teradata next-generation healthcare analytics · 2019-12-06 · • big data analytics . 5 new...
TRANSCRIPT
1
Teradata
Next-Generation Healthcare Analytics Nov 25th , 2016
Ozgur Kaynar
Senior Business Consultant [email protected]
2
• Healthcare Trends & Challenges
• Teradata Approach to Healthcare
• Teradata Advanced Analytics
• Health Sector Analytical Use Cases
Agenda
© 2014 Teradata
3
The Triple Aim of Healthcare for All Stakeholders
Better Care for Individuals • Safety and Effectiveness
• Patient-Centeredness
• Timeliness & Efficiency
• Equity
Better Health for Populations • Improve Outcomes
• Expand Access to Quality Care
• Influence Patient / Member Lifestyle Behavior
Reducing Per-Capita Costs • Administrative Efficiencies
• Value-Based Reimbursement
• Coordination of Care / Care Management
• Revenue Recovery
4
Healthcare Industry Challenges
Adaption Challenges • Expanding Coverage (ACA)
• New Products
• Payer and Provider Collaboration
Chronic Conditions Frequency & Cost Are Rising • Type 2 Diabetes
• Asthma
• Heart Disease
• Obesity
Population Requirements • Care Improvement&Coordination
• Patient / Member Engagement
• Lifestyle Choices Drive Costs Up
New Technology • New Communication Channels
• Social Media, Facebook, Twitter
• mHealth, Telemedicine
• DIY Healthcare (i.e. FitBits)
• Big Data Analytics
5
New Technologies, Adaption of Skillsets
Main Technology Challenges for Healthcare
© 2014 Teradata
Data & Application Silos
360 Patient Information
High Grow rate of Unstructured Data
Data Governance
Quality & Security
6
• Healthcare Trends & Challenges
• Teradata Approach to Healthcare
• Teradata Advanced Analytics
• Health Sector Analytical Use Cases
Agenda
© 2014 Teradata
7
A New Approach is Required…
KNOW MORE. DO MORE.
• Leverage your existing ecosystem and capabilities
• Provide an integrated approach to harness all your data
• Data reuse to speed time to market and improve operational expenditures
Enable healthcare systems and insurer
High Quality, Low Cost Care Coordination
Consumer Engagement Driving Wellness / Growth
Integrated Finance Performance Management
Improved Population Health
8
Healthcare Industry Framework
9
Teradata Framework
Industry Solutions
•Deep Industry Knowledge
• Industry Models, Partners and Teradata Services
Professional Services & Partners SIs
•Best in class
• Expertise
• 99% retention rate
Applications
• Leading Integrated Marketing Management Solutions
•Marketing Analytics and Customer Data Management
•Alliance Program
Analytics Platform
• The world’s most scalable analytics platform for structured, multi-structured and Big Data analytics
• Leader in Gartner
•Open standards
UNIFIED DATA ARCHITECTURE DATA MODELS SERVICES APPLICATIONS
Discovery Platform
Business Consulting
Architect & Implement
Optimize & Manage
Analytical
Partner
Solution Modeling Building Blocks
Data Integration Roadmaps
Health Care Data Model
APPLICATIONS
DATA MODELS
UNIFIED DATA ARCHITECTURE
SERVICES
Integrated Marketing
Management
Data Platform
Integrated Data
Warehouse
10
• Healthcare Trends & Challenges
• Teradata Approach
• Teradata Advanced Analytics
• Health Sector Analytical Use Cases
Agenda
© 2014 Teradata
11
Analyze all Data: Traditional + New Data Types
Business Transactions (claims, member billing,
enrollment, provider
payments)
Observations (clinical notes, UM and CM
notes, MRI’s, machine
sensors, medical charts)
Source: IDC, Gartner
Interactions (IVR, call center, hospital
admits, “likes”, authorizations,
tweets, weblogs, )
+ +
13 © 2014 Teradata
Healthcare Analytic Discovery Success
1. Invest in Discovery
Ask new (previously unanswerable) questions
Move from focus only on transactions to interactions
14 © 2014 Teradata
Healthcare Analytic Discovery Success
2. Analyze All Data
Start with your cleansed data that has been
integrated in your enterprise data warehouse
Access other valuable data in the ecosystem
Use new data which is never involved to analytics (
semi and unstructured data )
Use Analytics forAll Data – not subsets
15 © 2014 Teradata
Healthcare Analytic Discovery Success
3. Make it easy to use !
Analytics in Healthcare - Time sensitive - has to be quick
- With minimum effort & skillset – not a SW company ?
