To analyze risk factors of prolonged length of stay (PLOS) in patients with moderate and severe traumatic brain injury (TBI), and to construct a prediction model.

A retrospective cohort study was performed to analyze the clinical data of 533 patients with moderate and severe TBI admitted to the Neurosurgery Department of Fuyang Fifth People's Hospital from January 2018 to January 2023, which were divided into a training set (n=374) and an validation set (n=159) according to the ratio of 7 to 3. Patients in the training set were grouped into two groups according to the hospital stay, namely PLOS group (hospital stay≥28 d) and non-PLOS group (hospital stay<28 d). Multivariate Logistic regression analyses were used to assess the risk factors of PLOS. The R software was used to establish a nomogram prediction model based on the above risk factors. The receiver operating characteristic (ROC) curve, calibration curve and clinical decision curve analysis (DCA) were plotted in the training set and the validation set, and Hosmer-Lemeshow goodness-of-fit test was performed.

The first CT scan time after injury, GCS score at admission, body temperature, subdural hematoma, plasma potassium level, serum calcium concentration, C-reactive protein, thyroxine, prothrombin time, activated partial thromboplastin time, fibrinogen level, D-dimer, and combined upper gastrointestinal hemorrhage were correlated with PLOS in patients with moderate and severe TBI. Logistic regression analysis showed that body temperature>36.82℃, subdural hematoma, serum calcium concentration≤1.97 mmol/L, D-dimer>13.12 mg/L and combined upper gastrointestinal hemorrhage were independent risk factors of PLOS in moderate and severe TBI patients. The ROC of the nomogram prediction model indicated that area under curve (AUC) of the training set was 0.770 (95%CI: 0.699-0.840) and AUC of the validation set was 0.822 (95%CI: 0.754-0.889). The calibration curve showed that the predicted probability was consistent with the actual situation in both the training set and validation set. DCA showed that the nomogram prediction model presented excellent performance in predicting PLOS. In Hosmer-Lemeshow goodness-of-fit test, χ² value of the training set was 2.053 (P=0.979), with validation set of 4.566 (P=0.803).

Body temperature>36.82℃, subdural hematoma, serum calcium concentration≤1.97 mmol/L, D-dimer>13.12 mg/L and combined upper gastrointestinal hemorrhage were independent risk factors of PLOS in moderate and severe TBI patients. The nomogram prediction model based on these 5 predictive variables shows a good predictive performance, goodness-of-fit, and clinical applicability, which can provide a reference for clinical decision making.