Development of a Novel Venous Thromboembolism Risk Assessment Model for Perioperative Clinical Decision Support

7.项目总结/摘要 静脉血栓栓塞事件（VTE）仍然是一个主要的可预防的发病率和死亡率的原因， 术后患者国家指南建议根据个体化风险开始预防 分层尽管有这个标准，多项研究，包括埃默里大学医院的研究进行了 我们小组的研究表明，大多数高危患者没有接受推荐的预防措施。标准化 迄今为止，由于各种原因，在外科手术患者中使用VTE风险评估模型（RAM）受到限制。 这些因素包括对风险评估过程的混淆，认为出血风险增加， 给予预防性抗凝药物，以及RAM的繁琐的额外工作流程 计算.例如，一项研究发现，详细的病史和非常规实验室检查的记录， Caprini评分所要求的每名患者的时间增加了6分钟，使医生的诊所工作流程。的潜力 这一过程的自动化是朝着病人护理迈出的令人兴奋的一步， 以患者为中心，但它需要建立在数据基础上的模型，这些数据在前 护理的操作阶段，以便可以前瞻性地制定护理计划。我们小组以前的研究 证明了容易获得的数据，如患者的年龄，BMI，种族和合并症负担-代替 更复杂和耗时的RAM -可以用于有效地对患者进行风险分层。 我们假设，实施一个标准化的，简短的，自动化的静脉血栓栓塞风险评估模型， 术前工作流程将通过增加指定的 预防，从而降低静脉血栓栓塞率和提高围手术期的效率。为了实现这一长期目标，我们 提出的具体目标包括：（1）多变量逻辑回归和决策树机的应用 学习国家手术质量改进项目数据库，以开发一种新的术前VTE RAM和（2）VTE预防指南的范围审查，随后提出标准化的 该协议将促进可操作的风险分层。未来的研究将致力于将RAM构建到 电子健康记录和随后实施拟议的议定书，包括自动化的 开发模型并跟踪对相关过程措施和职业技术教育结果的影响。前瞻性 患者水平的风险评估驱动术前工作流程，有可能使手术护理更 标准化，更以病人为中心，更公平。F32将促进分析方面的进一步培训 方法和转化研究，这将使我追求这些目标，一个经验，将是 在我作为外科医生科学家和质量改进领导者的职业生涯中，这是非常宝贵的。

7. Project Summary / Abstract Venous thromboembolic events (VTE) remain a leading preventable cause of morbidity and mortality in postoperative patients. National guidelines recommend initiation of prophylaxis based on individualized risk stratification. Despite this standard, multiple studies, including an Emory University Hospital study conducted by our group, have shown that most high-risk patients do not receive recommended prophylaxis. Standardized use of VTE risk assessment models (RAMs) in surgical patients has been limited to date due to a myriad of factors including confusion around the risk assessment process, a perceived increased risk of bleeding from administration of prophylactic anticoagulant medications, and the cumbersome additional workflow for RAM calculation. For example, one study found that the detailed history and the capture of non-routine lab tests as required by the Caprini score adds 6 minutes per patient to a physician’s clinic workflow. The potential for automation of this process is an exciting step toward patient care that is both more standardized and more patient-centered, but it requires development of models built on data that is readily available in the pre- operative phase of care such that a care plan can be enacted prospectively. Our group’s previous research demonstrated that readily available data such as a patient’s age, BMI, race, and comorbidity burden – in lieu of a more complex and time-consuming RAM – could be utilized to effectively risk-stratify patients. We hypothesize that implementation of a standardized, brief, and automated VTE risk assessment model in the pre-operative workflow will improve the quality of surgical care by increasing delivery of indicated prophylaxis, thereby decreasing VTE rates and improving perioperative efficiency. With that long-term goal, our proposed specific aims include (1) the application of multivariable logistic regression and decision tree machine learning to the National Surgical Quality Improvement Project database to develop a novel pre-operative VTE RAM and (2) a scoping review of VTE prophylaxis guidelines with the subsequent proposal of a standardized protocol that would facilitate actionable risk stratification. Future research will aim to build the RAM into the electronic health record and subsequent implementation of the proposed protocol, including automation of the developed model and tracking of impact on pertinent process measures and VTE outcomes. Prospective patient-level risk assessment driving pre-operative workflows has the potential to make surgical care more standardized, more patient-centered, and more equitable. The F32 will facilitate further training in analytic methods and translational research that will allow me to pursue these aims, an experience that will be invaluable on my path toward a career as a surgeon scientist and leader in quality improvement.