Identifying patient subgroups and processes of care that cause outcome differences following ICU vs. ward triage among patients with acute respiratory failure and sepsis

项目总结 决定收治患有急性呼吸衰竭(ARF)和脓毒症(最常见和致命的原因)的患者 (急性呼吸窘迫综合征)到重症监护病房(ICU)在全美范围内差异很大。 然而，这些分诊决定对患者的预后有重大影响。在我们之前的工作中，我们使用了 920万住院患者的详细电子健康记录(EHR)数据，并发现入院决定 ARF患者到病房的死亡率绝对值增加3.8%。相比之下，选择承认 脓毒症患者到ICU的住院时间显著延长，死亡率绝对增加5.1%。 这种可自由决定的分类在全国范围内的影响将是指数级的。我们的发现强调了 通过确定患者亚组和治疗方法改善ARF和脓毒症预后的巨大机会 与病房护理相比，对ICU护理的益处和危害贡献最大的护理流程。 此应用程序建议更新我们的ARF和脓毒症队列，使其包括2013年以来的所有入院 到2022年，在北加州凯撒永久医院和宾夕法尼亚大学的29家医院 医疗系统，并将100多个数据字段合并到每个患者。这种高度颗粒化的精选 电子病历数据将使我们能够识别：(1)这些会议中不同的患者亚组和表型 ‘ARF’和‘败血症’的症状标准；以及(2)护理过程和(3)住院并发症 因果性地解释观察到的ICU与病房分诊与患者预后的关系。我们的多学科 团队将在工具变量回归、中介分析、机器学习、 复杂的电子病历数据和概率表型，以完成三个目标，促进我们的长期目标 改善ARF和脓毒症患者的护理，从而改善结果，无论他们在哪里接受治疗。 若干方法创新将使我们能够实现这些目标，进而不仅超越 先前的研究试图确定哪些危重患者从ICU入院中受益， 但找出这种分流效应背后的机制。这些数据还将使我们能够量化影响 新冠肺炎对急性肾功能衰竭和脓毒症患者的ICU和病房分诊模式、护理过程和预后的影响 从而使我们的成果现代化，并使其适用于大流行时代。 完成这项研究的目标将通过确定紧急情况下 科室、ICU和病房可以改善每个住院的400多万美国人的结局 患有ARF和/或脓毒症的一年。这样的结果将使个性化分类的开发和测试成为可能 算法，并指导患者的最佳护理，而不总是需要ICU入院，从而提高 减少医疗成本，并为真正需要的患者保留ICU能力。

PROJECT SUMMARY Decisions to admit patients with acute respiratory failure (ARF) and sepsis (the most common and lethal cause of the acute respiratory distress syndrome) to intensive care units (ICUs) are highly variable across the US. And, yet, these triage decisions have a substantial impact on patient outcomes. In our prior work, we used detailed electronic health record (EHR) data from 9.2 million hospitalizations and found that decisions to admit ARF patients to wards were associated with a 3.8% absolute increase in mortality. In contrast, choices to admit sepsis patients to ICUs resulted in considerably longer length of stay and a 5.1% absolute increase in death. The nationwide impact of such discretionary triage would be exponentially greater. Our findings highlight tremendous opportunities to improve ARF and sepsis outcomes by identifying the patient subgroups and processes of care that most strongly contribute to the benefits and harms of ICU- versus ward-based care. This application proposes to update our ARF and sepsis cohort such that it includes all admissions from 2013 through 2022 across 29 hospitals in the Kaiser Permanente Northern California and University of Pennsylvania health systems, and incorporate more than 100 more data fields per patient. This curation of highly granular EHR data will enable us to identify the: (1) distinct patient subgroups and phenotypes among those meeting the syndromic criteria of `ARF' and `sepsis;' and the (2) processes of care and (3) inpatient complications that causally explain the observed associations of ICU vs. ward triage with patient outcomes. Our multidisciplinary team will apply diverse expertise in instrumental variable regression, mediation analyses, machine learning, complex EHR data, and probabilistic phenotyping to complete three aims that promote our long-term goal of improving care, and hence outcomes, for patients with ARF and sepsis regardless of where they are treated. Several methodological innovations will enable us to achieve these goals, and, in turn, to not only surmount key limitations of prior studies that sought to determine which acutely ill patients benefit from ICU admission, but identify the mechanisms underlying such triage effects. These data will also allow us to quantify the impact of COVID-19 on ICU and ward triage patterns, care processes, and outcomes among ARF and sepsis patients, thereby modernizing our results and enabling their applicability to pandemic eras. Completing the aims of this study will improve public health by identifying ways in which emergency departments, ICUs, and wards can improve outcomes for the more than 4 million Americans hospitalized each year with ARF and/or sepsis. Such results will enable development and testing of personalized triage algorithms, and guide optimal care for patients without always requiring ICU admission, thereby improving patient outcomes, reducing health care costs, and preserving ICU capacity for patients who truly need it.