Vizient delivers smart data-driven resources and insights from benchmarking and predictive analytics to cost-savings for their members. The firm employs a modern data architecture utilizing Hadoop with Data Virtualization to power their data discovery and analytics initiatives.
Discover Vizient’s success in:
· Helping members apply data and insights in new ways to achieve sustainable results
· Integrate Member Spend and Supplier Sales data from all Vizient organizations to identify opportunities for increasing contract utilization
· Enable Single source-of-truth and consistent view of data from distributed data assets
Don’t miss the opportunity to learn from this healthcare innovator!
2. AGENDA
- Who is Vizient
- Modern Data Architecture
- Self Service discovery examples
- Recommendations
- Q&A
3. 3
Who is Vizient?
• Combination of VHA, University
HealthSystem Consortium, Novation,
MedAssets Spend and Clinical Resource
Management and Sg2
• Experts with the purchasing power, insights
and connections that accelerate
performance for members
4. 4
Vizient serves thousands of health care organizations
across the nation, from independent, community-based
organizations to large, integrated systems including
• Acute care hospitals
• Academic medical centers
• Non-acute community health care providers
• Pediatric facilities
Vizient members span the care continuum
5. 5
MEMBERSHIP BENEFITS
• Harness powerful insights
• Accelerate performance
• Achieve scale and efficiency
• Make innovative connections
We measure our success by our members’ success. We fuel
powerful connections that help members focus on what they
do best: deliver exceptional, cost-effective care.
Member-owned, member-driven
• Be more agile
• Build knowledge
• Gain advocates on important
policy issues
6. 6
Unmatched insight and expertise
Top 14
Vizient serves the top 14 hospitals
named to US News and World Report’s
2016-17 Best Hospitals Honor Roll
~$100B
Vizient represents over
$100 billion in annual purchasing
volume — the largest in the
industry.
200+
Vizient member hospitals have
achieved remarkable
improvements in quality and
patient safety through our
Hospital Engagement Network.
More
than 1/3
Vizient provides services
to more than one-third of the
nation’s hospitals.
Information is inclusive of MedAssets Spend and Clinical Resource
Management segment, including Sg2.
We deliver brilliant, data-driven resources and
insights — from benchmarking and predictive
analytics to cost-savings — to where they’re
needed most.
12. Unify disparate financial data marts
relational sources
Primary Use Case: Unify disparate accounting and finance data marts across various
legacy organizations into a logical data warehouse
Secondary Use Cases
• Provide a unified source for key BI initiatives like the GPO Dashboard
• Support reporting needs as legacy systems are migrated or replaced during integration of
Vizient and L-MDAS (dbVision, etc.)
• Provide a final resting place for archived legacy sources like Solomon, Epicor, etc.
12
VHA
MedAssets
UHC
Unified balance
sheet
13. Virtual Financial Data Mart
Architectural Approach
• Denodo was selected as the data platform in
order to utilize the following features of the
software:
– Data Virtualization allows sources in various mediums and locations to be
integrated without physically moving the data
– Data Abstraction allows data to be represented consistently within the
datamart while data sources are moved or replaced behind the scenes
– Data Integration allows for a single seamless view to be created across a
subject area (e.g. “Supplier Sales”) with varied data transformation rules
for each data source within the subject area (PRS, dbVision) allowing a
logical data warehouse to be created without the need to instantiate a
physical on
13
14. Data Intake and Standardization
Varied sources
14
Primary Use Case: Consolidate member data feeds and simplify member data
submission experience
Other Use Cases:
• Support consistent internal data standards
• Feed data to systems regardless of downstream tech stack
• Metadata approach to security, rights of use
• Ground for data governance and data mastering
Architectural Approach
• Data repository utilizes Hortonworks to persist input data as raw data that can be
schematized for format and validation work. Paxata for Data Quality and data
validation. Reuse overlapping datasets while allowing separate schematized views to
be published as needed. Abstract data source from data use while preserving security
and rights of use
• Reporting components match vary by downstream product
Goal of reducing onboarding
time between 10% to 50%
15. Data Intake and Standardization
Key Challenges
• Successful consumption of many formats and business processes into shared
delivery
– Ensuring varied processes can be supported
– Providing flexible mapping capabilities to support many formats
• People change management
– Be sure to manage how people experience the change as well as the
technology of the change
• Scalability/Configuration Management
– Process and tool needs to support parallel development of this project and
continued efforts to fold in other sources
– Process guidelines are being authored to allow for multiple development efforts
on the same datasets
15
16. Consolidated view of sales data
Varied sources
Primary Use Case: GPO Dashboard - Provide a consolidated view of supplier sales data
across all customers of legacy Vizient & Med Assets organizations.
