Rapid Clinical Snapshots from the EMR among Pneumonia Patients

描述（申请人提供）：目前对住院患者的研究主要集中在重症监护室（ICU）的患者，这些患者是电子病历（EMR）的早期采用者。由于人工提取生理数据，特别是生命体征的成本很高，对普通内科-外科病房患者的研究受到限制。由于缺乏此类数据，临床医生缺乏定量工具来衡量不同时间点的医院护理过程，更不用说特定患者在住院期间的不同时间点可能具有的恶化风险水平了。本项目采用住院EMR提供这样的工具。它侧重于具有极高发病率和死亡率的特定患者人群：患有社区获得性肺炎（CAP）的住院成人，他们经历了计划外转移到更高水平的护理（例如，从普通内科外科病房到ICU）。大约70%的转移发生在医院的前72小时，这些患者的死亡率在10%至40%之间，严重程度调整后的观察到的预期死亡率高达16。我们的长期目标是利用住院EMR的力量进行质量监控，质量改进，并确定旨在防止住院恶化的有效实践和干预措施。为了实现这一目标，我们有两个具体目标：（1）使用病例队列方法，我们将开发适合嵌入EMR的模型，以预测最初未入住ICU的CAP患者在入院后72小时内发生危重疾病。使用来自20家北方加州Kaiser Permanente医院的综合住院和门诊EMR数据，我们将确定一个约13，700名符合以下标准的住院患者队列：年龄≥ 18岁，入院诊断为CAP;并且不是仅因姑息治疗或舒适治疗而入院。危重疾病定义为（a）休克，（B）需要辅助通气的呼吸衰竭，和/或（c）心脏骤停。我们将使用大约8，800例患者住院记录（我们估计其中485例，即5.5%，将在72小时内发展为危重病）开发预测模型，并在大约4，900例患者记录（其中270例在72小时内发展为危重病）上对其进行验证。(2)使用特定目标1的结果，我们将生成CAP患者发生和未发生危重疾病的特征的时间限定的“快照”。“快照”将在入院时和住院12、24和48小时时表征CAP患者的以下方面：（a）他们的人口统计学、临床和生理特征（包括生命体征、实验室检查结果和疾病严重程度评分）;（B）关键护理过程，例如是否和何时进行特定检查（例如，胸部X线照片、脉搏血氧仪、乳酸盐）和干预（例如，提供补充氧气，全身抗生素治疗，静脉输液）;和（c）他们的住院结果，包括恶化，死亡，住院时间（LOS），以及幸存者的出院处置。

DESCRIPTION (PROVIDED BY APPLICANT): Current research on hospitalized patients has focused on patients in intensive care units (ICUs), which have been early adopters of electronic medical records (EMRs). Research on general medical-surgical ward patients has been limited due to the high cost of manual abstraction of physiologic data, particularly vital signs. Given the paucity of such data, clinicians lack quantitative tools to gauge the process of hospital care at different points in time, let alone the level of risk of deterioration a given patient may have at different points in the course of a hospital stay. This project employs the inpatient EMR to provide such tools. It focuses on a specific patient population that has extremely high morbidity and mortality: hospitalized adults with community-acquired pneumonia (CAP) who experience an unplanned transfer to a higher level of care (e.g., from a general medical surgical ward to the ICU). Approximately 70% of these transfers occur in the first 72 hours in the hospital, and death rates among these patients range from 10 to 40%, with severity-adjusted observed to expected mortality ratios as high as 16. Our long term goal is to harness the power of the inpatient EMR for quality monitoring, quality improvement, and the identification of effective practices and interventions designed to prevent in-hospital deterioration. To achieve this goal, we have two specific aims: (1) Using a case-cohort methodology, we will develop models, suitable for embedding in the EMR, to predict the occurrence of critical illness within 72 hours of hospital admission among CAP patients who were not initially admitted to the ICU. Using comprehensive inpatient and outpatient EMR data from 20 Northern California Kaiser Permanente hospitals, we will identify a cohort of approximately 13,700 hospitalized patients meeting the following criteria: age =18 years, admission diagnosis of CAP; and not admitted only for palliative or comfort care. Critical illness is defined as (a) shock, (b) respiratory failure requiring assisted ventilation, and/or (c) cardiac arrest. We will develop predictive models using approximately 8,800 patient hospitalization records (of which we estimate 485, or 5.5%, will develop critical illness within 72 hours) and validate them on approximately 4,900 patient records (with 270 developing critical illness within 72 hours). (2) Using the results of Specific Aim 1, we will generate time-delimited "snapshots" of the characteristics of CAP patients who did and who did not develop critical illness. The "snapshots" will characterize CAP patients on admission and at 12, 24, and 48 hours into their hospital stay with respect to (a) their demographic, clinical, and physiologic characteristics (including vital signs, laboratory test results, and severity of illness scores); (b) key processes of care, such as whether and when specific tests (e.g., chest roentgenograms, pulse oximetry, lactates) and interventions (e.g., provision of supplemental oxygen, treatment with systemic antibiotics, intravenous fluid boluses) were performed; and (c) their hospital outcomes, including deterioration, death, length of stay (LOS), and discharge disposition for survivors.