Meguid Robert A, Bronsert Michael R, Juarez-Colunga Elizabeth, Hammermeister Karl E, Henderson William G
*Surgical Outcomes and Applied Research program, University of Colorado School of Medicine, Aurora, CO†Department of Surgery, University of Colorado School of Medicine, Aurora, CO‡Adult and Child Center for Health Outcomes Research and Delivery Science, University of Colorado School of Medicine, Aurora, CO§Department of Biostatistics and Informatics, Colorado School of Public Health, Aurora, CO¶Division of Cardiology, Department of Medicine, University of Colorado School of Medicine, Aurora, CO.
Ann Surg. 2016 Jul;264(1):23-31. doi: 10.1097/SLA.0000000000001678.
To develop accurate preoperative risk prediction models for multiple adverse postoperative outcomes applicable to a broad surgical population using a parsimonious common set of risk variables and outcomes.
Currently, preoperative assessment of surgical risk is largely based on subjective clinician experience. We propose a paradigm shift from the current postoperative risk adjustment for cross-hospital comparison to patient-centered quantitative risk assessment during the preoperative evaluation.
We identify the most common and important predictor variables of postoperative mortality, overall morbidity, and 6 complication clusters from previously published prediction analyses that used forward selection stepwise logistic regression. We then refit the prediction models using only the 8 most common and important predictor variables, and compare the discrimination and calibration of these models to the original full-variable models using the c-index, Hosmer-Lemeshow analysis, and Brier scores.
Accurate risk models for 30-day outcomes of mortality, overall morbidity, and 6 clusters of complications were developed using a set of 8 preoperative risk variables. C-indexes of the 8 variable models are between 97.9% and 99.2% of those of the full models containing up to 28 variables, indicating excellent discrimination using fewer predictor variables. Hosmer-Lemeshow analyses showed observed to expected event rates to be nearly identical between parsimonious models and full models, both showing good calibration.
Accurate preoperative risk assessment of postoperative mortality, overall morbidity, and 6 complication clusters in a broad surgical population can be achieved with as few as 8 preoperative predictor variables, improving feasibility of routine preoperative risk assessment for surgical patients.
使用一组简约的常见风险变量和结局，为广泛的手术人群开发适用于多种不良术后结局的准确术前风险预测模型。
目前，手术风险的术前评估很大程度上基于临床医生的主观经验。我们提议从当前用于跨医院比较的术后风险调整模式，转变为术前评估期间以患者为中心的定量风险评估。
我们从先前发表的使用向前选择逐步逻辑回归的预测分析中，确定术后死亡率、总体发病率和6个并发症集群最常见且重要的预测变量。然后仅使用8个最常见且重要的预测变量重新拟合预测模型，并使用c指数、Hosmer-Lemeshow分析和Brier评分，将这些模型的区分度和校准度与原始的全变量模型进行比较。
使用一组8个术前风险变量，开发出了针对30天死亡率、总体发病率和6个并发症集群结局的准确风险模型。8变量模型的c指数在包含多达28个变量的全模型的c指数的97.9%至99.2%之间，表明使用较少的预测变量具有出色的区分度。Hosmer-Lemeshow分析显示，简约模型和全模型的观察事件率与预期事件率几乎相同，两者均显示出良好的校准度。
对于广泛的手术人群，仅需8个术前预测变量就能实现对术后死亡率、总体发病率和6个并发症集群的准确术前风险评估，提高了外科患者常规术前风险评估的可行性。
