© Cloudera, Inc. All rights reserved.
DIGGING FOR GOLD: DELIVERING IMPROVED PATIENT
OUTCOMES THROUGH ADVANCED ANALYTICS
Ryan Swenson, Dir. of Healthcare and Life Sciences Solutions, Cloudera
John Spooner, Sr. Analyst Internet of Things, 451 Research
Nathan Salmon, Chief Architect, MetiStream
Jawad Khan, Dir. Data Services and Knowledge Management, Rush University Medical Center
© Cloudera, Inc. All rights reserved. 2© Cloudera, Inc. All rights reserved.
TODAY’S AGENDA
• Industry overview: John Spooner
• MetiStream Ember demo: Nathan Salmon
• Rush University Medical Center use case: Jawad Khan
• Q&A
The best-in-class organizations use Cloudera
© Cloudera, Inc. All rights reserved.
Achieving Healthcare Outcomes
with AI
• John Spooner, Senior Analyst, Internet of Things
•June 26, 2018
© Cloudera, Inc. All rights reserved.
Insight
Market Insight
Your radar into the competitive IoT landscape, with daily analysis of
the market delivered to your inbox.
Technology &
Business Insight
In-depth analysis of key technologies and players driving the IoT
market: Connectivity, Security, Software, Cloud, etc.
Data
Customer Data
Enterprise and consumer perspective on the early adoption and
demonstrable benefit of IoT, captured in quarterly surveys.
Market Data
Market sizing and forecasting of the myriad device types and radios
driving the IoT revolution, built from bottoms-up analysis.
M&A
A complete database of M&A activity for IoT technology vendors and
service providers.
TODAY’S SPEAKER
John Spooner
Senior Analyst,
Internet of Things
Research Channel
INTERNET OF THINGS
3
© Cloudera, Inc. All rights reserved.
4
© Cloudera, Inc. All rights reserved.
7
5
Healthcare is a
continuum of journeys
© Cloudera, Inc. All rights reserved.
Setting the stage for AI in healthcare
• AI/ML promises to be an effectiveness multiplier
for clinicians/researchers, speeding diagnosis and
treatment for patients and caregivers
• Clinical impact:
• Quicker decisions, based on more data, increase
efficiency/throughput, saving costs and improving revenue.
• Supports new remote/proactive monitoring and treatment;
complements the EHR.
Patient impact:
• Quicker diagnosis and treatment for immediate needs.
• Emphasis on remote monitoring and long-term care planning.
6
© Cloudera, Inc. All rights reserved.
What is AI?
• The terms AI, machine learning and deep learning are used
interchangeably and not always correctly.
• It’s best to think of them as a nested hierarchy.
7
© Cloudera, Inc. All rights reserved.
What is AI in healthcare?
The application of AI in healthcare has three modes:
8
© Cloudera, Inc. All rights reserved.
AI adoption in healthcare?
• Market conditions:
Providers/clinicians want AI & ML –
and are already broadly using it.
Researchers are also utilizing AI.
Payers: Beginning to support AI and
invest in the technology.
Vendors: Already moving to productize
the three tiers we outlined.
9
© Cloudera, Inc. All rights reserved.
AI adoption in healthcare?
• Market opportunity:
451 Research data shows only half of the healthcare
companies we have surveyed use AI or are considering it;
the rest have no AI strategy!
Legislation generally supports AI:
• US regulators generally support AI (e.g., the FDA fast tracking
AI/algorithm and AMA AI guidelines).
• Global attention; e.g., EU declaration on AI.
Payers will continue to invest.
10
© Cloudera, Inc. All rights reserved.
Things to consider going forward
11
© Cloudera, Inc. All rights reserved. 14© Cloudera, Inc. All rights reserved.
METISTREAM EMBER
Nathan Salmon, Chief Architect
Founded 2014
~5K members
Open Source Advocates
Spark & Hail
Trainers
MetiStream Highlights & Key Clients
Rush Univ Medical Center:
▪ Improved Rush’s Adenoma
Detection Rate (ADR) KPI by
50% by enhancing their
ability to process and
analyze unstructured clinical
notes data.
▪ Processed and annotated
two years of clinical notes
data in less than 36 hours.
▪ Scaled up Cloudera nodes
on an Azure cloud from 15
nodes to 40 nodes.
▪ Cloud spend was only $1K a
day.
▪ Established Data Lake;
leveraging Ember to simplify
and expedite data ingest
and model development.
Sharp Healthcare:
▪ Processed and annotated a
10-year span of historical
clinical notes into Cloudera
(Hadoop platform) in a little
over 8 hours.
