediction tools (CPTs) to assess the prognosis of elderly TBI patients. The goal of this study was to compare the predictability of survival to hospital discharge among tree-based algorithms in elderly TBI. Methods: An observational study was performed to study individuals aged 60 years and above who suffered a TBI. To develop and validate the predictive models, included patients were divided into a training set and a testing set, with a ratio of 7:3. The tree-based techniques including Chi-square Automation Interaction Detection (CHAID), Conditional Inference Trees (CIT), and Classification and Regression Tree (CART) were used to develop the predictive models using the training dataset. Therefore, the testing dataset was used to estimate the tree-based models' prediction capabilities. Results: There were 2,391 patients in the entire cohort and 25 predictors were analyzed for building predictive models. The CART model had the highest sensitivity of 0.99, while CHAID, CIT, and CART models had a specificity of 0.64, 0.73, and 0.47, respectively. Moreover, the CART algorithm had acceptable performance; the area under the receiver operating characteristic curve (AUC) was 0.736. Moreover, the F1 score of the model of CART was 0.98. Conclusions: In summary, tree-based algorithms demonstrated acceptable prognostication performance in elderly TBI with outstanding sensitivity. These algorithms' models are straightforward and user-friendly, making them appropriate for use as screening tools in general practice. Future research could be focused on validating these models with unseen data and comparing predictability across different CPTs.