Presentation by Rich Pollack, VP and Chief Information Officer, VCU Health, at the marcus evans National Healthcare CIO Summit held in Pasadena, CA March 13-14 2017
An Analytics Journey: Case Study over Seven Years-Rich Pollack, VCU Health
1. AN ANALYTICS JOURNEY: CASE
STUDY OVER SEVEN YEARS
RICH POLLACK CHCIO, FCHIME
VP&CIO
VCUHEALTH
1
2. THE VCU HEALTH SYSTEM
2
• VCU Health System
• Only academic medical center in Central Virginia
• 2014 McKesson-AHA Quest for Quality Prize Recipient
• Referral center for the state and Mid-Atlantic
• 37,000 admissions and 650,000 outpatient visits
• MCV Hospitals and Clinics
• Teaching hospital of the VCU Health System
• 1125 licensed beds
• 90,000 patients are treated annually in the hospital’s emergency department,
which is the region's only Level I Trauma Center
• 100+ primary care, specialty and sub-specialty outpatient clinics
• NCI designated Cancer Center
• 11,000 employees
• MCV Physicians
• 720-physician, faculty group practice
• Virginia Premier Health Plan
• 200,000 member Medicaid Health Plan
3. 7 YEAR JOURNEY
3
2009-11
2012
2013
2014
2015
2016
Outside Consultant
Engaged an
outside firm with
deep expertise to
conduct an
assessment
(current state –
future needs) and
optional
recommendations
for moving
forward including
management and
governance
structures.
Educating my peers
Slowing building
the case thru 1:1
sessions with
over 20 leaders
to help them
understand the
current state and
future need and
how to close the
gaps
Approval of EA Effort
After two abortive
attempts, was able to
get the endorsement
of senior leadership
to elevate analytics,
consolidate siloes,
rename the service,
establish new
leadership, setup
data governance
structure, and
increase investment
in Analytics
Enriching the EDW
Began populating the
Oracle tables with
EMR data via direct
ODBC feed from
EMR production
node. Began transfer
of EMR
data/reporting
expertise from EMR
team to EA team
Current State
17 person EA team.
Director reporting to
corp VP of Strategy.
Data Architect, ETL
specialist, Tableau
visualization tool,
genomic data,
serving researchers
with clinical data.
Mature data
governance
Maturing Data Gov.
Brought in
outside experts in
data gov., and
healthcare
analytics to help
catalyze our
efforts.
Developed robust
committee
structure led by
COO with broad
representation.
Started sub
committees
focused on MDM
and data integrity.
4. CURRENT STATE IN 2006
4
• Small 5 person decision support team reporting to CMO
• EDW migrated from mainframe to C/S Oracle and Cognos
• Primarily populated with billing data in star schema
• 4 to 5 other “siloed” data marts existed
5. ABORTIVE ATTEMPT TO USE EMR VENDORS DATA
WAREHOUSE FOR CLINICAL ANALYTICS
5
• Incompatible tools Business Objects versus Cognos
• Inability to easily import non EMR data
• Little focus on addressing non clinical reporting needs
• Lack of data management tools (MDM/Data Dictionary)
6. DISCUSSIONS WITH CEO 2008 – 2009
6
• Change in IT strategic focus to emphasize Analytics
• Predicted dramatic growth in need and resources
• Discussed need for different structure, elevation in the org,
elimination of siloes and enlargement of team
• Need for data governance and data management structure
and process
• Charged CIO to begin peer educational effort
8. 2008 DISTRIBUTION OF DATA MARTS:
8
CTSA DW
I2B2
(1 FTE)
Cerner DW
Business Objects
(2 FTE)
IDX
Globalworks
Cognos8
(2 FTE)
DSS
Cognos8
(6 FTE)
Core Transaction Systems
GE/Lawson/Cerner/etc
Cancer Center
DW
(4 FTE)
VPHP
DW
(~20 FTE)
OHI
Population Health
(1 FTE)
If left unaddressed,
continued future
growth of data silos
9. PEER EDUCATION FROM 6/2009 THRU 6/2011
9
• Development of a comprehensive educational presentation
illustrating where we needed to go and the gap that existed
between that future state and current capability.
