The Development of an EHR-based Measure of Orthopaedic Treatment Success

项目摘要 尽管骨科疾病的患病率和治疗费用很高， 非常少的I级证据支持治疗方法， 整形外科移位肱骨近端骨折（PHF）的治疗决定，第三个 老年人最常见的骨折，仍然具有挑战性，医生之间差异很大。 目前缺乏数据来评估用于以下疾病的矫形治疗的有效性： PHF患者因此，骨科医学结局评估的现状 仅限于过程测量和终点，如生存率和手术并发症。 不幸的是，这些数据点描绘了一个病人的医疗经验的不完整的画面 以及治疗是否成功地实现了患者优先考虑的结果目标。 患者感知获益指标，称为患者报告结局指标 （PROM）没有以标准化的方式收集。因此，迫切需要 有更好的骨科治疗有效性证据，以提高安全性和质量 为PHF提供的护理。我们对这个应用程序的总体目标是制定一个衡量标准， 从常规采集的电子健康记录（EHR）数据中获取骨科治疗成功率。 在每次交互过程中生成EHR系统中捕获的叙述性临床记录 在病人和医生之间，从而产生一个病人的历史记录，身体 调查结果、医学推理和病人护理。在骨科手术中， 临床文档，骨科医生撰写了一个不断发展的患者故事， 对骨科治疗的反应。临床笔记记录了改善或缓解的程度 患者直接经历和报告的情况，除了症状 尚未解决，正在徘徊，或出现后续并发症。一直 增加了研究，以促进自然语言处理（NLP）的使用 对在非结构化临床笔记中发现的医学概念进行分类的方法。深度学习NLP 模型已经用于临床文本分类，并且可以用于识别 经历治疗成功或失败。该项目的基本原理是， 从常规获取的EHR数据中衡量骨科治疗成功率， 我们如何评估骨科护理质量的范式转变， 治疗是否成功地实现了患者优先考虑的目标。一旦这个 项目完成并采用我们的新方法后，可以确定患者的结局 比黄金标准的PROM更容易、更有效、成本更低。

Project Summary Despite the high prevalence and treatment costs associated with orthopaedic conditions, there is remarkably little Level I evidence supporting the treatment approaches used in orthopaedics. Treatment decisions for displaced proximal humerus fractures (PHF), the third most common fracture in the elderly, remain challenging and highly varied between physicians. There is a dearth of data to evaluate the effectiveness of orthopaedic treatments used for patients with PHFs. Thus, the status quo for outcome assessment in orthopaedic medicine has been limited to process measures and end points such as survival and surgical complications. Unfortunately, these data points paint an incomplete picture of a patient’s medical experience and whether treatment was successful in achieving the outcome goals prioritized by the patient. Measures of patient-perceived benefits, known as Patient-Reported Outcome Measures (PROMs) are not collected in a standardized manner. Therefore, there is an urgent need to have better evidence on orthopaedic treatment effectiveness to improve the safety and quality of care provided for PHF. Our overall objective for this application is to develop a measure of orthopaedic treatment success from routinely captured electronic health record (EHR) data. Narrative clinical notes captured in EHR systems are generated during each interaction between patients and physicians, thereby producing a record of a patient’s history, physical findings, medical reasoning, and patient care. During orthopaedic encounters and through clinical documentation, orthopaedic surgeons author an evolving patient story of patient response to orthopaedic treatment. Clinical notes document the degree of improvement or relief experienced and reported directly by patients, in addition to scenarios in which symptoms have not been resolved, are lingering, or when subsequent complications have arisen. There has been an increase in research to advance the use of natural language processing (NLP) methods to classify medical concepts found in unstructured clinical notes. Deep learning NLP models have been used for clinical text classification and can be used to identify patients that experience treatment success or failure. The rationale for this project is that the development of a measure of orthopaedic treatment success from routinely captured EHR data will initiate a paradigm shift in how we evaluate the quality of orthopaedic care and enable assessment of whether treatment was successful in achieving the goals prioritized by patients. Once this project is completed and our new approach is adopted, patient outcomes can be determined more easily, more effectively, and with less cost than the gold-standard of PROMs.