Improving Outcomes and Saving Lives in Real Time: How Hospitals Can Use Predictive Analytics Across the Care Continuum Essential Hospitals Engagement

Transcription

1 Improving Outcomes and Saving Lives in Real Time: How Hospitals Can Use Predictive Analytics Across the Care Continuum Essential Hospitals Engagement Network February 18, 2015

2 CHAT FEATURE The chat tool is available to ask questions or comments at anytime during this event. 2

3 RAISE YOUR HAND If you wish to speak telephonically, please raise your hand. We will call your name, when your phone line is unmuted 3

4 AGENDA Overview» David Engler, PhD America s Essential Hospitals Improving Outcomes and Saving Lives in Real Time» Ruben Amarasingham, MD Parkland Center for Clinical Innovation Q & A Wrap-up 4

5 SPEAKER INFORMATION Ruben Amarasingham, MD, MBA President and CEO Parkland Center for Clinical Innovation 5

6 How hospitals can use real time predictive analytics across the care continuum: A case study at a Texas essential hospital Ruben Amarasingham, MD, MBA

7 Agenda Highlights Using Real-Time Prediction to Improve Sepsis Care and Outcomes Tackling Readmissions with Predictive Analytics Addressing the Social Determinants of Health Challenges in Predictive Modeling Applications Q&A 2/18/2015 7

8 Case Study Highlights Sepsis ADVERSE EVENT RESULTS Parkland Hospital Relative improvement in sepsis bundle compliance: 100% Relative reduction in mortality: 17.4% Estimated yearly savings: $1.15M - $5.25M Readmission Reduction Parkland Hospital 1115 Waiver revenue recovery (FY13-FY14): $8M On pace for 1115 revenue recovery (FY15-FY16): $30M Penalty avoidance to date: $5.7M 2013 Lowest CHF readmission rate (CMS peer group) Texas Health Resource (HEB Hospital) Chronic Kidney Disease outpatient management Parkland Hospital Relative reduction in HF readmissions: ~40% Savings for every $1.00 spent: $4.59 Increase in compliance achieving goal SBP, DBP, medication best practices: 82% 2/18/2015 8

9 What Clinicians Do in Medicine: Prediction 1. What does this patient have? 2. What will this patient develop? 3. What will be the effect of a given therapy? 2/18/2015 9

10 Prediction in the Context of Modern Medicine : 150 years Doubling Time of Medical Knowledge Year 2/18/

11 Prediction in the Context of Modern Medicine Doubling Time of Medical Knowledge We are here: 1 year Year 2/18/

12 Prediction in the Context of Modern Medicine Doubling Time of Medical Knowledge Staggering increase in medical information Increasing volume of decisions at multiple levels High fragmentation of care Increasing capacity for error We are here: 1 year Year 2/18/

13 What is Electronic Clinical Predictive Modeling and What is its Purpose? Using electronic data to predict future clinical events so that one can: 1. Discriminate between high and low risk patients 2. Prevent adverse events 3. Allocate scarce clinical resources under real-time demands 4. Suggest actions 2/18/

14 PCCI Organizational Background A 501c(3) non-profit research and development corporation specializing in the development of clinical prediction and surveillance software to help prevent adverse clinical events. 2/18/

15 PCCI Relationship with Parkland Revenue Share, Software Products & Services Funding, Infrastructure & Data PCCI 501(c)(3) Non- Profit 2/18/

16 PCCI History, Funding, and Research PCCI has obtained >$30M in scientific funding for predictive analytics. 2/18/

17 Every Adverse Event has a Timeline Hours 30 Days 90 Days Years Patient Quality and Safety Events Cardio-Pulmonary Arrest Sepsis Surgical Complication Hypoglycemia Readmissions All-Cause CHF AKI PNA DM Failure Mode and Effects Analysis Population Health Preventable Admissions Short-Term Diabetic Complications Asthma Complications End of Life Care Population Health Preventable Admissions Triad: CKD, Diabetes, Hypertension (NIH) 2/18/