▪ Coded and indexed the notes
to provide Sharp with rapid
text search on any text,
phrase, term, acronym, and
SNOMED code.
▪ 80% efficiency in their Quality
Payment Plan reporting to
CMS; accelerated
reimbursement process from
weeks to days.
▪ Provided free form
dashboards via Ember.
▪ Created Rapid Response
Model: Predicting Patient
Decline.
IQVIA:
• Re-platformed and migrated
to a new Next Generation
Data Warehouse (NDWx).
• Transformed and drove
efficiencies and quality for
data processing of patient,
claims, RX and DX
transactions.
• Simplified the analysis and
reporting of complex
healthcare data. The new
solution helps to forecast
and analyze the drugs usage
of patients, and helps
determine the accurate drug
prescriptions for the
diagnosis.
• Improved overall weekly
data processing from a
week to under 2 hours.
Be The Match:
▪ Increased collaboration,
internal and external for
genomic analysis
▪ Intuitive user experience for
cohort building and analysis
▪ Scalable compute to
empower fast genomic data
science on a large number of
samples
▪ Single source of truth –
created Genomic Data
Repository for all genomic
data (samples, variants, HLA,
KIR), and annotations
(analysis results, phenotypic
info)
▪ Cost effective storage for
petabytes of data
CENTRALIZED SOURCE OF TRUTH
Consider this…
HEALTHCARE DATA
DataData Data
Clinic
al
Notes
Configure Data
Sources
1
1010
0101
CENTRALIZED SOURCE OF TRUTH
Consider this…
HEALTHCARE DATA
DataData Data
Clinic
al
Notes
Configure Data
Sources
1
Code & Index
Notes at Scale
2
11010101
10100101
1010
0101
CENTRALIZED SOURCE OF TRUTH
Consider this…
HEALTHCARE DATA
DataData Data
Clinic
al
Notes
Code & Index
Notes at Scale
2
Develop &
Deploy Models
3
Configure Data
Sources
1
10100101
11010101
1010
0101
CENTRALIZED SOURCE OF TRUTH
Consider this…
HEALTHCARE DATA
DataData Data
Clinic
al
Notes
Note Search &
Model
Interaction
4
MODEL
SHARING
Configure Data
Sources
1
Code & Index
Notes at Scale
2
10100101
11010101
Develop &
Deploy Models
3
1010
0101
© Cloudera, Inc. All rights reserved. 22© Cloudera, Inc. All rights reserved.
RUSH UNIVERSITY MEDICAL CENTER
Jawad Khan, Director Data Services and Knowledge Management
© Cloudera, Inc. All rights reserved. 23
Rush System
Rush University
Medical Center
676 Beds
Rush Oak Park
237 Beds
Rush Copley
210 Beds
Rush University
Enrollment:
2,569
Rush Health
• Chartered March 2,
1837
• Over 900 providers
• 1100+ inpatient beds
© Cloudera, Inc. All rights reserved. 24
Maestro Custom Marts
Clarity
Caboodle
EDW
Maestro
DM-Views
BO &
Cubes
Tableau
Radar
© Cloudera, Inc. All rights reserved. 25
SilosSelf-service
(Near) Real-time
Good BI framework for
Descriptive reporting
Traditional
Analytics
Advanced
Analytics
© Cloudera, Inc. All rights reserved. 26
Free the data!
