髋部骨折手术预后的预测:Nottingham髋部骨折评分的独立验证和重新校准

Prognosis prediction after hip fracture surgery:independent validation and recalibration of the Nottingham Hip Fracture Score

目的探讨Nottingham髋部骨折评分(NHFS)在我国的应用价值,并建立适合我国髋部骨折患者术后30 d死亡风险的评估公式。方法回顾性分析2019年8月至2022年5月期间郑州大学第一附属医院骨科诊治的824例髋部骨折患者资料。男312例,女512例,年龄73(63,82)岁。比较不同生存结局患者的临床资料。根据患者NHFS,通过公式计算出患者术后30 d病死率,并与实际病死率进行比较,以验证原预测模型的有效性。将患者分为训练组(n=577)和验证组(n=247),对训练组患者采用二元logistic回归分析,建立新的预测模型。在训练组和验证组中对预测模型的区分度、校准度及临床有效性进行验证。结果多因素logistic回归分析结果显示:高龄(年龄≥86岁)(OR=3.775,95%CI:1.099~12.972,P=0.035)、男性(OR=3.151,95%CI:1.574~6.306,P=0.001)、入院血红蛋白浓度≤100 g/L(OR=2.402,95%CI:1.189~4.850,P=0.015)、入院前依赖他人看护(OR=2.673,95%CI:1.298~5.505,P=0.008)、合并症≥2种(OR=4.988,95%CI:1.874~13.274,P=0.001)是髋部骨折患者术后30 d死亡的危险因素。对原预测模型进行验证:C指数为0.764,区分度良好;原预测模型的预测病死率与实际病死率比较差异有统计学意义(P<0.05),校准度不佳。对预测模型重新校准:30 d病死率(%)=100/[1+e^(5.818-NHFS×0.599)],在训练组和验证组中对新预测模型进行验证:C指数分别为0.762和0.780,区分度良好;预测模型的预测病死率与实际病死率接近,差异无统计学意义(P>0.05),校准度良好;训练组和验证组阈值概率分别在0~26%和0~35%时干预患者可能受益,模型具有临床有效性。结论NHFS能对髋部骨折患者术后30 d死亡风险做出预测。校准后新的NHFS预测模型对我国髋部骨折患者术后30 d病死率有较好的预测价值。

Objective To investigate the application value of the Nottingham Hip Fracture Score(NHFS)in China and establish a formula specifically designed to assess the risk for 30-day mortality after surgery for hip fracture patients in China.Methods A retrospective study was conducted to analyze the clinical data of 824 hip fracture patients who had been treated at Department of Orthopaedics,The First Hospital Affiliated to Zhengzhou University from August 2019 to May 2022.There were 312 males and 512 females with a median age of 73(63,82)years.The clinical data were compared between patients with different survival outcomes.The 30-day mortality was calculated by the formula according to the patients'NHFS,and compared with the actual one to validate the effectiveness of the original prediction model.The patients were divided into a training group(n=577)and a validation group(n=247).Binary logistic regression analysis was performed to establish a new prediction model for the patients in the training group.The discrimination,calibration,and clinical effectiveness of the predictive model were assessed in both the training and validation groups.Results Multivariate logistic regression analysis showed that advanced age(≥86 years old)(OR=3.775,95%CI:1.099 to 12.972,P=0.035),male(OR=3.151,95%CI:1.574 to 6.306,P=0.001),admission hemoglobin concentration≤100 g/L(OR=2.402,95%CI:1.189 to 4.850,P=0.015),dependence on others for care before admission(OR=2.673,95%CI:1.298 to 5.505,P=0.008),and comorbidities≥2(OR=4.988,95%CI:1.874 to 13.274,P=0.001)were identified as risk factors for postoperative 30-day mortality(all P<0.05).In validation of the original prediction model,the C-index was found to be 0.764,indicating good discrimination.However,there was a significant discrepancy between the mortality forecast by the original prediction model and the actual mortality(P<0.05),indicating poor calibration.After the prediction model was recalibrated,30-day mortality(%)=100/[1+e^(5.818-NHFS×0.599)].After the new prediction model was validated in both the training and validation groups,the C-indexes were 0.762 and 0.780,indicating a good level of discrimination.The predicted 30-day mortality by the prediction model was closely aligned with the actual mortality(P>0.05),demonstrating good calibration.When the threshold probabilities of the training and the validation groups were 0 to 26%and 0 to 35%,respectively,the patients might benefit from clinical intervention,showing clinical effectiveness of the model.Conclusions The NHFS can predict the risk for 30-day mortality after hip fracture surgery.The new NHFS prediction model after calibration has a good predictive value for 30-day mortality after hip fracture surgery in Chinese population.