• Persistent 1 on 1 meetings and discussion with peer VPs
and other leadership across the health system.
10. ENTERPRISE BUSINESS INTELLIGENCE
AN EXPLORATION OF THE STRATEGIC IMPLICATIONS AND POSSIBILITIES FOR THE
HEALTH SYSTEM
10
Rich Pollack MS CPHIMS FHIMSS
VP & CIO
VCUHS
September 14, 2010
11. EXECUTIVE SUMMARY
11
To succeed in this new era of accountability and regulatory prescriptions, we need to possess a level of insight and analysis around
operations, finance, clinical care and research that has heretofore been absent. To achieve that future state there will be a need for
investment in either incremental or dramatic improvements in our core analytics capability.
Today, we believe there exists a “gap”, between current delivery of reports and analysis via DSS (and other data marts) – and the very
real needs of decision makers for reporting that is more multidimensional, for financial modeling capability and for easy patient cohort
queries in near real time. Accordingly we need to:
Understand the needed roadmap to realize the unmet potential of our of our rich underlying business/clinical data set to provide actionable real time
intelligence across patient populations and venues of service delivery
Calculate and establish the effort needed to raise our analytic delivery engine to equal those healthcare organizations much further along in executing best
practice in this space
Geissinger
Partners
Northwestern Memorial
Ohio State
Intermountain
Mayo
12. EXECUTIVE SUMMARY (CONT.)
12
Specify the particular use case objectives for the clinical, operational and financial domains based upon business/clinical goals and priorities clearly articulated by
leadership.
. Evaluate the current infrastructure and tools available in-house and from VCUHS’ core vendors, and understand the “gap” between anticipated needs and vendor
capabilities.
Finally, we will define VCUHS future data integration and warehouse strategy to ultimately provide robust business and clinical intelligence support for both current
and future needs.
Therefore, our intent is to initiate an early planning process to identify VCUHS’ critical business and clinical dashboard
metrics, performance reporting requirements and clinical research needs that may be met through improved data
warehouse technologies and data integration.
We will map those requirements to current analytics capabilities and infrastructure, and will develop the go-forward plan to
leverage current tools, as well as develop the future VCUHS analytics and data warehouse architecture and resources.
13. Strategy
Development
Capital
Allocation
Revenue Cycle
Management
• In what services do I make money? Lose money? Why?
• In which services should I invest? Divest? Improve financial performance?
• How do my hospitals services compare across the health system?
• How can I optimize service location? Should I consolidate services?
• What is my profitability across service lines by hospital?
• Where should I focus my cost reduction initiatives? How do I track success?
• Why are costs for X service increasing, is it labor, supplies or physician practices? How can we decrease
these costs?
• Are patient care processes following protocols? Why not? Is it physician or patient complexity driven?
• What physician should set standards for others?
• What impact to my bottom line will a rate change create? Can I afford to sign this managed care contract?
Can I afford not to?
• What types, how much and where are my denials originating?
• How relative is my pricing to cost? Comparatively throughout my system?
• Are my payors paying correctly? Timely?
• Can I compare my payors profitability and score them?
• Am I charging for all the services I am performing?
Operations
Improvement
• What services should I invest capital? What financial return can I expect?
• What is the ROI of a capital purchase?
• What incremental ancillary capacity will be affected by additional capital purchases?
Quality,
Patient Safety
and P for P
• What metrics do I measure and how do I measure them?
• How do I get information to front line caregivers and managers to make a difference?
• How can I efficiently collect and report quality and patient safety metrics?
INFORMATION NEEDS TO BE TIMELY AND RELIABLE TO BE USED FOR STRATEGIC DECISION-
MAKING AND DERIVING OPERATIONAL INSIGHT
13
14. SAMPLE CLINICALLY RELEVANT YET DIFFICULT QUESTIONS TO ANSWER
14
• What are the incremental cost reductions. Length of stay reductions and improved outcomes for
patents treated using the CAP protocol compared to those not on the protocol?
• How did treatment differ by physician for DRG stratified patient cohorts (using EMR data)?
