de Jong Louis, de Haan Eveline, van Rijckevorsel Veronique A J I M, Kuijper T Martijn, Roukema Gert R
Department of Surgery, Maasstad Hospital, Rotterdam, The Netherlands.
Department of Surgery, Franciscus Hospital, Rotterdam, The Netherlands.
J Bone Joint Surg Am. 2025 Mar 5;107(5):459-468. doi: 10.2106/JBJS.23.01397. Epub 2025 Jan 21.
The aim of this study was to develop an accurate and clinically relevant prediction model for 30-day mortality following hip fracture surgery.
A previous study protocol was utilized as a guideline for data collection and as the standard for the hip fracture treatment. Two prospective, detailed hip fracture databases of 2 different hospitals (hospital A, training cohort; hospital B, testing cohort) were utilized to obtain data. On the basis of the literature, the results of a univariable analysis, and expert opinion, 26 candidate predictors of 30-day mortality were selected. Subsequently, the training of the model, including variable selection, was performed on the training cohort (hospital A) with use of adaptive least absolute shrinkage and selection operator (LASSO) logistic regression. External validation was performed on the testing cohort (hospital B).
A total of 3,523 patients were analyzed, of whom 302 (8.6%) died within 30 days after surgery. After the LASSO analysis, 7 of the 26 variables were included in the prediction model: age, gender, an American Society of Anesthesiologists score of 4, dementia, albumin level, Katz Index of Independence in Activities of Daily Living total score, and residence in a nursing home. The area under the receiver operating characteristic curve of the prediction model was 0.789 in the training cohort and 0.775 in the testing cohort. The calibration curve showed good consistency between observed and predicted 30-day mortality.
The Rotterdam Hip Fracture Mortality Prediction-30 Days (RHMP-30) was developed and externally validated, and showed adequate performance in predicting 30-day mortality following hip fracture surgery. The RHMP-30 will be helpful for shared decision-making with patients regarding hip fracture treatment.
Prognostic Level II . See Instructions for Authors for a complete description of levels of evidence.
本研究的目的是开发一种用于预测髋部骨折手术后30天死亡率的准确且具有临床相关性的预测模型。
将先前的研究方案用作数据收集指南以及髋部骨折治疗标准。利用两家不同医院（医院A，训练队列；医院B，测试队列）的两个前瞻性、详细的髋部骨折数据库来获取数据。基于文献、单变量分析结果和专家意见，选择了26个30天死亡率的候选预测指标。随后，使用自适应最小绝对收缩和选择算子（LASSO）逻辑回归在训练队列（医院A）上进行模型训练，包括变量选择。在测试队列（医院B）上进行外部验证。
共分析了3523例患者，其中302例（8.6%）在手术后30天内死亡。经过LASSO分析，26个变量中的7个被纳入预测模型：年龄、性别、美国麻醉医师协会评分为4、痴呆、白蛋白水平、日常生活活动能力Katz独立指数总分以及居住在养老院。预测模型在训练队列中的受试者操作特征曲线下面积为0.789，在测试队列中为0.775。校准曲线显示观察到的和预测的30天死亡率之间具有良好的一致性。
开发并外部验证了鹿特丹髋部骨折死亡率预测-30天（RHMP-30）模型，该模型在预测髋部骨折手术后30天死亡率方面表现良好。RHMP-30将有助于在髋部骨折治疗方面与患者进行共同决策。
预后II级。有关证据水平的完整描述，请参阅作者指南。