18 Every Adverse Event has a Timeline Admission Discharge 30 Days 90 Days 24 hours 2/18/

19 Every Adverse Event has a Timeline Admission Discharge 30 Days 90 Days 24 hours 7 days EMR Pieces 2/18/

20 Sepsis is Common, Deadly, Expensive and On the Rise Major cause of morbidity and mortality 750,000 cases of severe sepsis per year in the United States with a nearly 40 percent mortality rate 1,2 Hospitalization rates have increased steadily 3 In-hospital mortality for septic patients has been estimated at 17% by NCHS 4 Approximately $15.4 billion was spent in 2009 for septicemia hospitalizations alone Sepsis diagnoses are projected to outpace population growth 1 1. Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Critical care medicine. Jul 2001;29(7): Infection and Sepsis-Related Mortality Hotspots Identified across the U.S. 2013; Accessed 5/30/ National Hospital Care Survey. Data Uses. National Hospital Care Survey 2012; Hall MJ, Williams SN, DeFrances CJ, Golosinskiy A. Inpatient care for septicemia or sepsis: a challenge for patients and hospitals. NCHS data brief. Jun 2011(62):1-8 2/18/

21 Sepsis Treatment Timeline: Overview 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement 2/18/

22 Sepsis Treatment Timeline: Overview 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement 6 EMR Pieces 2/18/

23 Sepsis Treatment Timeline: Overview 0 Hours Critical time for identification 3 Hours 6-24 Hours Discharge and early therapy Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement EMR Pieces 2/18/

24 Sepsis Treatment Timeline: Arrival through Triage 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders 1 2 Inpatient Intervention Evaluation & Improvement Patient Arrival PCCI Sepsis Risk Evaluation EMR Pieces 2/18/

25 Developing the Sepsis Model: Variable Splines Sepsis Risk WBC Sepsis Risk Respiratory Rate 2/18/2015 Lucena et al, Manuscript in preparation 25

26 Sepsis Treatment Timeline: Sepsis Alert through Admission 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement Alert and Provider Evaluation EMR Pieces 2/18/

27 Sepsis Treatment Timeline: Sepsis Alert through Admission 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement Activation of Care Pathway EMR Pieces 2/18/

28 Pieces Continues to Monitor the Patient after Admission 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement Ongoing monitoring EMR Pieces 2/18/

29 Reporting for Performance Improvement 0 Hours 3 Hours 6-24 Hours Discharge Arrival Admit Transfer ID Risk Alert Orders Inpatient Intervention Evaluation & Improvement Evaluation and Improvement 3 hour sepsis bundle 6 hour sepsis bundle Length of Stay Mortality EMR Pieces 2/18/

30 Improved Compliance with Bundle Metrics Percent Compliance 40% +77.8% % % Hr Metrics Hr Metrics Cumulative Metrics 2014 Post-Implementation /18/

31 Substantial Early Impact on Mortality Sepsis POA Mortality Rate 8% -17.4% Potentially ~21 lives saved at Parkland Post-Implementation 2/18/

32 Cost Savings Total Allocated Cost $30,000 $27,124-12% -5% $23,799 20,000 10,000 Average Cost per Sepsis Patient $14,219 Median Cost per Sepsis Patient Average Cost per Sepsis Patient $13,490 Median Cost per Sepsis Patient Post-Implementation 2/18/

33 Readmission Model Derivation Samples Validation Samples * Day Readmission (%) Very Low Low Intermediate High Very High Predicted Readmission Risk Category Amarasingham et al, Medical Care, /18/

34 Natural Language Processing for Readmission Models 68 yo WF presents with acute on chronic non ischemic systolic and diastolic chf, severely depressed ef and grade ii diastolic dysfunction. Disease/ Symptom Time Attribute Acute Heart Failure Chronic Heart Failure current and primary historic Systolic, significant depression in ejection fraction; Diastolic dysfunction, grade 2 Non-ischemic 2/18/