Architectural Approach
• Financial Datamart (on Denodo) for data source
• Denodo TDE Exporter Tool for daily data extracts to Tableau:
– Report Data
– Report User Security
• Tableau for report development and distribution
16
Over 400 active users across 6
departments with one story
17. Consolidated sales data view in virtual data mart
Key Challenges
• Balance between data timeliness and report performance
– Tableau reports performed best utilizing the TDE format (cached/extracted dataset) as
opposed to a live connection
– This meant that the report caches required daily refreshes, and data extraction had to
be appropriately tuned
– Denodo features such as dataset statistics and indexing greatly contributed to this
performance tuning
• Provisioning user security at cell level
– The requirement for some internal report users to be restricted to the
members/customers to which they are assigned meant that a new report security
approach was needed
– Reliance on TDEs for report data necessitated the integration of security in the
reporting layer
– Tableau’s “data blending” feature allows user security to be specified within a separate
dataset
– This also supports reuse of the security view across logical data warehouse views
17
18. Integrate member spend and supplier sales
Varied
Primary Use Case: Contract Sales Analyzer Dashboard - Integrate Member Spend
and Supplier Sales data from all Vizient organizations to identify opportunities for
increasing contract utilization
Other Use Cases:
• Maintain consistency (Single Source Of Truth) with GPO dashboard regarding:
– Supplier Sales Data
– Dimension Data
– User Security
Architectural Approach
• Data source utilizes Denodo to reuse overlapping datasets (sales, dimensions,
security) while allowing separate virtualized views to be created for new datasets
(member spend) which can be also be reused by future projects via a logical data
warehouse
• Reporting components match approach used by GPO Dashboard
18
19. Contract Sales Actualizer Dashboard
Key Challenges
• Successful integration of Exadata RDM as a data source for Denodo
– Approach utilizes the strength of Exadata RDBMS for aggregating large
quantities of data quickly
– Denodo to integrate the data with similar legacy SQL Server data sources to
create a comprehensive view of Vizient member spend
• Scalability/Configuration Management
– Advances were made to support parallel development of this project and
continued efforts on GPO dashboard
– Compartmentalization features within Denodo allow for code changes in each
project to be version controlled and assessed for dependencies
– Process guidelines are being authored to allow for multiple development efforts
on the same datasets
19
20. Other examples
Data Science comparatives – Determine groups analyze with
various Machine Learning tools and publish via accessible sql
Quick service virtual data marts to publish data via API or SQL to
support consulting engagements
Support for “test area” hack day data sets while maintaining
security and confidentiality
20
21. Recommendations
(AKA learn from my mistakes before your own)
Mastering data is hard… publish trusted sources
• One source to rule them all isn’t practical in a distributed architecture world... Use data
services to make it “look like it’s together” (If it looks like it works... It works)
• Don’t make your system complexity your users problem... Abstract complexity in layers
Attaching and publishing isn’t point and click… you need to understand your use
cases or your performance will suffer
Don’t underestimate the people side of technology change... Usability and ease will
win out (work will find a way)
Hadoop is special but not that special... Hadoop is a source of data that has smarts,
don’t treat it as if it’s just another SQL source but also don’t make differentiation
your users problem
21
22. Our central focus is helping members
apply data and insights in new ways to
achieve sustainable results. Our
success is ultimately defined by the
success of our members in serving their
patients and communities.
Curt Nonomaque, President and CEO, Vizient
23. This information is proprietary and highly confidential. Any unauthorized dissemination,
distribution or copying is strictly prohibited. Any violation of this prohibition may be subject
to penalties and recourse under the law. Copyright 2016 Vizient, Inc. All rights reserved.
23
Contact Chuck DeVries at chuck.devries@vizientinc.com
for more information.
Notes de l'éditeur
The next generation of data intake goes beyond automated data submission (ADS) and provides
Flexible toolsets to make data submission simpler
Make data persistence easier
Provide flexibility in how applications consume member information
Differentiation between adaptive coding vs. machine learning vs. full deep learning – Current tools allow separation and capability in these areas
Difference between learning basic data rules and differentiated machine learning problems