RWJBarnabas将实时Epic EDI警报直接联动快速响应团队
RWJBarnabas Links Real-Time Epic EDI Alerts Directly to Rapid Response Teams
您需要了解的要点
由RWJBarnabas Health和Rutgers Robert Wood Johnson Medical School牵头的一项多医院、23,132名患者的研究（DOI: 10.1056/AIoa2500973）评估了Epic恶化指数（EDI）的企业级推广。
在11家急症护理医院实施后，高风险患者（EDI评分>60）的院内死亡率从23.1%降至18.6%，风险调整后死亡几率降低18%。
并非仅依赖被动的EHR仪表板，高风险“红色警报”会触发自动、即时的推送通知，发送至医院快速响应团队（RRT）携带的移动设备。
高风险患者的RRT评估率从25.3%显著提升至37.5%，但转入重症监护病房（ICU）的数量并未激增——这证明更早的床旁评估可防止不必要的重症监护升级。
RWJBarnabas与Rutgers如何扩展Epic的EDI以将院内死亡率降低18%医疗系统信息学、重症护理运营和临床AI领域面临一个已被充分证实的部署差距。尽管数百种机器学习早期预警系统（EWS）已被回顾性验证可预测患者恶化，但证明降低院内死亡率的真实世界前瞻性研究仍然极其罕见。
大多数临床AI干预停滞不前，是因为它们依赖于碎片化、被动的通知：评分在桌面屏幕上变化，但忙碌的床旁护士和主治医生直到生理崩溃已经发生时才注意到信号。
此外，未校准的警报可能引发“警报疲劳”或使重症监护病房（ICU）因不必要的转入而过载。为了将算法预测与即时临床行动相连接，RWJBarnabas Health和Rutgers Robert Wood Johnson Medical School的研究人员在11家医院执行了为期多年的Epic恶化指数（EDI）系统范围实施。
他们的23,132名患者研究发表于NEJM AI，证明将实时EHR预测评分与自动化的快速响应团队（RRT）移动推送通知相结合，可使风险调整后院内死亡率降低18%。
15分钟重新计算与移动RRT推送基础设施该举措将标准EHR供应商模型转变为自动化、系统范围的早期干预管道：
15分钟动态重新计算：
改良版Epic恶化指数持续分析31个EHR变量——包括生命体征趋势、实验室结果、护理评估和患者年龄——每15分钟重新计算恶化评分。
校准的风险分诊阈值：
分级评分将患者分为绿色（<30）、黄色（30–59）和红色（>60）风险带。评分>60对严重恶化或死亡具有较高的阳性预测价值。
统一移动推送集成：
当患者达到“红色警报”阈值（>60）时，EHR自动绕过静态桌面视图，向值班快速响应团队的移动设备发送即时推送通知。
警报疲劳抑制逻辑：
内置抑制规则会阻止重复的推送警报，如果患者已在接受ICU级别护理、舒适护理，或在6小时内已触发最近的快速响应或脓毒症警报。
“我们的目标是在患者到达干预变得更为困难的节点之前，更早地识别他们，”RWJBarnabas Health卫生信息学副总裁兼重症护理医师、Rutgers Robert Wood Johnson Medical School助理教授Thomas Nahass医学博士表示。“恶化指数给了我们一个更早的时间点。如果我们能更早让重症护理专家关注患者，我们就能改变他们的预后轨迹。”通过证明自动化的RRT推送通知增加了快速响应评估（从25.3%提升至37.5%）而未导致ICU转入增加，RWJBarnabas团队证明了早期床旁重症护理干预可以在内科-外科病房逆转患者恶化。
随着RWJBarnabas Health和Rutgers进入下一阶段——专注于速度追踪，以在患者风险评分达到红色阈值之前捕捉快速上升的风险——他们为全国各医疗系统提供了一个宝贵的、可复制的运营蓝图，帮助将EHR数据转化为挽救生命。
What You Should Know
A multi-hospital, 23,132-patient study led by RWJBarnabas Health and Rutgers Robert Wood Johnson Medical School (DOI: 10.1056/AIoa2500973) evaluated an enterprise rollout of the Epic Deterioration Index (EDI).
