Assuring High Quality Patient Outcomes through Clinical Systems Integration and Analytics:

Transcription

1 Assuring High Quality Patient Outcomes through Clinical Systems Integration and Analytics: Organization wide Integration and Networking for Population Management The Seventh Annual Predictive Modeling Summit December 5, 2013 Margaret J. Holm R.N., Ph.D., F.A.C.H.E. Charles G. Macias M.D., M.P.H.

2 Conflict of Interest Disclosure Margaret J. Holm R.N., Ph.D., F.A.C.H.E. Director Quality and Clinical Systems Integration Texas Children s Hospital Charles G. Macias M.D., M.P.H. Medical Director, Center for Clinical Effectiveness and Evidence Based Outcomes Center, Texas Children's Hospital/Baylor College of Medicine Has no real or apparent conflicts of interest to report.

3 Learning Objectives Discuss the current and future utility of predictive modeling and its clinical and financial implications through Clinical Systems Integration. Outline an organization wide population health strategy that steadily builds permanent, integrated teams inspiring the adoption of the science of improvement, evidence based medicine and the utilization of clinically relevant data for prediction and action. Demonstrate an iterative approach used to create meaningful data for the clinicians and eliminate unproductive time for analytics personnel. Describe how expanding analytical capabilities into the organization creates opportunities to further identify clinical and operational variation and view clinically relevant financial data.

4 Predictive Modeling: Challenges in Healthcare Burgeoning information age Clinical challenges Wide variations in clinical practice Wide variations in beliefs, interpretation of evidence Lack of agreement of relevant metrics Reliability and validity Slow uptake of research in health care delivery systems Inadequate training to develop robust and clinically relevant models Lack of integration between technological, scientific, and data management infrastructures Adaptability of models to varied and changing populations

5 Current Data Systems Traditional database systems Data storage Data retrieval Traditional and early attempts at applying predictive modeling Data analysis and reporting Decision support Investments more heavily committed to managing costs e.g. health plans identification of high consumers Query: structure for systematic translation of predictive models to improve outcomes

6 About Texas Children s Hospital Statistics (Annual) Number of Beds 658 Annual Inpatient Admissions Patient Encounters Emergency Room Visits 31, million 116,000 Surgeries 26,802 Births 5,074 Employees 9,199

7 Population at Texas Children s Hospital: Children and Women Health Plan Pediatric Hospital Clinics Women s Pavilion

8 Board Quality Strategy and Governance: Clinical Systems Integration Implementation integrated and evidence based quality and safety program measurable improvements Relevance meaningful data information about clinical outcomes and operational processes CLINICAL SYSTEMS INTEGRATION EXECUTIVE COMMITTEE CLINICAL PROGRAM COUNCIL Structure enterprise wide data management infrastructure leverage the clinical systems easy to access assist in accelerating improvements in clinical and operational processes. EDW TEAM CARE PROCESS TEAMS WORK GROUPS

9 The clinical challenge: Jenny Jones and a Fragmented System Within six months, Jenny had visited: One PCP Two Hospitals Three ERs Leading to: Six different Asthma Action Plans with conflicting discharge instructions

10 Quality Defined

11 Variation

12

13 Building Capacity for Population Health Content System Evidence Based Care Evidence Integrated practice via guidelines, order sets and measures Evidence based Care is the standard of care in the organization Improved Population Health Governance Structure for the Organization Deployment strategy Care Process Teams Deployment System Measurement System Care Process Teams Define Measures

14 Clinical Systems Integration and Predictive Modeling Trending system level population data to predict process and outcomes metrics at current state then strategize and implement to improve outcomes.

15 Prioritization of Populations: Pareto Principle

16 Care Process Analysis: Data from EDW Amount of Variation Asthma Diabetes Sickle Cell Appendectomy Prematurity Resources Consumed Copyright HealthCatalyst 2013 Improvement Opportunity: Large processes with significant variation Bubble Size = Case Count

17 Targeting Opportunities for Improvement Mean 1.96 std # of Cases 1 box = 100 cases in a year # of Cases Excellent Outcomes Poor Outcomes Excellent Outcomes Poor Outcomes Option 1: Focus on Outliers the prescriptive approach Strategy. Identify extreme cases with the potential for high costs from bad outcomes and eliminate the unfavorable tail of the curve ( executive dashboard approach) Result. If the outlier trim point is set at 1.96 standard deviations, only 2.5% of cases fall under the adverse outcome tail, so the impact is minimal Copyright HealthCatalyst

18 Targeting Opportunities for Improvement Mean # of Cases 1 box = 100 cases in a year # of Cases Excellent Outcomes Poor Outcomes Excellent Outcomes Poor Outcomes Option 2: Focus On Inliers improving quality outcomes across the majority Strategy. Identify best practices through research and analytics and develop guidelines and protocols to reduce inlier variation Result. Shifting the cases which lie above the mean toward the excellent end of the spectrum produces a much more significant impact Copyright HealthCatalyst

19 Iterative Process Exercise

20 Success Factors for an Iterative Approach Integrated team: Frequent interaction between clinical and technical experts. Flexible : Adjustments to the analytics system are easy and almost immediate. Iterate: Check and act. Check and act. Repeated adjustments make the data better and more meaningful.