External integrated source of
truth
Standardize platform
Big Data
NLP
Data Science / Machine
Learning
© Cloudera, Inc. All rights reserved. 27
© Cloudera, Inc. All rights reserved. 28
© Cloudera, Inc. All rights reserved. 29
Data
Science
Data
Engineering
Domain
Expertise
Code
Developers
Visualization
Advanced
Computing/AI
Statistics
Math
Scientific
Method
© Cloudera, Inc. All rights reserved. 30
Big Data
Engineering
Acquire
Data
Prepare
Data
Computational
Big Data
Science
Analyze
Data
Act on Data
Business
Intelligence
RDBMS
Operational
reporting
© Cloudera, Inc. All rights reserved. 31
© Cloudera, Inc. All rights reserved. 32
© Cloudera, Inc. All rights reserved. 33
Cluster 1: Patients with neurological and psychological diagnoses and a long inpatient
stay
Cluster 2/3: Patients with complicated spinal surgeries and complicated major joint
replacement surgeries, particularly those with underlying medical comorbidities such
as COPD and poorly controlled diabetes
Cluster 4: Patients with complicated neurosurgical procedures
Cluster 5: Medical patients with cardiovascular disease and/or chronic kidney
disease;
Cluster 6: Patients with specific hematologic malignancies;
Cluster 7: NICU patients, particularly neonates of extreme pre-maturity
© Cloudera, Inc. All rights reserved. 34
© Cloudera, Inc. All rights reserved. 35
© Cloudera, Inc. All rights reserved. 36
Data
Acquisition
Machine
Data
Attribute
Data
People
Data
Data
Governance
• Privacy and Lifetime
• Curation and quality
• Sources of Truth
• Interoperability and regulation
• Emphasize foundational
capabilities of EMR
Acquire all data
• Sensor data from machines
• Data from fitness devices
• Data from medical devices
• Social media
• Mobile apps
• External Sets
• Cost Data
• Patient Records
• Self reported novel
people data
• Genomic data
Data (Sources of Truth)
Data Analysis (Aggregates, Reports, NLP, Prediction, Machine Learning, AI)
Data Presentation (Apps first, Real time, Mobile, Self Service)
© Cloudera, Inc. All rights reserved. 37
• Genomic Data
• Clinical Data
(caboodle)
• Patient
Unstructured
Data (NLP)
People
Data
• Provider Directory
• Cost Data
• ERP
• Nomenclatures/
Ontology
• Risk Models
• Scheduling
• Geography
• Environmental
Attribute
Data
• Medical
Device
• Health Device
• Sensor (IOT)
Machine
Date
Machine Learning + AI
Predictive + Prescriptive
Analytics
Timeline – 1 year to 3 years
Genomic + Clinical + Cost +
Unstructured Data = Precision Medicine
& Prescriptive Decisions
Streaming for real-time
capability
Azure + Hadoop/Cloudera
Align with Epic Cognitive Computing Roadmap
Real time data streaming to analytics platform
Rules Engine with Bidirectional Flow of Data to EMR
AI Layer applied to streaming data
API Based App development leveraging FHIR/EMR and Streaming
Data/HDFS
© Cloudera, Inc. All rights reserved. 38
• User community
• Tools
Security
• Hadoop
challenges
• Network
challenges
Cloud
Integration
• Continuous
Integration
• Team
Collaboration
DevOps &
Data
Science
© Cloudera, Inc. All rights reserved. 39
Team resources – Train Vs. Hire
Acceleration – Build Vs. Buy
Cloud Enablement – Agility & Speed Vs. $$
Fail early!
Network & Security team alignment
© Cloudera, Inc. All rights reserved. 40
Questions?
© Cloudera, Inc. All rights reserved.
THANK YOU

Delivering improved patient outcomes through advanced analytics 6.26.18

  • 1.
    © Cloudera, Inc.All rights reserved. DIGGING FOR GOLD: DELIVERING IMPROVED PATIENT OUTCOMES THROUGH ADVANCED ANALYTICS Ryan Swenson, Dir. of Healthcare and Life Sciences Solutions, Cloudera John Spooner, Sr. Analyst Internet of Things, 451 Research Nathan Salmon, Chief Architect, MetiStream Jawad Khan, Dir. Data Services and Knowledge Management, Rush University Medical Center
  • 2.
    © Cloudera, Inc.All rights reserved. 2© Cloudera, Inc. All rights reserved. TODAY’S AGENDA • Industry overview: John Spooner • MetiStream Ember demo: Nathan Salmon • Rush University Medical Center use case: Jawad Khan • Q&A
  • 3.
  • 4.
    © Cloudera, Inc.All rights reserved. Achieving Healthcare Outcomes with AI • John Spooner, Senior Analyst, Internet of Things •June 26, 2018
  • 5.
    © Cloudera, Inc.All rights reserved. Insight Market Insight Your radar into the competitive IoT landscape, with daily analysis of the market delivered to your inbox. Technology & Business Insight In-depth analysis of key technologies and players driving the IoT market: Connectivity, Security, Software, Cloud, etc. Data Customer Data Enterprise and consumer perspective on the early adoption and demonstrable benefit of IoT, captured in quarterly surveys. Market Data Market sizing and forecasting of the myriad device types and radios driving the IoT revolution, built from bottoms-up analysis. M&A A complete database of M&A activity for IoT technology vendors and service providers. TODAY’S SPEAKER John Spooner Senior Analyst, Internet of Things Research Channel INTERNET OF THINGS 3
  • 6.
    © Cloudera, Inc.All rights reserved. 4
  • 7.
    © Cloudera, Inc.All rights reserved. 7 5 Healthcare is a continuum of journeys
  • 8.
    © Cloudera, Inc.All rights reserved. Setting the stage for AI in healthcare • AI/ML promises to be an effectiveness multiplier for clinicians/researchers, speeding diagnosis and treatment for patients and caregivers • Clinical impact: • Quicker decisions, based on more data, increase efficiency/throughput, saving costs and improving revenue. • Supports new remote/proactive monitoring and treatment; complements the EHR. Patient impact: • Quicker diagnosis and treatment for immediate needs. • Emphasis on remote monitoring and long-term care planning. 6
  • 9.
