Modeling informatics data to track maternal risk and care quality

摘要 虽然近几十年来孕产妇严重发病率和死亡率显著增加，但尚不清楚原因是什么。 遵循针对高风险临床场景的推荐安全实践，并降低不良反应的风险 产妇结局。减少孕产妇风险的一项关键战略是实施“安全包”， 统一的协议，以标准化的高风险临床条件的护理。虽然标准化的临床措施 这些捆绑包中支持的是基于证据的，但在实施、护理 产后出血、高血压病和高血压病的安全性方案的质量监测和结局评估 妊娠期疾病、败血症和严重异常生命体征（孕产妇预警系统）。产科 护理涉及服务、临床医生和资源的复杂协调， 跟踪结果并在真实的时间内大规模识别高质量护理的能力。尽管管理清晰 建议，孕产妇死亡率和安全审查已确定，护理不足往往发生， 其次是供应商偏离建议，系统问题，包括延迟识别， 反应，以及医院层面的影响，其中非最佳做法是正常的。这种程度 是否遵循指南以及是否可以避免不良后果尚不清楚，许多医院在 他们系统地审查护理的能力。从电子健康记录（EHR）收集的数据可能会 即时分析，以识别高危患者和并发症，并跟踪护理和管理， 人口数量。我们的研究组对超过40，000例分娩进行了关于产科出血的先前EHR研究 住院治疗表明，调整后的围产期子宫切除术的几率减少了一半， 实施出血安全捆绑包。该提案的首要假设是，EHR数据 能够可靠地识别与孕产妇抢救失败相关的临床管理因素， 紧急情况，如：（i）严重高血压，（ii）产科出血，（iii）败血症，和（iv）明显异常 产妇生命体征（产妇早期预警系统）。未能救助被定义为未能防止 临床上重要的恶化，如死亡或永久性残疾，从基础疾病或 医疗护理的并发症。我们将分析护理在多大程度上遵循捆绑建议和估计 不遵循指南时，存在抢救失败的风险。我们将利用丰富的EHR数据， 描述提供者的行为并对患者进行风险分层。我们的研究小组包括信息学，临床 研究、围产期流行病学、决策分析和生物统计学。来自八家医院的EHR数据 研究团队将进行分析。我们将描述临床管理，结果和护理质量， 严重的高血压，产科出血，败血症，以及明显的异常生命体征。这些分析的数据 将用于一些模拟，以告知临床试验和干预措施的开发。

ABSTRACT While maternal severe morbidity and mortality increased significantly over recent decades, it is unclear to what degree recommended safety practices for high-risk clinical scenarios are followed and reduce risk for adverse maternal outcomes. A key strategy to reducing maternal risk has been implementation of `safety bundles' and uniform protocols to standardize care for high-risk clinical conditions. While the standardized clinical measures supported in these bundles are evidence based, there are major knowledge gaps related to implementation, care quality surveillance, and outcomes assessment for safety protocols for postpartum hemorrhage, hypertensive diseases of pregnancy, sepsis, and grossly abnormal vital signs (maternal early warning systems). Obstetrical care involves complex coordination of services, clinicians, and resources, and leadership are limited in their ability to track outcomes and identify high quality care in real time at scale. Despite clear management recommendations, maternal mortality and safety reviews have identified that deficiencies in care often occur secondary to providers deviating from recommendations, systems issues including delayed identification and response, and hospital-level effects where non-optimal practices are normalized. The degrees to which guidelines are followed and adverse outcomes can be averted are not known, and many hospitals are limited in their ability to systematically review care. Data collected from the electronic health record (EHR) may be instantaneously analyzed to identify at-risk patients and complications and track care and management in large populations. Prior EHR research on obstetric hemorrhage by our study group of over 40,000 delivery hospitalizations demonstrated that adjusted odds for peripartum hysterectomy decreased by half after implementation of a hemorrhage safety bundle. The overarching hypothesis of this proposal is that EHR data can reliably identify clinical-management factors associated with failure to rescue in the setting of maternal emergencies such as: (i) severe hypertension, (ii) obstetric hemorrhage, (iii) sepsis, and (iv) frankly abnormal maternal vital signs (maternal early warnings systems). Failure to rescue is defined as a failure to prevent a clinically important deterioration, such as death or permanent disability, from an underlying illness or a complication of medical care. We will analyze to what degree care follows bundle recommendations and estimate risk for failure to rescue when guidelines are not followed. We will leverage the richness of EHR data to characterize provider behavior and risk stratify patients. Our study group includes expertise in informatics, clinical research, perinatal epidemiology, decision analysis, and biostatistics. EHR data from eight hospitals in a research consortium will be analyzed. We will characterize clinical management, outcomes, and care quality for severe hypertension, obstetric hemorrhage, sepsis, and frankly abnormal vital signs. Data from these analyses will be used for a number of simulations to inform development of clinical trials and interventions.