Information may be one of the most crucial and differentiating resources in the health insurance industry. Advances in big data technology provide a great opportunity to effectively respond to member inquiries and anticipate health care needs. It can also increase the timeliness and accuracy of provider reimbursements and more efficiently integrate all participants within health care delivery. In this presentation, Andy Ashta, Executive Director, Information Architecture and Data Management at Health Care Service Corporation, will discuss how the organization has strategically integrated big data technologies into their data topology, heavily leveraging streaming data to connect applications and services, notifications, and alerts to improve the customer experience.
Delivering Healthcare Value Through Transformation to Big Data Streams
1. Health Care Service Corporation, a Mutual Legal Reserve Company
Delivering Healthcare Value
Through Transformation to Big
Data Streams
ANDY ASHTA
Executive Director, Information Architecture & Data Management
3. 3FOOTER
Who We Are?
#6on Diversity MBA’s 50 Out Front for
Diversity Leadership Best Places to Work
for Women & Diverse Managers
Our Purpose
To do everything in our power to stand with
our members in sickness and in health ®
208.3million
claims processed
annually 1936
year foundedLARGESTcustomer-
owned health insurer in
the U.S. and 4th largest
overall
+$1billion
in IT spend
Over
21,000
employees
15million
members
2,100
IT employees
Operating Blue
Cross and Blue
Shield plans in
FIVE states: IL,
MT, NM, OK, TX
4. 4FOOTER
About Me
Executive Director,
Information Architecture & Data Management
Industries
Financial Services, Healthcare, Digital, Publishing
Roles
Data Development, Business Intelligence, Data Architecture, Application Development
Andy Ashta
6. 6FOOTER
✓ Changing analytics in healthcare
✓ Organizational focus
✓ Data Stream Use Cases &
Outcomes
Healthcare Data Ambitions
7. 7FOOTER
User Expectations Have Evolved
Clinical notes, call transcript,
web logs, etc. Variety of data
has increased.
As soon as an event occurs, “can
you make me aware”. “Can I interact
with the member now?” Velocity of
information exchange has increased.
Analyzing 200M, 300M rows. Analyze all
medical and pharmacy claims over a 5
year period. Verbose logs from
application servers for security purposes.
Volume has increased.
“This data can’t be wrong.”
Confidence (veracity) of data
needs to be ironclad.
8. 8FOOTER
Data Processing In Healthcare Has Slowly Evolved
Transactional
Reporting
Dashboards
Descriptive
Predictive
Prescriptive
AI/ML
Juncture
point
where the
“game has
changed”
The
technologies
that got us
here …
… will not get
us here!
9. 9FOOTER
Matching Technology For These Processes Have Evolved
Transactional
Reporting
Dashboards
Descriptive
Predictive
Prescriptive
AI/ML
Mainframe
Relational
Database
Distributed
Relational
Database,
Postgres
Hadoop Hadoop,
NoSQL
Data Science
Data Streaming,
Machine Learning
10. 10FOOTER
Value Drivers Are Increasingly Realtime
Understand
population
activity
Understand
individual
behavior
Possess
location and
current activity
Connected
Ability to interact in a
meaningful and
valuable dialog
11. 11FOOTER
Value Drivers Are Increasingly Realtime
ACA
Enrollment
Claims
monitoring/
dash-boarding
Hospital
admission
Complexity
of claims
processing
Pharmacy
abuse
Negligence/
Fraud/
Abuse
18. Data Awareness
• Logging data changes
• Machine learning of those
changes and events
Claims Processing
• Realtime understanding of
claims inventory
• Machine learning of claim
complexity
• Irregularities in claims
Digital Interaction
• Realtime updates
• Relevant data availability
PURSUIT OF USE CASES AND CAPABILITIES
Data Stream Outcomes At HCSC
19. Need
• Create proactive data
processing capability
• Reduce anomalies
and irregularities
within HCSC data
Pattern
• Machine learning of data
stream
Outcome
• 360M records / day
• 2KB record size
• Enhanced visibility of
data processing
PURSUIT OF USE CASES AND CAPABILITIES
Data Awareness
20. Need
• Create a more expansive and
dynamic view of claims inventory
• Improve processing efficiency of
more complicated claims.
• Find irregularities in medical
claims submissions
Pattern
PURSUIT OF USE CASES AND CAPABILITIES
Claims Processing
• Machine learning of data stream
• Collection, analysis, and
visualization
21. Execution Outcome
• Python reading Kafka claims topic.
• Unstructured learning with
KMeansClustering in TensorFlow
• Cluster samples to-
– Claims Complexity
– Claims Irregularity
• Dashboarding
– Later this year
PURSUIT OF USE CASES AND CAPABILITIES
Claims Processing
22. Need
• BlueAccess for Members
(BAM)
• Real-time member and
coverage updates
Pattern
• Spring Cloud Stream to Gemfire
• Moving to Spring Data Flow
PURSUIT OF USE CASES AND CAPABILITIES
Digital Interaction
23. Execution Outcome
• 2M full member canonicals / day
PURSUIT OF USE CASES AND CAPABILITIES
Digital Interaction
24. Need
• Member Preference Interaction
History (MPIH) application
• Real-time member preference and
incentive updates
Pattern
PURSUIT OF USE CASES AND CAPABILITIES
Digital Interaction
• Spring Kafka
25. Execution Outcome
• 320 messages / min
PURSUIT OF USE CASES AND CAPABILITIES
Digital Interaction
27. DESIGNING AND IMPLEMENTING THESE CAPABILITIES AT HCSC
Changing Landscape
• Value Drivers
are Real-time
• Most healthcare
firms possess
some degree of
legacy data
platforms
Needs Organizational
Alignment
• Technology +
(Business)
Functional
Domains
• Tactical +
Strategic Scope
Strategic Outcomes
• Increased Data
Awareness
• Improved
Medical Claims
Processing
• Digital
Interaction
Delivering Healthcare Value