构建颅脑损伤（TBI）去骨瓣减压术后新发出血的风险预测列线图模型。

选取重庆市开州区中医院外二科自2020年1月至2022年12月行去骨瓣减压术治疗的187例TBI患者，记录术后新发出血情况。将患者按照术后是否新发出血分为新发出血组和非新发出血组，采用二元Logistic回归分析筛选新发出血的危险因素。通过危险因素构建TBI去骨瓣减压术后新发出血的风险预测模型，采用重采样1000次的bootstrap进行内部验证（验证集63例）。采用Hosmer-Lemeshow评估模型的拟合优度，受试者工作特征（ROC）曲线分析模型的价值，校准曲线验证模型的准确性，决策曲线评估模型的临床获益情况。

187例患者中，新发出血组23例（12.3%），非新发出血组164例（87.7%），2组患者的合并颅骨骨折占比、术前血肿量、术前Rotterdam CT评分、术前伴硬膜下血肿占比、脑疝占比、术后低血压占比、手术时间、受伤至手术时间、凝血酶时间比较，差异有统计学意义（P<0.05）。二元Logistic回归分析显示，合并颅骨骨折、术前血肿量≥20 cm³、Rotterdam CT评分4~6分、术后低血压、手术时间长是术后新发出血的独立危险因素。将危险因素通过R语言构建预测模型，Hosmer-Lemeshow检验显示训练集校准度良好（χ²=1.944，P=0.963）。ROC分析显示，训练集和验证集的AUC分别为0.905（95%CI：0.849~0.960）、0.925（95%CI：0.839~1.000）。校准曲线与理想曲线拟合良好，提示预测模型准确性高；决策曲线分析显示，该模型预测TBI去骨瓣减压术后新发出血的净获益区间均较大。

合并颅骨骨折、术前血肿量≥20 cm³、Rotterdam CT评分4~6分、术后低血压、手术时间长是术后新发出血的影响因素，基于此构建的风险预测模型可为临床评估术后新发出血风险提供辅助参考。

To construct a nomogram model for predicting the risk of new bleeding for traumatic brain injury (TBI) after decompressive craniectomy.

A total of 187 TBI patients who underwent decompressive craniotomy at Second Department of Surgery of Chongqing Kaizhou Hospital of Traditional Chinese Medicine from January 2020 to December 2022 were selected, the occurrence of postoperative new blood emission was recorded. Patients were divided into newly generated blood group and non-newly generated blood group based on whether they experienced new bleeding after surgery, and binary Logistic regression analysis was used to screen for risk factors for new blood emission. The risk prediction model of new blood flow after TBI was constructed by risk factors. Internal validation was performed using bootstrap resampled 1000 times (validation set of 63 cases). Hosmer-Lemeshow was used to evaluate the fit degree of the model, receiver operating characteristic (ROC) curve was used to analyze the value of the model, calibration curve was used to verify the accuracy of the model, and decision curve was used to evaluate the clinical benefits of the model.

Among the 187 patients, there were 23 cases (12.3%) in the newly generated blood group and 164 cases (87.7%) in the non- newly generated blood group. There were differences in the combined skull fracture, preoperative hematoma volume, preoperative Rotterdam CT score, preoperative subdural hematoma, cerebral hernia, postoperative hypotension, surgery time, injury to surgery time, and thrombin time between the two groups of patients (P<0.05). Binary Logistic regression analysis confirmed that combined skull fractures, preoperative hematoma volume≥20 cm³, Rotterdam CT score (4-6 points), postoperative hypotension, and surgery time were independent risk factors for new postoperative bleeding. Constructing a predictive model for risk factors using R language, Hosmer Lemeshow test showed good calibration of the training set (χ²=1.944, P=0.963). ROC analysis showed that the AUC of the training set and validation set were 0.905 (95%CI: 0.849-0.960) and 0.925 (95%CI: 0.839-1.000), respectively. The calibration curve fits well with the ideal curve, indicating high accuracy of the prediction model; The decision curve analysis shows that the model predicts a relatively large net benefit range for new bleeding after decompressive craniectomy.

Combined with skull fracture, preoperative hematoma ≥20 cm³, Rotterdam CT score (4-6 points), postoperative hypotension, and long operation time are the influencing factors for postoperative new blood emission. The risk prediction model based on this construction can provide an auxiliary reference for clinical assessment of the risk of postoperative new blood emission.