Choose big data analytics technologies that - Support iterative analysis with very short cycles
- Connect to familiar BI & stats tools
- Shorten time-to-answer (and run-time) How do we do that? We connect your favorite BI &
Stats tools to ALL the data using
the latest analytics power tools
16 © 2014 Teradata
Healthcare Analytic Discovery Success
4. Embrace multi-structured data
Nothing to be afraid of – text, sensor, web-log, social
media data doesn’t bite
You may have to learn new data management or
analytics techniques
There is a lot of it – estimates range from 70%-90% of
enterprise data is multi-structured
17 © 2014 Teradata
Healthcare Analytic Discovery Success
5. Ramp up with an experienced team
You have a very capable IT staff and brilliant
healthcare analytics leaders – still, ask for use cases
Specific technology, tool and data scientist
experience will jump-start your project
Shadow the experienced team, learn and take
over if you like
How do we do that? We have the largest professional
workforce dedicated to big data
analytics, +250 by 2016
18 © 2014 Teradata
Healthcare Analytic Discovery Success
6. Plan to democratize analytics
Don’t limit analysis to a few highly paid data
scientists
Don’t have to send every business analyst to
training for Hbase, MapReduce, Impala, Hive & Pig
This early decision will affect the tools and partners
you choose How do we do that? your favorite BI & Stats tools
connect to all the data using the
latest analytics tools
19 © 2014 Teradata
Healthcare Analytic Discovery Success
7. Measure the VALUE
Estimate business Value & priority and try to reach
that in a limited time/resources/simplicity
Talk with Value and Benefit for Sustaning analytics in
the company
Try before you buy !
How do we do that? We want you to be convinced
with the value by answering your
questions on your data
20
• Healthcare Trends & Challenges
• Teradata Approach
• Teradata Advanced Analytics
• Health Sector Analytical Use Cases
Agenda
© 2014 Teradata
21
Population Management
Out of Area/Network Utilization
Atypical Antipsychotic Utilization Trends
Predictive Path to Surgery
Patient
Admission to
Discharge
Analysis
Drug & Medical Condition Affinity Analysis
Missing
Diagnosis
(Risk)
Prediction
Operations
Claims Overpayment Review
Text Analysis -HL7
Call Center
IVR
Optimization
Marketing Sales
Web Based Insurance Path to Purchase
Attribution of
marketing
campaigns
Patient
Satisfaction
Reduce Complaints to Medicare:
Member & Provider Interactions
Healthcare Discovery Use Cases
22 © 2014 Teradata
HL7 DATA PARSER PATIENT AFFINITY & LENGTH OF STAY PATIENT COMPARE ADMISSION DIAGNOSIS PROCEDURE PATHS PATHS TO SURGERY DIABETES PATIENT READMISSION DRUG PRESCRIPTION AFFINITY ANALYSIS HEALTHCARE IMPACT ANALYSIS SUSPICIOUS DOCTORS
SHAREABLE, REPEATABLE APPS WITH TERADATA ASTER APPCENTER
23
HL7 DATA PARSER
PARSE SEMI-STRUCTURED DATA FOR CARE INSIGHTS
24
25
RAW HL7 DATA IS TOUGH TO ANALYZE
26
HL7 PARSER APP MAKES THE DATA READABLE FOR ANALYSIS
WHICH IS USEFUL FOR ; DETECTING DISCREPANCY between HL7 and TREATMENT or DETECTING ANOMALIES caused by FRAUD
27
Population Management
Out of Area/Network Utilization
Atypical Antipsychotic Utilization Trends
Predictive Path to Surgery
Patient
Admission to
Discharge
Analysis
Drug & Medical Condition Affinity Analysis
Missing
Diagnosis
(Risk)
Prediction
Operations
Claims Overpayment Review
Text Analysis -HL7
Call Center
IVR
Optimization
Marketing Sales
Web Based Insurance Path to Purchase
Attribution of
marketing
campaigns
Patient
Satisfaction
Reduce Complaints to Medicare:
Member & Provider Interactions
Healthcare Discovery Use Cases
28
HEALTHCARE PATIENT READMISSION
PREDICT READMISSION BASED ON PATIENT METADATA AND USING PREDICTIVE ANALYTICS
29
30
SELECT DIFFERENT METADATA FIELDS TO PERFORM THE ANALYSIS AGAINST