Following implementation across 11 acute care hospitals, in-hospital mortality among high-risk patients (EDI score >60)dropped from 23.1% to 18.6%, representing an 18% reduction in the risk-adjusted odds of death.
Rather than relying solely on passive EHR dashboards, high-risk “red alerts” triggered automated, immediate push notifications to mobile devices carried by hospital Rapid Response Teams (RRTs).
RRT evaluations among high-risk patients increased significantly from 25.3% to 37.5%, yet transfers to intensive care units (ICUs) did not surge—proving that earlier bedside evaluation prevents unnecessary critical care escalation.
How RWJBarnabas & Rutgers Scaled Epic’s EDI to Cut In-Hospital Mortality by 18% The health system informatics, critical care operations, and clinical AI sectors face a well-documented deployment gap. While hundreds of machine learning early warning systems (EWSs) have been retrospectively validated to predict patient deterioration, real-world prospective studies demonstrating reduced in-hospital mortality remain exceptionally rare.
Most clinical AI interventions stall because they rely on fragmented, passive notifications: a score changes on a desktop screen, but busy bedside nurses and attending physicians miss the signal until physiological breakdown has already occurred.
Furthermore, uncalibrated alerts risk triggering “alert fatigue” or over-saturating Intensive Care Units (ICUs) with unnecessary transfers. To bridge algorithmic prediction with immediate clinical action, researchers at RWJBarnabas Health and Rutgers Robert Wood Johnson Medical School executed a multi-year, systemwide implementation of the Epic Deterioration Index (EDI) across 11 hospitals.
Published in NEJM AI, their 23,132-patient study proved that pairing real-time EHR predictive scoring with automated Rapid Response Team (RRT) mobile push notifications drove an 18% reduction in risk-adjusted in-hospital mortality.
15-Minute Recalculations and Mobile RRT Push Infrastructure
The initiative transformed a standard EHR vendor model into an automated, systemwide early intervention pipeline:
15-Minute Dynamic Recalculation:
The modified Epic Deterioration Index continuously analyzes 31 EHR variables—including vital sign trends, laboratory results, nursing assessments, and patient age—recalculating deterioration scores every 15 minutes.
Calibrated Risk Triage Thresholds:
Tiered scores categorize patients into Green (<30), Yellow (30–59), and Red (>60) risk bands. A score of >60 carries a high positive predictive value for severe decline or death.
Unified Mobile Push Integration:
When a patient hits the “Red Alert” threshold (>60), the EHR automatically bypasses static desktop views to send an instant push notification directly to the mobile devices of on-duty Rapid Response Teams.
Alert Fatigue Suppression Logic:
Built-in suppression rules block redundant push alerts if the patient is already receiving ICU-level care, comfort care, or has triggered a recent rapid response or sepsis alert within 6 hours.
“Our goal was to identify patients earlier, before they reached a point where intervention becomes much more difficult,” stated Thomas Nahass, MD, VP of Health Informatics and intensive care physician at RWJBarnabas Health, and Assistant Professor at Rutgers Robert Wood Johnson Medical School. “The deterioration index gives us an earlier point in time. If we can get a critical care eye on the patient sooner, we can change the course of their outcome.” By demonstrating that automated RRT push notifications increased rapid response evaluations (from 25.3% to 37.5%) without swelling ICU transfers, the RWJBarnabas team proved that early bedside critical-care interventions can reverse patient deterioration right on the medical-surgical floor.
As RWJBarnabas Health and Rutgers advance into the next phase—focusing on velocity tracking to catch patients whose risk scores are rising rapidly before reaching the red threshold—they provide an invaluable, replicable operational blueprint for health systems nationwide seeking to convert EHR data into saved lives.
Reader Interactions