• How many pneumonia cases were re-admitted for pneumonia within 6 month by provider by co-
morbidities?
• How many patients on heparin have experienced a platelet count drop of 15% in the last 24 hrs?
• Which nurses have the most contact with patients who test positive for MRSA?
• How often does each resident internist ignore drug interaction alerts?
15. 15
Understanding Levels of Business Intelligence and Advanced Analytics
Analytics go beyond standard business intelligence capabilities and now draw on
advanced analytics and predictive modeling techniques from efforts developed in various
industries
What best practices or innovation should we
adopt?
Evidence-based innovation
What outcomes should we expect? Predictive modeling
What patterns can we leverage? Information discovery
What if these trends continue? Forecasting / extrapolation
Why is this happening? Statistical analysis
Alerts What actions are needed?
Dashboards / benchmarking How are things going - relative to plan?
Drill down reports & analytic pathways Where exactly is the problem?
Ad hoc reports How many, how often, where?
Standard reports What happened?
Advanced
Analytics
Standard
BI
CompetitiveAdvantage
Source: Adapted from a graphic used in Tom Davenport’s Competing on Analytics
16. • Engagement of a 3rd party consultant to independently lay
out roadmap for organizational governance and structure.
fully engaged all leadership. 6/2011 – 4/2012
16
Next:
Outside Validation and Guidance.
17. OUTSIDE CONSULTANT FINDINGS
17
• Significant number of able “analytics employees” (>35) are already in
place/funded (just missing a few key skills (enterprise data architect/modeler;
ETL/parser developers)
• Immediate opportunities exist to merge Core system teams and applicable
data marts
• Further opportunities exist to consider leveraging non-Core systems; health
systems need to increasingly be more payer-like (e.g., value based purchasing)
• Enterprise data governance is nascent, ad-hoc and informal contributing to
some stakeholder data mistrust (ugly data syndrome)
23. PRESENTED PROPOSAL TO SENIOR LEADERSHIP -
11/2012 - 1/2013
23
• 2 tries required before endorsement
• DSS renamed EA with new VP at corporate level
• Commitment made to greatly expand staffing, consolidate
siloes and improve technology
25. 25
Data Governance Executive Committee
COO (Chair), CIO, CMIO and most operational VPs
Representatives from university, research and HMO
Data Management Committee (Additional Members TBD)
Abdoul Sousseh (chair) Jeremy Utz Jo Weller Lori Lynch Scott Wead Rusty Pitts Arthur Palmer
Bonita Walker Melissa Wimmer Liz Locus David Summers
DM Task
Groups:
•Master Data
management
•Data
Dictionary/Metadata
•Data
Stewardship/Ownership
•Data Integrity
Technical
Advisory Panel:
Overarching Governance for Analytics Efforts
Overarching Data Governance Council & Process
Provider Side Virginia Premier VCU School of Medicine
This approach allows VCUHS to continue the current highly responsive delivery of information to departmental leadership while deploying the tools and teams to move
eventually towards a single enterprise data warehouse
At the team level, each employee reports to a single manager to support departmental BI/Analytics needs while simultaneously being
available for collaborative exchange of knowledge and joint solution efforts with other departmental BI/Analytics teams.
Team Level
MCVP
GlobalWork (2)DSS (6)
Cerner
EDW (2) I2B2 (2)20
Overarching Data Governance Council & Process
26. ENRICHING EDW WITH CLINICAL EMR DATA 2013 - 2015
26
• Achieved thru a direct OBDC connection (Oracle to Oracle
table joins) from production EMR (remote hosted) to EDW
• Re-align EMR report writing specialists from EMR team to
EA to provide clinical data SME badly needed by EA analysts
27. LEVERAGING EA AND EDW TO MEET MU 1 AND MU 2
REQUIREMENTS – 2012 - 2015
27
• Because clinical data and billing data exist in
disparate systems and because physician-patient
relationship differs in both venues, the EDW was
the only solution to merging the clinical data with
billing based patient-physician identity
• Result was a highly effective and accurate
physician performance dashboard out of the
EDW that is still in use today
29. FURTHER ENRICHING THE EDW WITH NEW DATA SOURCES
2016 - 2017
29
• Have now (2016) begun to populate structured genomic test
data (patient specific) into the EDW
• Began more extensive engagement with the research side of the
health system to include helping lead a research EDW effort.
