骨盆髋臼骨折术后并发症发生概率预测模型的建立

Establishment of a model for predicting the postoperative morbidity in patients with pelvic and acetabulum fractures

目的建立骨盆髋臼骨折术后并发症发生概率的预测模型。方法回顾分析我院近10年来94例骨盆髋臼骨折手术病历资料完整病例,选择与并发症相关生理学及手术严重度评分为变量,以Logistic回归分析计算其回归系数,建立其并发症发生概率预测模型,以ROC曲线分析检测其预测价值。结果本组资料数据建立的骨盆髋臼骨折并发症发生概率预测模型为:lnP/(1-P)=-13.83+0.47×PS+0.33×OS(P:预测并发症发生概率;PS:术前生理学评分;OS:手术严重度评分),ROC曲线下面积0.888±0.037、截断点P=37%、灵敏度84.21%、特异度82.14%、符合率82.98%、误判率15.79%及漏判率17.86%。结论本组资料建立的预测模型对骨盆髋臼骨折术后并发症发生概率有较高的预测能力,可指导临床治疗,降低手术风险。

Objective To establish a model for predicting the postoperative morbidity in patients with pelvic and acetabulum fractures. Methods The clinical of 94 patients with pelvic and acetabulum fractures were retrospectively reviewed for analysis of the postoperative complications and duration of hospital stay. The preoperative physiological scores （PS） and operative severity scores （OS） of the patients were selected as the variables, and their regression coefficients were calculated by logistic regression analysis to establish the model for predicting the postoperative morbidities. The predictive value of the model was evaluated according to the ROC curve. Results The model for predicting postoperative morbidities of pelvic and acetabulum fractures, In P/（ 1-P）=-13.83＋0.47×PS＋0.33×OS, has an area under ROC curve of 0.888±0.037 with the cut-offpoint of 37%, sensitivity of 84.21%, specificity of 82.14%, concordance rate of 82.98%, misclassification rate of 15.79% and omission classification rate of 17.86%. Conclusion This model possesses a high predictive capability for postoperative morbidities of pelvic and acetabulum fractures, and can be helpful for clinical therapeutic decision-making and lower the risks in operation.