    © Cloudera, Inc.All rights reserved. What is AI? • The terms AI, machine learning and deep learning are used interchangeably and not always correctly. • It’s best to think of them as a nested hierarchy. 7
  • 10.
    © Cloudera, Inc.All rights reserved. What is AI in healthcare? The application of AI in healthcare has three modes: 8
  • 11.
    © Cloudera, Inc.All rights reserved. AI adoption in healthcare? • Market conditions: Providers/clinicians want AI & ML – and are already broadly using it. Researchers are also utilizing AI. Payers: Beginning to support AI and invest in the technology. Vendors: Already moving to productize the three tiers we outlined. 9
  • 12.
    © Cloudera, Inc.All rights reserved. AI adoption in healthcare? • Market opportunity: 451 Research data shows only half of the healthcare companies we have surveyed use AI or are considering it; the rest have no AI strategy! Legislation generally supports AI: • US regulators generally support AI (e.g., the FDA fast tracking AI/algorithm and AMA AI guidelines). • Global attention; e.g., EU declaration on AI. Payers will continue to invest. 10
  • 13.
    © Cloudera, Inc.All rights reserved. Things to consider going forward 11
  • 14.
    © Cloudera, Inc.All rights reserved. 14© Cloudera, Inc. All rights reserved. METISTREAM EMBER Nathan Salmon, Chief Architect
  • 15.
    Founded 2014 ~5K members OpenSource Advocates Spark & Hail Trainers MetiStream Highlights & Key Clients
  • 16.
    Rush Univ MedicalCenter: ▪ Improved Rush’s Adenoma Detection Rate (ADR) KPI by 50% by enhancing their ability to process and analyze unstructured clinical notes data. ▪ Processed and annotated two years of clinical notes data in less than 36 hours. ▪ Scaled up Cloudera nodes on an Azure cloud from 15 nodes to 40 nodes. ▪ Cloud spend was only $1K a day. ▪ Established Data Lake; leveraging Ember to simplify and expedite data ingest and model development. Sharp Healthcare: ▪ Processed and annotated a 10-year span of historical clinical notes into Cloudera (Hadoop platform) in a little over 8 hours. ▪ Coded and indexed the notes to provide Sharp with rapid text search on any text, phrase, term, acronym, and SNOMED code. ▪ 80% efficiency in their Quality Payment Plan reporting to CMS; accelerated reimbursement process from weeks to days. ▪ Provided free form dashboards via Ember. ▪ Created Rapid Response Model: Predicting Patient Decline.
  • 17.
    IQVIA: • Re-platformed andmigrated to a new Next Generation Data Warehouse (NDWx). • Transformed and drove efficiencies and quality for data processing of patient, claims, RX and DX transactions. • Simplified the analysis and reporting of complex healthcare data. The new solution helps to forecast and analyze the drugs usage of patients, and helps determine the accurate drug prescriptions for the diagnosis. • Improved overall weekly data processing from a week to under 2 hours. Be The Match: ▪ Increased collaboration, internal and external for genomic analysis ▪ Intuitive user experience for cohort building and analysis ▪ Scalable compute to empower fast genomic data science on a large number of samples ▪ Single source of truth – created Genomic Data Repository for all genomic data (samples, variants, HLA, KIR), and annotations (analysis results, phenotypic info) ▪ Cost effective storage for petabytes of data
  • 18.
    CENTRALIZED SOURCE OFTRUTH Consider this… HEALTHCARE DATA DataData Data Clinic al Notes Configure Data Sources 1 1010 0101
  • 19.
    CENTRALIZED SOURCE OFTRUTH Consider this… HEALTHCARE DATA DataData Data Clinic al Notes Configure Data Sources 1 Code & Index Notes at Scale 2 11010101 10100101 1010 0101
  • 20.
    CENTRALIZED SOURCE OFTRUTH Consider this… HEALTHCARE DATA DataData Data Clinic al Notes Code & Index Notes at Scale 2 Develop & Deploy Models 3 Configure Data Sources 1 10100101 11010101 1010 0101
  • 21.
    CENTRALIZED SOURCE OFTRUTH Consider this… HEALTHCARE DATA DataData Data Clinic al Notes Note Search & Model Interaction 4 MODEL SHARING Configure Data Sources 1 Code & Index Notes at Scale 2 10100101 11010101 Develop & Deploy Models 3 1010 0101
  • 22.
