Singh Mandeep, Arruda-Olson Adelaide M, Lewis Bradley R, Johnson Bradley K, Chaudhry Rajeev, Arghami Arman, Alkhouli Mohamad, Rihal Charanjit S
Department of Cardiovascular Medicine, Mayo Clinic, Rochester, Minnesota, USA.
Department of Cardiovascular Medicine, Mayo Clinic, Rochester, Minnesota, USA.
JACC Adv. 2025 Jun 25;4(7):101865. doi: 10.1016/j.jacadv.2025.101865.
Automated individualized risk prediction tools linked to electronic health records (EHRs) are not available for patients undergoing percutaneous coronary interventions (PCIs).
Our goal was to automatically extract data elements used in the Mayo Clinic PCI models from EHR to enable point of care risk assessment.
Using the Mayo Clinic PCI registry, variables in the Mayo Clinic PCI risk score were trained and tested in an EHR to identify in-hospital death, stroke, bleeding, acute kidney injury (AKI) in patients who underwent PCI from 2016 to 2024. Least absolute shrinkage and selection operator regression was utilized to train (data building) and test (assessing performance) prediction models and to estimate effect sizes that were weighted and integrated into a scoring system.
Death, stroke, bleeding, AKI occurred in 157 (1.8%), 43 (0.5%), 157 (1.8%), and 682 (7.6%), respectively. The C-statistics (95% CI) from the training and testing data sets were 0.83 (95% CI: 0.80-0.86) and 0.84 (95% CI: 0.78-0.89); 0.76 (95% CI: 0.65-0.84) and 0.77 (95% CI: 0.65-0.86); 0.80 (95% CI: 0.75-0.83) and 0.75 (95% CI: 0.68-0.81); and 0.82 (95% CI: 0.80-0.84) and 0.80 (95% CI: 0.77-0.84) for in-hospital death, stroke, bleeding, and AKI, respectively. Bootstrap analysis indicated that the models were not overfit to the available data set. The probabilities estimated from the models matched the observed data well, as indicated by the calibration curve slope and intercept and across subgroups, including women, acute coronary syndrome, cardiogenic shock, and diabetes mellitus.
Real-time, automated, point of care PCI risk assessment is feasible in an EHR environment.
对于接受经皮冠状动脉介入治疗（PCI）的患者，尚无与电子健康记录（EHR）相关联的自动化个体化风险预测工具。
我们的目标是从电子健康记录中自动提取梅奥诊所PCI模型中使用的数据元素，以实现即时护理风险评估。
利用梅奥诊所PCI登记处的数据，在电子健康记录中对梅奥诊所PCI风险评分中的变量进行训练和测试，以识别2016年至2024年接受PCI治疗患者的院内死亡、中风、出血、急性肾损伤（AKI）情况。采用最小绝对收缩和选择算子回归来训练（数据构建）和测试（评估性能）预测模型，并估计效应大小，将其加权并整合到评分系统中。
死亡、中风、出血、AKI的发生例数分别为157例（1.8%）、43例（0.5%）、157例（1.8%）和682例（7.6%）。训练数据集和测试数据集的C统计量（95%CI）分别为：院内死亡为0.83（95%CI：0.80 - 0.86）和0.84（95%CI：0.78 - 0.89）；中风为0.76（95%CI：0.65 - 0.84）和0.77（95%CI：0.65 - 0.86）；出血为0.80（95%CI：0.75 - 0.83）和0.75（95%CI：0.68 - 0.81）；AKI为0.82（95%CI：0.80 - 0.84）和0.80（95%CI：0.77 - 0.84）。自助法分析表明，模型对可用数据集没有过度拟合。校准曲线斜率和截距以及在包括女性、急性冠状动脉综合征、心源性休克和糖尿病等亚组中显示，模型估计的概率与观察数据匹配良好。
在电子健康记录环境中，实时、自动化的即时护理PCI风险评估是可行的。
