Early Recognition of In-Hospital Patient Deterioration Outside of The Intensive Care Unit: The Case For Continuous Monitoring Israeli Society of Internal Medicine Meeting July 5, 2013 Eyal Zimlichman MD, MSc Deputy Director and Chief Quality Officer Sheba Medical Center 1
Overview 1. Defining the problem 2. Importance of early recognition of clinical instability 3. Rapid Response Systems: rationale, design and current concerns 4. Effective monitoring the afferent limb of RRSs 5. A possible solution: The Earlysense continuous monitoring system 2
The Facts Patients treated in hospitals are becoming more and more complex, still, most are hospitalized in a non-icu setting. 10-20% of hospitalized patients develop complications and 5-8% of all patients die inhospital. Importantly, an estimated 37% of these events may be preventable 1. Up to 50% of cardiopulmonary arrests on the floors could be prevented by earlier ICU transfer. 1 Leape LL et al, N Engl J Med 1991, 324:377-384. 3
More Facts Cardiac Arrests/shock occur in 0.6% of medical patients and 0.5 % of surgical patients 1 17% of cardiac arrest patients survive to discharge 2 For a hospital with 20,000 medical and surgical admissions we expect ~100 cardiac arrests that would end with 83 in-patient mortalities annually. 1 - Needleman J et al. NEJM 2002:346:1715-1722 2 - Peberdy MA et al. Resuscitation 2003;58:297-308 4
The Joint Commissions National Patient Safety Goals 2009 Recognition and response to changes in patient s condition Accuracy of patient identification Communication among caregivers Health care associated pressure ulcers SAFETY Safety of using medications Patients active involvement Health careassociated infections Patient harm resulting from falls Reconcile medications 5
Timing is Everything The timing of several acute care interventions has a substantial impact on mortality. Thrombolytic, PPI, aspirin, beta-blockers in myocardial infarction. Emergency resuscitation. Angiography, thrombolytic Tx in acute stroke. Early goal-directed therapy for sepsis. 6
Can we predict clinical deterioration? 60% of in-hospital cardiac arrests, deaths and emergency unplanned ICU admissions are preceded by hemodynamic antecedents - The ACADEMIA study. Kause J et al. Resuscitation 2004 84% of patients who developed cardiac arrest had instability within the 8h window preceding the event - Schein RM et al. Chest 1990 Of 150 cardiac arrest cases, 99 (66%) had documented instability in the 6h window preceding the event Franklin C et al. Crit Care Med 1994 7
The Solution: Rapid Response Teams 8
Rapid Response Systems: Why aren t we getting the results we expect? Outcome studies looking into the benefits of RRS have generally reported mixed results. MERIT trial 1 (randomized controlled trial of 23 hospitals in Australia) No substantial affect on incidence of cardiac arrest, unplanned ICU admissions, or unexpected death. Implementation of a RRS at Brigham and Women s Hospital A recent Meta-analysis 3 concluded that the data supporting RRSs are not "robust." 1 Hillman K et al. Lancet 2005: 365. 2 Rothschild JM et al. Jt Comm J Qual Patient Saf. 2008: 34. 3 Chan PS et al. Arch Intern Med. 2010;170. 9
Rapid Response Systems: Why aren t we getting the results we expect? The patient Efferent limb: Planned response Quality analysis Afferent limb: Event detection and response triggering Staff s skills Adequate therapy at hand? RRT Sufficient staff? Staff s skills Sufficient monitoring? 10
Reasons for Failure to Rescue Addressed by RRS Not addressed by RRS Jones DA et al. N Engl J Med 2011;365:139-146. 11
Opportunities for Intervention Taenzer AH et al. Anesthesiology 2011:115 12
So what should we look for? SpO2 oxygen saturation PaCO2 arterial carbon dioxide RR respiration rate Ve minute ventilation Lynn and Curry Patient Safety in Surgery 2011, 5:3 13
Monitoring of Low-Risk Hospitalized Patients Has to be: Continuous Easy to use for the staff Places as little limitations on the patients Integrated monitoring system (IMS) Cost-effective 14
The EarlySense Monitoring system 15
16
Validation Preliminary validation in sleep lab patients. Follow up validation in an ICU setting. Performance testing: Patient s weight Bed position Patient position / activities Feasibility studies Sheba Medical Center, Israel Meir Medical Center, Israel MetroWest Medical Center, Framingham, MA Ben-Ari J, Zimlichman E et al. J Med Eng Technol. 2010 Oct-Nov;35(7-8):393-8. 17
First clinical implementation The Sheba/Sapir Study Non-interventional monitoring of patients in a medical ward environment. Two acute care medical centers in Israel Sheba Medical Center and Sapir Medical Center. Overall, 113 patients with an acute respiratory disease participated in the study. Patients were followed-up for major clinical events (death, ICU transfer or on the floor intubation). 18
