扁桃体切除术后患者出血的影响因素及列线图预警模型构建与验证

Construction and validation of a nomogram prediction model for postoperative bleeding in patients undergoing tonsillectomy

摘要:

目的 分析扁桃体切除术后患者出血的影响因素，并构建及验证列线图预警模型，以预测术后出血风险。方法 回顾性分析2018年1月—2023年1月在西安市人民医院耳鼻咽喉头颈外科行扁桃体切除术的508例患者临床资料。根据术后是否出血将患者分为出血组和未出血组，对两组患者的临床资料进行单因素分析，筛选出具有统计学意义的变量后进行多因素一般Logistic回归分析，构建列线图预警模型，绘制受试者工作特征（ROC）曲线，计算曲线下面积（AUC）以评估模型的预测效能。结果 两组患者性别构成、吸烟饮酒史、手术季节、既往反复发作炎症反应、术中出血量、手术位置、凝血酶原时间、活化部分凝血活酶时间、血小板计数、红细胞计数、中性粒细胞计数水平比较，差异均无统计学意义（P >0.05）。两组患者年龄、扁桃体包埋程度、术后感染、手术时间、术者资历、白细胞计数（WBC）比较，差异均有统计学意义（P <0.05）。多因素一般Logistic回归分析结果显示，年龄< 18岁［O^R=214.062（95% CI：2.536，18 065.451）］、重度扁桃体包埋［O^R=34.602（95% CI：1.592，751.844）］、有术后感染［O^R=16.817（95% CI：1.088，259.896）］、手术时间长［O^R=3.062（95% CI：1.588，5.907）］、术者手术经验< 6年［O^R=24.520（95% CI：1.795，334.98）］、WBC水平高［O^R=2.625（95% CI：1.161，5.935）］均是扁桃体切除术后患者出血的危险因素（P <0.05）。构建扁桃体切除术后患者出血的列线图模型结果显示，模型具有良好的拟合校准曲线（P <0.05）。ROC曲线分析结果显示，联合预测曲线下面积最大，为0.979（95% CI：0.950，1.000），敏感性最高，为95.2%（95% CI：0.762，0.999），特异性最高，为94.9%（95% CI：0.925，0.967）。结论 患者年龄< 18岁、重度扁桃体包埋、有术后感染、手术时间长、术者手术经验< 6年、WBC水平高是扁桃体切除术后患者出血的独立危险因素。该列线图预警模型可作为临床医生评估患者术后出血风险的有效工具。

Abstract:

Objective To analyze the factors influencing postoperative bleeding in patients undergoing tonsillectomy and to develop and validate a nomogram prediction model to estimate the risk of postoperative bleeding.Methods A retrospective analysis of the clinical data from 508 patients who underwent tonsillectomy in the Department of Otolaryngology-Head and Neck Surgery of Xi'an People's Hospital from January 2018 to January 2023 was conducted. Patients were divided into bleeding and non-bleeding groups based on the occurrence of postoperative bleeding. Univariable analysis was performed on clinical data of both groups to identify variables with statistical significance, which were then included in multivariable Logistic regression analysis to construct a nomogram prediction model. Receiver operating characteristic (ROC) curve was plotted and the area under the curve (AUC) was calculated to evaluate the predictive efficiency of the model.Results There was no statistically significant difference between the two groups in terms of sex composition, history of smoking and alcohol consumption, season of surgery, history of recurrent inflammatory reactions, intraoperative blood loss, surgical sites, prothrombin time, activated partial thromboplastin time, platelet count, red blood cell count, and neutrophil count (P > 0.05). The age, degree of tonsil embedding, postoperative infections, surgical duration, surgeon's experience, and white blood cell (WBC) count were different between the two groups (P < 0.05). The multivariable Logistic regression analysis indicated that age < 18 years [O^R = 214.062 (95% CI: 2.536, 18 065.451)], severe tonsil embedding [O^R = 34.602 (95% CI: 1.592, 751.844) ], postoperative infections [O^R=16.817 (95% CI: 1.088, 259.896) ], long surgical duration [O^R = 3.062 (95% CI: 1.588, 5.907) ], surgeon's experience < 6 years [O^R = 24.520 (95% CI: 1.795, 334.98) ] and high WBC count [O^R = 2.625 (95% CI: 1.161, 5.935) ] were risk factors for postoperative bleeding in patients undergoing tonsillectomy (P < 0.05). The nomogram model for postoperative bleeding after tonsillectomy demonstrated a good fit as assessed with the calibration curve (P < 0.05). The ROC curve analysis exhibited that the AUC of the combined detection was the highest being 0.979 (95% CI: 0.950, 1.000), with the highest sensitivity being 95.2% (95% CI: 0.762, 0.999), and the highest specificity being 94.9% (95% CI: 0.925, 0.967).Conclusions Age under 18 years, severe tonsil embedding, postoperative infections, prolonged surgical duration, surgeon's experience less than 6 years, and high WBC count are independent risk factors for postoperative bleeding in patients undergoing tonsillectomy. The nomogram prediction model serves as an effective tool for clinicians to assess the risk of postoperative bleeding.