    © Cloudera, Inc.All rights reserved. 22© Cloudera, Inc. All rights reserved. RUSH UNIVERSITY MEDICAL CENTER Jawad Khan, Director Data Services and Knowledge Management
  • 23.
    © Cloudera, Inc.All rights reserved. 23 Rush System Rush University Medical Center 676 Beds Rush Oak Park 237 Beds Rush Copley 210 Beds Rush University Enrollment: 2,569 Rush Health • Chartered March 2, 1837 • Over 900 providers • 1100+ inpatient beds
  • 24.
    © Cloudera, Inc.All rights reserved. 24 Maestro Custom Marts Clarity Caboodle EDW Maestro DM-Views BO & Cubes Tableau Radar
  • 25.
    © Cloudera, Inc.All rights reserved. 25 SilosSelf-service (Near) Real-time Good BI framework for Descriptive reporting Traditional Analytics Advanced Analytics
  • 26.
    © Cloudera, Inc.All rights reserved. 26 Free the data! External integrated source of truth Standardize platform Big Data NLP Data Science / Machine Learning
  • 27.
    © Cloudera, Inc.All rights reserved. 27
  • 28.
    © Cloudera, Inc.All rights reserved. 28
  • 29.
    © Cloudera, Inc.All rights reserved. 29 Data Science Data Engineering Domain Expertise Code Developers Visualization Advanced Computing/AI Statistics Math Scientific Method
  • 30.
    © Cloudera, Inc.All rights reserved. 30 Big Data Engineering Acquire Data Prepare Data Computational Big Data Science Analyze Data Act on Data Business Intelligence RDBMS Operational reporting
  • 31.
    © Cloudera, Inc.All rights reserved. 31
  • 32.
    © Cloudera, Inc.All rights reserved. 32
  • 33.
    © Cloudera, Inc.All rights reserved. 33 Cluster 1: Patients with neurological and psychological diagnoses and a long inpatient stay Cluster 2/3: Patients with complicated spinal surgeries and complicated major joint replacement surgeries, particularly those with underlying medical comorbidities such as COPD and poorly controlled diabetes Cluster 4: Patients with complicated neurosurgical procedures Cluster 5: Medical patients with cardiovascular disease and/or chronic kidney disease; Cluster 6: Patients with specific hematologic malignancies; Cluster 7: NICU patients, particularly neonates of extreme pre-maturity
  • 34.
    © Cloudera, Inc.All rights reserved. 34
  • 35.
    © Cloudera, Inc.All rights reserved. 35
  • 36.
    © Cloudera, Inc.All rights reserved. 36 Data Acquisition Machine Data Attribute Data People Data Data Governance • Privacy and Lifetime • Curation and quality • Sources of Truth • Interoperability and regulation • Emphasize foundational capabilities of EMR Acquire all data • Sensor data from machines • Data from fitness devices • Data from medical devices • Social media • Mobile apps • External Sets • Cost Data • Patient Records • Self reported novel people data • Genomic data Data (Sources of Truth) Data Analysis (Aggregates, Reports, NLP, Prediction, Machine Learning, AI) Data Presentation (Apps first, Real time, Mobile, Self Service)
  • 37.
    © Cloudera, Inc.All rights reserved. 37 • Genomic Data • Clinical Data (caboodle) • Patient Unstructured Data (NLP) People Data • Provider Directory • Cost Data • ERP • Nomenclatures/ Ontology • Risk Models • Scheduling • Geography • Environmental Attribute Data • Medical Device • Health Device • Sensor (IOT) Machine Date Machine Learning + AI Predictive + Prescriptive Analytics Timeline – 1 year to 3 years Genomic + Clinical + Cost + Unstructured Data = Precision Medicine & Prescriptive Decisions Streaming for real-time capability Azure + Hadoop/Cloudera Align with Epic Cognitive Computing Roadmap Real time data streaming to analytics platform Rules Engine with Bidirectional Flow of Data to EMR AI Layer applied to streaming data API Based App development leveraging FHIR/EMR and Streaming Data/HDFS
  • 38.
    © Cloudera, Inc.All rights reserved. 38 • User community • Tools Security • Hadoop challenges • Network challenges Cloud Integration • Continuous Integration • Team Collaboration DevOps & Data Science
  • 39.
    © Cloudera, Inc.All rights reserved. 39 Team resources – Train Vs. Hire Acceleration – Build Vs. Buy Cloud Enablement – Agility & Speed Vs. $$ Fail early! Network & Security team alignment
  • 40.
    © Cloudera, Inc.All rights reserved. 40 Questions?
  • 41.
    © Cloudera, Inc.All rights reserved. THANK YOU