21 TCH s EDW Architecture Copyright HealthCatalyst 2013 Metadata: EDW Atlas Security and Auditing FINANCIAL SOURCES (e.g. EPSi,) Common, Linkable Vocabulary DEPARTMENTAL SOURCES (e.g. Sunquest Labs) Financial Source Marts Departmental Source Marts ADMINISTRATIVE SOURCES (e.g. API Time Tracking) Administrative Source Marts Clinical Asthma Appendectomy Deliveries Pneumonia Diabetes Surgery + others Operations Labor productivity Radiology Practice Mgmt Financials Patient Satisfaction + others Patient Source Marts PATIENT SATISFACTION SOURCES (e.g. NRC Picker, EMR Source Marts HR Source Mart EMR SOURCE (e.g. Epic) Human Resources (e.g. PeopleSoft) More Transformation Less Transformation

22 Information Management Skills Knowledge Managers (Workflow) DATA CAPTURE Acquire key data elements Assure data quality Integrate data capture into operational workflow Application Administrators (e.g., EMR Administrators, Financial System Administrators) = Subject Matter Expert DATA ANALYSIS Interpret data Discover new information in the data (data mining) Evaluate data quality DATA PROVISIONING Move data from transactional systems into the Enterprise Data Warehouse Build visualization for use by clinicians Knowledge Managers (Evidence) Data Architects (Analysis) Data Architects (infrastructure) Data Architects (Visualization) Copyright HealthCatalyst 2013

23 Clinical Programs: Quality & Clinical Evidence based Team Clinical Program MD Lead RN Lead MD Lead RN Lead MD Lead RN Lead MD Lead RN Lead MD Lead RN Lead Operations Director Clinical Director MD Lead #5 Care Process #4 Care Process #3 Care Process #2 Care Process #1 Care Process Knowledge Manager Data Architect (Analysis) Data Architect (Visualization and Infrastructure) Application Service Owner Copyright HealthCatalyst 2013 = Subject Matter Expert = Data Capture = Data Provisioning = Data Analysis

24 Texas Children s Clinical Programs Surgery Clinical Program Medicine Clinical Program Women s Clinical Program Copyright HealthCatalyst 2013

25 Asthma: Team Specific Cohort Definition Overlap (19,776) Potential Rules (101,389) Medications (72,581) Supplemental ICD9 (38,250) Copyright HealthCatalyst 2013

26 Metric Development Process 1. Care Process Defined 2. Current Literature Research 3. Individual Ratings 5. Group Creates Final Scorecard 4. Aggregate Ratings

27 Chronic Asthma Metrics: A Balanced Report Card Measure Site of Care Quality Domain Asthma severity and control is assessed by health professional during a patient visit Documentation of asthma control medications prescribed (e.g., corticosteroids) Percent of population that has had a health professional show them how to use an inhaler and spacer Planned care visits for asthma are completed at least every 6 months, or more frequently for more severely ill patients or those with comorbidities Patients hospitalized for asthma or asthma-related issue Asthma patients who have had a visit to the ED/Urgent Care site for asthma in the past "x" months H, ED, IP Effective Efficient Safe H, ED, IP Efficient Safe Timely H, ED, IP, PCP Effective Efficient Patient centered Safe ED Effective Efficient Patient centered Safe H IDS Effective Patientcentered Safe (Care coordination or lack thereof) Effective Patientcentered Safe KEY IP = Inpatient, PCP = Primary Care Provider, ED = Emergency Department OR = Operating Room

28

29

30 Financial Modeling Efficiency shapes quality definitions IOM: avoidance of waste Cost modeling can be linked to clinical care, expectations and outcomes

31

32 Providing financial data to link clinical processes

33 Proactively Positioning for New Requirements Beyond operations, deep into clinical care delivery Future Care Delivery Actions to Position TCH Data driven Specific aim to improve care and reduce costs Targeting specific population Clinical, data, operational experts in a team Managing populations and care across the continuum Multi professional Cross departmental Multi dimensional improvements Linking care across the continuum

34 Clinical Systems Integration: Changing Care Delivery for Jenny Jones Moving theoretical to practical interventions Characterizing population trends Outcomes and financial burden Modeling to understand impact of change with real populations using near real time data Infusing all levels of evidence into models EDW: accelerating data retrieval and transformation Empowering patients across the system

35 Key Points Predictive modeling in health care has limited real time applicability at the patient level without a dedicated strategy Structures to effectively integrate data management/transformation, best science/evidence, and operational improvement can link models to outcomes Adaptability and iterative improvement is necessary to achieve targeted outcomes Assure organization and leadership readiness to assure effective spread Assure IS resources

36 Thank You QUESTIONS & ANSWERS Margaret J. Holm RN, PhD, FACHE Charles G. Macias MD, MPH