31
PLOT PREDICTIVE OUTCOMES BASED ON DIFFERENT FACTORS
32
NAÏVE BAYES DASHBOARD SHOWS READMISSION PREDICTIONS
33
GLM DASHBOARD SHOWS READMISSION PROPENSITY
34
Aster Analytics Evolving Use & Value Examples Affinity & Influencer Analysis: (Product, Service, Social, Warranty)
Predictive Analysis (Behaviors, Components, Social,…)
Behavioral (paths & pattern sequences)
Text Analytics
(sentiment, documents, voice of customer )
36
Start building solutions with BUSINESS MIND
Consider always BUSINESS PAINPOINTS, STRATEGY, IMPROVEMENT AREAS
Prioritization and Measure with VALUE
Use Frameworks, Templates and have a ROADMAP
Fail Fast, Succeed Fast – Become AGILE
Implement, Evaluate, Improve with BUSINESS CONSULTING
Plan and Pilot a BIG DATA Program
Ask TERADATA to become YOUR STRATEGIC PARTNER
Closing Thoughts
© 2014 Teradata
37
For more information, please visit:
www.teradata.com
www.teradata.com/blogs
www.teradatamagazine.com
www.forbes.com/sites/teradata
38
• Real Use Cases
Appendix
© 2014 Teradata
39
• Could you use a predictive model to assess the likelihood of readmission?...there’s an App for that
© 2014 Teradata
Patient Readmissions
A Closer Look
65% Medicare Readmissions
deemed Avoidable
$26B Medicare Costs (2015) from
Readmissions within 30 days
Predictive Model
Quality of Care
Patient Satisfaction
Operational Efficiency
Cost Control
Population Management
Profitability
40
• Predicting patient readmission using several analytic methods – Naïve Bayes
– GLM
– Random Forest
• Using various features of the patient – Time in hospital
– Number of procedures in last year
– Medications and past diagnosis history
– Etc.
• Visualize output in AppCenter and Tableau server
Diabetes Patient Readmission
© 2014 Teradata
Predictive Model
41
• Want to identify doctors with suspicious behavior?...there’s an App for that
© 2014 Teradata
Anomalous Profiles: Medicare/Medicaid Fraud
A Closer Look
$60B+ Conservative estimate of
Medicare Fraud each year Behavioral Analytics
Cost Control
Population Management
Fraud Prevention
42
• Uncover doctors that are out of step with their expected behavior
• Analytics performed on public CMS Part B provider dataset
• Utilize SQL-GR function pSALSA to explore the distinct features of similarity between doctors
• Visualize using Tableau server and AppCenter Sigma
Healthcare Fraud Anomalous Doctors
© 2014 Teradata
Behavioral Analytics
43
• Do you need to illuminate all clinical pathways traversed to optimize a patients journey?...there’s an App for that
Barriers impeding Quality of Care and Patient Satisfaction:
• Poor continuity and coordination across disciplines
• Specialists acting in silos yield suboptimal care
• Lack of an aggregate view across all patients
© 2014 Teradata
Path Analysis: Admission Diagnosis Procedure
A Closer Look
Path
and
Pattern
Quality of Care
Patient Satisfaction Operational Efficiency
Cost Control Population
Management
Processing Efficiency
44
• Discovery analysis to better understand patient path/touch points
• What are patients doing after they are admitted with x disease? – What is my golden path(common expected behavior)?
– Are there any outlier sequences of events?
– Could patterns of behavior be utilized for a predictive model?
• Utilize patented n-Path SQL-MR function for path and pattern analytics
• Visualized with Sankey, Tree, and Sunburst visuals
Admission Diagnosis Procedure Pathways
© 2014 Teradata
Path and Pattern
45
AppCenter Healthcare Prebuilt Analytics Templates Repeatable Applications, Shareable Results
Analytic Templates
Available Now:
• Admission Diagnosis Procedure Paths
• HL7 Parser
• Impact Analysis • Paths to Surgery
• Patient Affinity and LOS
• Patient Compare
• Diabetes Patient Readmission
• Drug Affinity Analysis • Suspicious Doctors
• Suspicious Doctors Similarity graph