Next Generation Data Analytics Platform

Transcription

1 Next Generation Data Analytics Platform NJ/DVHIMSS Fall 2018 Jim Beinlich Associate Vice-President Corporate Information Services

2 What is our GOAL? Create an integrated Ecosystem of clinical, financial, research, IoT and other data that allows users to work in a common environment to speed discovery through data 2

3 What s driving change? Complexity, Velocity, and Demand for Self-Service 3

4 4

5 What s driving change? External demand drivers Penn s Response What we need Mounting pressure on rates Increased focus on quality and value metrics (and cost thereof) Investment in Epic everywhere at Penn Demand for service line integration across continuum of care Shift to at-risk contracting, population management, and value-based care Increased consolidation and competition DAI: Improve efficiency & effectiveness of existing resources Groups adding FTEs to support needs Ad hoc analytics (approach, tools, data, systems) Frustration Step change in capability to meet existing and known nearterm demands Clinical Business Research All with an accelerating pace of change 5

6 Data Access Capabilities Chronology of Data Access Capabilities Consolidate and Expand Data Access Resources Consolidate EMR Establish Research Data Warehouse Institute Executive Analytics Governance Enable Search and Discovery of Clinical Notes Establish Centralized Data Access Team Create Specialized Datamarts and Dashboards Expand Population Health & Value Based Care Implement Analytics Self Service Storefront Implement Integrated EMR DW - Epic Caboodle Epic Cosmos Research Network Integrate Princeton Medical Center Implement Pathways of Care Institute Executive Analytics Prioritization Broaden Use of Organization Wide Internal Registries and Alerts Fully Integrated EMR Establish Common Dashboards for Ambulatory KPI s Consolidate Data Access Requests Across Organization Integrate External Claims Data Integrate Chester County Hospital and Practices Integrate Phenotype and Genotype Data Implement Organization Data Lake Strategy Collaborate in Research Network Facilitating Clinical Trial Recruitment Financial DW Aggregate Patient Data into Central Data Warehouse Implemented Common Ambulatory Applications

7 Describing the architecture 7

8 How we re looking at the solution Interchangeable modules that provide specific features yet work together so that the whole appears to be a single solution to an end user Requires a platform to run on. Like a smartphone is a platform upon which many different apps operate a unified analytics platform provides the foundation that ties the individual pieces together Custom building an in-house solution would be analogous to building our own smartphone introduces cost and delays that don t make sense 8

9 So what s different? Old Paradigm Warehouse Fixed Data Model Ingest Scrub and Transform Retrospective Report New Paradigm Lake Flexible Data Model Connect Late-Binding Real Time Discover 9

10 Flexible and Secure Data Model Non-proprietary, interoperable, easily adaptable Proposed Ecosystem for Analytics Significantly improving access to data Libraries, Widgets, Visualization Tools Provides most common tools, allows for bring your own widget plug and play variety of visualization tools stored libraries available for re-use Tools can access any data in the lake Penn Data Lake Curated Data Clinicians, students, Basic research needs Curated Data Sets Curated Data Sets Curated Data Sets RAW Data Informaticists, Advanced researchers Genomics Somatic Bio-bank CPD Germline SNP Structured EMR Labs Encounters ICD & CPT Meds Demographics Unstructured Cath Reports Imaging Rad Reports Medical Devices External NIAGADS dbgap UKBiobank KidsFirst 10

11 Microsoft Next Generation Architecture 11

12 Analytics Project Team Structure Governance Analytics Governance Committee Project Directors: Chief Data Info Officer (IS), AVP Strategic Operations) Project Managers: 1 Full time, 1 half-time Technology Platform Data Lake Curated Data Store User Facing Products Data Analytics Center Sr. Manager Data Engineer Data Analytics Center Sr. Manager Data Architect Data Engineers Data Analyst (IS) Ops Director Associate CIO (co-leads) Data Architect Data Engineer Data Analyst (IS) Data Analysts (Ops) Ops Director Associate CIO (co-leads) Data Architect Data Engineer Data Analyst (IS) Data Analysts (Ops) Infrastructure Integrated Areas Security Education Testing 12

13 Analytics Project Team Grid 13

14 Data Domain Committees Ambulatory Operations Hospital Operations Clinical / Quality Operations Access Scheduling Productivity Throughput ED Peri-op Radiology Mortality Complications Readmissions Safety Reported Metrics Finance Marketing / Patient Experience Research Hospital finance Ambulatory Finance Physicians Referrals HCAHPS Research 14

15 Project Timeline Data Domain Committees Define metrics Reporting & dashboard guidelines Ownership / governance Develop Governance Model Analytics Storefront Dashboards: Quality Ambulatory Hospital Ops PPA Data Lake Build Curated Data Store Build User-Facing Products Dashboards Tools Reports Initial Product Launch Apr-Jun Apr-Sept Jul-May 2019 Sept -Apr 2019 May 2019 April 2018 July 2018 January 2019 May

16 Storefront 16

17 Storefront 17

18 Storefront 18

19 Storefront - Education 19

20 Storefront 20

21 Storefront 21

22 Analytics as a core service 22

23 Ongoing Support Move from project to product mentality Regular release cycles Need to organize teams for agile development Need to assign project managers as product managers Big challenge for the PMO 23

24 ROI Stories Denials Analysis 24

25 Claims Total PB $ Claims/ year ~$2B Total PB $ Denials/year at historical rate of ~10% = ~$200M Opportunity Too much Data for them to process without use of an organized approach and system. Struggle with teasing out information from 835 to find root cause of a Denial Some common Denial reasons are: Wrong coding Service is not Medical necessity Wrong insurance vendor No pre-certification Looking for actionable denials data, by payor, by specialty, by department other Dimensions 25

26 Rev Cycle Process 26

27 Denials Explorer 27

28 Denials Explorer 28

29 Denials Explorer 29

30 Denials Explorer 30

31 Denials Explorer 31

32 Denials Explorer 32

33 Denials Explorer 33

34 Model the Denials 34