Concussions are complex, as patients often present with nonspecific symptoms, requiring timely evaluation and accurate diagnosis. This study, using the National Health Interview Survey database, aimed to explore and validate a concussion risk model to support diagnostic decision-making and patient treatment supervision. This study included demographic and clinical data of 14,275 subjects in 2023. Predictive indicators were selected using least baseline characteristics and least absolute shrinkage and selection operator regression analysis, and a risk nomogram model was constructed. The model was evaluated using calibration curves, the area under curve of receiver operating characteristic, and decision curve analysis. The eligible concussion group (n = 363) and the nonconcussion group (n = 13,912) from the National Health Interview Survey database exhibited significant differences in 9 baseline characteristics (P <.05). Age, education level, general health, family income-to-poverty ratio, marital status, mental health, anxiety, behavior, and industry were found to be predictive indicators for patients with concussion. The model built using these predictive indicators demonstrated an area under curve of 0.712 in the receiver operating characteristic curve (95% CI: 0.68647 - 0.73671), indicating good predictive performance. The nomogram showed a strong association between the predicted and actual risks, with high calibration. Decision curve analysis confirmed strong discriminative ability of the model. The exploratory model based on 9 predictive indicators served as a valuable decision-making tool for clinicians. In concussion patients, these predictive indicators could be closely monitored in clinical practice, allowing for timely intervention to improve prognosis.
Yang et al. (Fri,) studied this question.