1 Joint Session: Sepsis Expanding your Toolbox to Help Sepsis OutcomesHawai‘i Pacific Health Melinda Ashton, M.D. Arys Nogueron

2 Serving our communities for4 hospitals (566 beds) >34,000 admissions 44+ clinic sites >1600 physicians >1700 nurses Serving our communities for more than a century Wilcox Health Kapi‘olani Medical Center for Women & Children Pali Momi Medical Center To begin, I’d like to share a little about Hawai‘i Pacific Health. We are four flagship hospitals and 44 outpatient clinics. Our health care system is unique in Hawaii because of our breadth, depth and level of expertise. Being part of a health care system means when you visit one of our 1,300 physicians, you’re not just seeing one person, you’re gaining access to best practices, extensive knowledge, up-to-date technology, and resources of our entire network. Straub Clinic & Hospital 2 FY09 data. 2

3 Why Sepsis? Motivation to reduce length of stayMotivation to improve outcomes

4 Average Length of Stay for All Sepsis Patients13.81 days 8.55 days

5 Average Length of Stay Severe Sepsis/Septic Shock Patients18.76 days 11.32 days

6 Raw Mortality Rate for All Sepsis Patients19% 6.8%

7 Raw Mortality Rate Severe Sepsis/Shock Patients42.1% 17.6%

8 The beginning High Value Healthcare Collaborative provided 3 hour and 6 hour bundles of surviving sepsis campaign strategies We decided to build tools to drive bundle adoption and better care for all sepsis patients

9 Quality Improvement Hawai‘i Pacific Health system initiativeThen each hospital created their own sepsis team

10 Build will…….

11 ICU admission: 48% Mortality: 49% ICU admission: 88% Mortality: 26%94 88 85 81 73 73 75 76 76 69 63 58 53 42 31 ICU admission: 48% Mortality: 49% ICU admission: 88% Mortality: 26% ICU admission: 88% Mortality: 46% 2013 Chart review of 59 patients who were septic on arrival and did not transfer

12 Build Tools: Ordersets

13 Ordersets Sepsis admissionAdded sepsis bundle orders to other admission orders Other issues Blood culture trigger for BPA Nursing initiated orders

14 Build Tools: Best Practice Advisories

15 ED Triage BPA Activation: Some common early sepsis criteria Required:yes to ‘systemic infection’ nursing question OR blood culture ordered OR Yes to ‘fever within 24 hours’ Three of the following with the last set of vitals: Temp less than 36 C or above 38 C HR >105 RR >22 SBP < 90 Altered mental status ED Wards only

16 Sepsis BPA Activation: Some common early sepsis criteriaRequired: yes to ‘systemic infection’ nursing question OR blood culture ordered OR yes to ‘Fever within 24 question’ Two of the following with the last set of vitals: Temp less than 36 C or above 38 C HR >105 RR >22 SBP < 90 Altered mental status One of the following in the last 24 hours WBC above 12 K or below 4 K Lactate above 2.2 (CLH abnormal range) WBC differential with bands >10% ED Physician, Inpatient Nursing and Inpatient Physician BPAs Use this criteria. ED Physician is specific to the ED Wards Inpatient BPAs are at most sites Exceptions: Family Birth Center, OR, Cath Labs

17 Behind the Scenes Monitored and gathered data on BPAs with active patients no alerts to front line clinicians Corrected any abnormalities before go-live

18 Severe Sepsis AccuracyBPA Methodology applied to All Pts Sepsis BPA 6 months of data Sepsis Accuracy Severe Sepsis Accuracy Misfires PMMC 63.3% (34 per month) 66.7% (7 per month) 3.6% (147 per month) SCH 29.4% (65 per month) 48.3% (16 per month) 1.9% (44 per month) WMH 50.4% (6 per month) 65.4% (2 per month) 2.5% (42 per month) SCH does not have a ‘suspected infection’ question gathered in the ED. This makes a significant difference

19 BPA Data Post Go-Live

20 BPA accuracy: Sepsis Early results Recent results n = 64,298Sensitivity: BPA fired when patient had sepsis 69.3% 66.9% Specificity: BPA didn’t fire when patient didn’t have sepsis 98.9% Positive predictive value: when BPA fired patient had sepsis diagnosed 50.3% 52.2% Negative predictive value: when BPA didn’t fire sepsis was not diagnosed 99.8% 99.3%

21 BPA accuracy: Severe SepsisEarly results n = 64,298 Recent results n = 170,293 Sensitivity: BPA fired when patient had sepsis 79.6% 95.9% Specificity: BPA didn’t fire when patient didn’t have sepsis 98.2%

22 Measure Current Practice

23 Epic Clinical Data Reporting (ECDR) Team

24 ECDR Multidisciplinary team tasked to extract data, and develop datamarts (with data dictionaries) to allow regular and routine reporting. Recommend needed changes to EHR to capture desired data Meets twice a month

25 Steps of Process Clinical Data Reporting Team Members EHRBusiness Analyst Reporting Team Pt Safety/Quality Team HIM Clinical Experts and Frontline Users 1 Outline Fields of Interest x 2 Work as Team to Identify Fields in EHR 3 Discuss and Agree on Data Definitions 4 Design Datamart and Data Dictionary 5 Validate Datamart 6 Implement into Production Steps of Process Combining the various teams was a true collaborative effort. In addition to the teams that we just showed you, we added the Inpatient Business Analysts, and Clinical Experts who are Frontline Users to form the ECDR Team. Each member or group under the ECDR had its own roles. First we had to outline the fields of interests, primarily this came from the Patient Safety and Quality Team, the Frontline users and also we at the Reporting Team are aware of the specifications via the requests. We work together as a team in a Computer Lab to id the fields in EMR. Sometimes we do this on our own between meetings, and other times we do collaborate on the data definitions. Next we discuss the fields, review the descriptions in the specifications to determine the appropriate clinical definition and data definition. In the event of the Birth Weight, we decided to capture both the Birth Weight in the Delivery Summary, and the Birth Weight found in the Electronic Birth Certificate to provide flexibility for reporting. This way the user who needs the report can have the flexibility of using one or the other field depending on their reporting needs. The Reporting team then takes the specifications and pulls it all together in a datamart. The Data Dictionary is set up after the Datamart has been designed and Validated. Validation is a team effort – we split up data results and have all the teams validate. We tweak the logic for the datamart as needed, go through validation again for those fields, until we get it right, then we put it into production.

26 Field: Fluid Delivery HIM MD RN IT RN Rev cycle QualitySome orders get changed or canceled Measure what is ordered I can’t link MAR charting with flowsheet charting ED RNs don’t always chart volume HIM MD RN IT Field: Fluid Delivery RN Rev cycle Quality How can we fix the process to better serve the patient? They chart ‘start’ and ‘stop’ time for billing

27 Fluids!! Our biggest AH HA!Our clinicians in ED had learned to provide care without accurate fluid data. We had to approach this in multiple ways: Audits CNE “HPH fluid week” Documentation approach

28

29

30 Continued Sharing of Results

31 3 Hour Bundle Compliance

32 3 Hour Bundle Compliance

33 3 Hour Bundle Compliance

34 Current work Dissemination of resultsModified reporting to mimic CMS core measure requirements for 3 hour bundle

35 arys. nogueron@hawaiipacifichealth. org melinda