The Sheba-Sapir study: Results Vital signs alerts Out of 113 patients, we identified 10 major clinical events. Through retrospective monitoring data analysis we have calculated the accuracy of the EarlySense monitor in identifying these events. Sensitivity Specificity Positive predictive value Negative predictive value 40 HR 115 82% 67% 21% 97% 8 RR 40 64% 81% 26% 95% HR and RR 55% 94% 50% 95% Zimlichman E et al. J Hosp Med. Aug 2012. 19
The Sheba-Sapir study: Results Trend Alerts For trend analysis we grouped together HR and RR readings for running 6 hour periods throughout the day and compared the median of the readings for each period with the corresponding period on the previous day. Sensitivity Specificity Positive predictive value Negative predictive value HR 20 78% 89% 41% 97% RR 5 100% 69% 25% 100% HR and RR 78% 93% 54% 98% Zimlichman E et al. J Hosp Med. Aug 2012. 20
The California Hospital Medical Center study Full implementation of the system in a 33 bed Medical- Surgical unit, including a wireless connected central nurse station and pagers for nurses. 21
Study Objectives To evaluate the ability of continuous monitoring on a medical-surgical unit to detect adverse events and help reduce risk. Study outcomes: Transfer to intensive care units Length of stay (total and ICU) Code Blue events Return on investment ($) Design: 9-month intervention, comparing pre-post and to a sister control unit. 22
Study Outcomes Control Unit Intervention Unit CU-IU post All Units LOS in med/surg unit (days) ICU transfers Baseline (pre) Control (post) p value Baseline (pre) Intervention (post) p value p value p value 3.80 3.61 0.26 4.00 3.63 0.03 0.91 0.10 Transfers / 1000pt 18.89 19.06 1.00 26.52 25.93 0.92 0.12 0.21 ICU Days / 1000pt 32.69 85.36 ICU LOS, mean (median) 120.11 63.44 0.01 1.73 (1.32) 4.48 (2.12) 4.53 (2.33) 2.45 (1.85) 0.05 0.02 0.014 APACHE II score 13.08 14.06 0.59 15.19 13.38 0.25 0.61 0.65 Code blue events n (/1000pt) 6 (3.9) 5 (2.1) 0.36 9 (6.3) 2 (0.9) <0.01 0.45 0.02 Zimlichman E et al. American Thoracic Society Annual Meeting, San Francisco 2012. 23
Results Number of Alerts No. of patients 73 No. of Heart Rate (HR) Alerts average per week No. of Respiration Rate (RR) Alerts average per week No. of Turn Patient Alerts average per week No. of Bed Exit Alerts average per week Total no. of alerts per week (168 hours) Average number of alerts / hour Estimated True Alerts Estimated False Alerts 10 60 60 37 167 1.0 alert / hour 117 50 Average number of alerts per 12 hours shifts (for all nurses) Average number of alerts per 12 hours shift per nurse (assuming 6 nurses on shift) Estimated false alerts per nurse per shift 12 2 0.60 24
Alert Burden on Staff Study Current study In-patient setting Medical-surgical units Type of alerts Alerts per 100 recording hours Heart rate (18%) and respiratory rate (82%) Chambrin et al., 1999 27 ICU Ventilators (38%), cardiovascular monitors (37%), pulse oximeters (15%) and capnography (14%) Lawless et al., 1994 28 Pediatric ICU Pulse oximeter (44%), ventilators (31%), cardiovascular monitors (24%), capnography (1%) Görges et al., 2009 29 ICU Ventilators (40%), cardiovascular monitors (21%), pulse oximeters (15%), infusion pumps (12%) Siebig et al., 2010 30 ICU Cardiovascular monitors (66%), pulse oximeters (26%), respiration Wiklund et al., 1994 31 Malviya et al., 2000 32 Postanesthesia care unit Pediatric Postanesthesia Care Unit 25 2.2 161 230 636 604 rate (3%) Pulse oximeters 730 Pulse oximeters 167 Masimo Signal Extraction Technology pulse oximeters 200
CASE STUDIES 26
Case #1: Gastrointestinal Bleeding A 60 y/o male was admitted to the unit with respiratory failure, tachycardia, fever, diabetes and alcoholism. The Event: The system detected consistent and gradual HR increases from 110/min to 155/min and generated five high HR alerts. The Outcome: The patient was diagnosed with new GI bleeding and was transferred to the ICU. After two days the patient s condition improved and he was discharged from the ICU. 1:00 am to 6:00 am - HR increasing up to 155 BPM. High HR alerts 27
Case #2: Sepsis and Metabolic Acidosis A 76 y/o female was admitted with acute pancreatitis. Her chief complaint was severe, non-radiating abdominal pain. The Event: The system alerted for high RR. The Outcome: The patient was transferred to telemetry unit and later to the ICU. She was intubated and diagnosed with necrotizing pancreatitis and sepsis. The high RR was secondary to metabolic acidosis. Her condition gradually improved. High RR alerts 28
THANK YOU QUESTIONS? Eyal Zimlichman MD, MSc Eyal.zimlichman@sheba.health.gov.il 29