• Engage with the population health team to provide a targeted
data mart derived from the EDW.
30. EAW
Oracle DB
Cerner Standby
Cerner
Millennium
Production
GE HPA/MCVP
IDX
Midas
Surginet
EnterpriseAnalytics InternalDataSources
HPD
Pharmacy
HR - ULTI
INTRANET
VCC
Teletrackinga
nd Transport
Lawson
Clinical
Billing
OR
Pathology
Registry
OHI
BIC
CTED/
MDAS
Subject
Matter Areas
Analytic
Group
Partnerships
Clinical
Data
Sources
Operational
Data Sources
Outbound Connection
Inbound Data Extract
30
UNOS
31. MATURING DATA GOVERNANCE PROCESS - TODAY
31
• DG Committee meets monthly with broad exec participation
as members. Chaired by COO.
• Sub-committees formed for data integrity, MDM etc.
• Extends beyond health system to Research and University
community
32. Major IT Systems, e.g., Cerner, Lawson, IDX, etc.
Overarching Data Governance Council & Process
Enterprise Analytics Virginia Premier VCU School of Medicine
This approach allows VCUHS to continue the current highly responsive delivery of information to departmental leadership while deploying the tools and teams to move
eventually towards a single enterprise data warehouse
At the team level, each employee reports to a single manager to support departmental BI/Analytics needs while simultaneously being
available for collaborative exchange of knowledge and joint solution efforts with other departmental BI/Analytics teams.
Team Level
13 I2B2 (2)20
Overarching Data Governance Council & Process
Data Governance Executive Committee
Chief Operation Officer (Chair)
Data Management
Committee
Data Architect(Co-Chair) ,
Director of Analytics(Co-Chair)
Data Steward Workgroup
Patient Access Systems Manager (Chair)
EDW Steering:
CIO (Co-Chair)
CRIO (Co-Chair)
DATA GOVERNANCE - STRUCTURE
32
33. Data
Governance
Documentation
Stewardship
Policy
Quality
33
Charter- Data Governance Steering Committee
Purpose
Recognizing that data is a key organizational asset, the Data Governance Executive
Steering Committee will provide guidance, direction and prioritization for the
enterprise data governance program.
Scope
Set goals and priorities for the Data Management Committee
Communicate the importance of data governance as a priority for
all IT projects
Remove barriers that might prevent data governance from being
fully accepted by the organization
Sponsor initiatives designed to improve data quality and drive
towards the enterprise goal of “One Source of Truth in Reporting”
Annually measure the performance of the Data Management
Committee and the extent of organizational adoption of key principles
34. 7 YEAR JOURNEY
34
2009-11
2012
2013
2014
2015
2016
Outside Consultant
Engaged an
outside firm with
deep expertise to
conduct an
assessment
(current state –
future needs) and
optional
recommendations
for moving
forward including
management and
governance
structures.
Educating my peers
Slowing building
the case thru 1:1
sessions with
over 20 leaders
to help them
understand the
current state and
future need and
how to close the
gaps
Approval of EA Effort
After two abortive
attempts, was able to
get the endorsement
of senior leadership
to elevate analytics,
consolidate siloes,
rename the service,
establish new
leadership, setup
data governance
structure, and
increase investment
in Analytics
Enriching the EDW
Began populating the
Oracle tables with
EMR data via direct
ODBC feed from
EMR production
node. Began transfer
of EMR
data/reporting
expertise from EMR
team to EA team
Current State
17 person EA team.
Director reporting to
corp VP of Strategy.
Data Architect, ETL
specialist, Tableau
visualization tool,
genomic data,
serving researchers
with clinical data.
Mature data
governance
Maturing Data Gov.
Brought in
outside experts in
data gov., and
healthcare
analytics to help
catalyze our
efforts.
Developed robust
committee
structure led by
COO with broad
representation.
Started sub
committees
focused on MDM
and data integrity.