35 Success Highlights: All-Cause Readmissions at PHHS Readmission Rate 10.5% % 9.9% FY13 Benchmark % FY15 Goal % Hospital Average Monthly Actual 8.0 Oct-12 Feb-13 Jun-13 Oct-13 Feb-14 Jun-14 Oct-14 2/18/

36 Community Hospital Results: HF Readmission Readmission Rate 40% 30 23% National Average ' % 18% 13% Goal Monthly Actual Hospital Average 0 Oct-11 Feb-12 Jun-12 Oct-12 Feb-13 Jun-13 Oct-13 Feb-14 Jun-14 Oct-14 * This readmission rate does not account for: terminal illness, hospice, elective readmissions which would lower the CMS readmission rate. Full review of pre- and post-intervention states recommended. Source: THR Clinical Informatics 2/18/

37 Connecting the DFW Community We thank the W.W. Caruth, Jr. Foundation at Communities Foundation of Texas for the generous grant of up to $12 million grant to build, operationalize, and launch the Dallas IEP 2/18/

38 Partnerships 2/18/

39 Connecting the Community Low-cost and simple electronic case management and client tracking solution for connecting organizations working on social determinants of health 2/18/

40 Pieces Analytics for the Community Leverages predictive and prescriptive analytics on medical and social data to identify at risk individuals 2/18/

41 Complexities of Predictive Modeling in Healthcare 2/18/

42 The Complexities of Predictive Modeling 1. Interventions for highest risk patients * 2. Considering clinical vs. social risk 3. Explanation vs. Prediction 4. Non-health care data sources * 5. Changing EMR data models 6. Changing clinical interventions 7. Changing populations Amarasingham et al, Health Affairs, /18/

43 Thank You! 2/18/

44 Questions Contact Information: Ruben Amarasingham Spencer Ballard 2/18/

45 Sample Sepsis Treatment Timeline: Actual Events ID Risk Alert Orders 8:12 PM patient arrives in the Parkland ED 1 8:28 PM labs ordered 2 8:33 PM labs drawn 3 4 9:14 PM subset of labs result EMR Admission 5 9:25 PM Pieces Sepsis alert model fires 9:27 PM page sent to nurse, provider receives Best Practice Alert (BPA) 9:37 PM BPA accepted, order set submitted Pieces 9:38 PM Sepsis activation page to ED Nurse, ED phlebotomy, ED Pharmacy Inpatient Intervention 6 10:08 PM patient receives first dose of broad spectrum antibiotics Evaluation & Improvement 7 11:27 PM Order placed for hospital admission 1:28 AM Patient transferred to floor 2/18/

46 PHHS Sepsis POA Volumes FY 2013 No Intervention FY 2014 Post-Implementation October 1, 2012 September 30, 2013 June 2,,2014 September 30, , 445 patients 579 patients 2/18/

47 Six Key Components Every System Needs to Achieve Results Detect PHHS Sepsis Solution Components Accurately ID confirmed sepsis patients Predict Warn Act Monitor Learn Real-time prediction of likely sepsis patients Notification to the appropriate care team at the right time and place Rapid coordination of defined sepsis interventions Continuous monitoring: real-time feedback and tracking of intervention and workflow compliance Performance improvement tracking, model refinement, FMEA analysis, end user feedback 2/18/

48 QUESTIONS 48

49 THANK YOU FOR ATTENDING Upcoming Webinars and Events: The Ebola Outbreak: Essential Hospitals on the Front Line February 25 Webinar Register Here Policy Assembly March Washington, DC Register today at PolicyAssembly.essentialhospitals.org Vital2015 June San Diego, CA Register today at vital.essentialhospitals.org Evaluation: When you close out of WebEx following the webinar, an evaluation will open in your browser. Please take a moment to complete. We greatly appreciate your feedback! 49