59
Encontro e-Saúde
Big Data & Business Analytics
em Saúde UNIMED-PR
PUC-PR
Umberto Tachinardi, MD, MS, FACMI Associate-Dean for Biomedical Informatics – SMPH – UW-Madison
Chief Knowledge Officer – UW-Health
Consortium on Technology for Proactive Care Northeastern University
Healthcare is Changing
EPISODIC, REACTIVE
FOCUS ON DISEASE
PROACTIVE and PREVENTIVE
FOCUS ON WELLBEING
QUALITY OF LIFE
HOSPITAL-CENTRIC PATIENT-CENTRIC, HOME-BASED
FRAGMENTED, LOCAL DATA
MORE EVIDENCE – BASED
DECISION SUPPORT
INTEROPERABLE, EHR AVAILBLE
ANYWHERE, ANYTIME
EMPOWERED, ENAGAGED,
INFORMED, PARTICIPATING PASSIVE PATIENTS
TRAINING & EXPERIENCE
BASED
Framework for Analytics Capability
Development: the Eight Levels of Analytics
Michael Walker; http://www.rosebt.com/1/post/2012/09/eight-levels-of-analytics-for-competitive-advantage.html
I’m aware
I understand
I know how to
create great
outcomes
Know
the
Past
Shape
the
Future
5
Analytics-Driven Healthcare Enterprise
Optimization
Predictive Analytics
BI Reporting and
Ad Hoc Analysis
• What happened?
• When and where?
• How much?
• What is the best choice?
• What will happen?
• What will the impact be?
Predictive
analytics
• Personalized healthcare
• Dynamic fraud detection
• Patient, member behavior
Decision support
analytics
• Enterprise analytics
• Evidence-based medicine
• Clinical outcomes
analytics
Data integration
Data warehouse
• Dashboards
• Clinical data repositories
• Departmental data marts
Transaction
reporting
• Basic reporting
• Spreadsheets
Current analytics level
© 2011 IBM Corporation
http://tamaradull.com
Hospital Infection
PCOR
Readmission
Precision/Personalized Medicine
The new EHR ecosystem
Transactional database
Pre-Analytical database
Data Warehouse
IT (CIO)
Data Center Networks Support
Training Security Applications
Analytics (CRIO/CKO)
Data Warehouse Reporting Semantic Mgt
Registries/Panels/Datasets NLP
IT/Informatics Governance
The abridged/biased map of the domain(s)
Proprietary & Confidential
1
10
Weber GM, Mandl KD, Kohane IS. Finding the Missing Link for Big Biomedical Data. JAMA. Published online May 22, 2014
Data in Healthcare
• Types and sources
• Structured
• Coded (e.g. ICD, SNOMED, CPT)
• Numerical values (e.g. temperature, BP, HR, age)
• Unstructured
• Text (e.g. discharge summary, email)
• Image (e.g. radiological, microscopy, photos)
• Audio (e.g. dictation, heart murmur)
Big Data in Healthcare
• Definitions
• The “3Vs”: "Big data is high volume, high velocity,
and/or high variety information assets that require
new forms of processing to enable enhanced
decision making, insight discovery and process
optimization.” (Gartner)
Clinical/Populational Data Challenges:
Granularity
Genetics, Genome Science, Systems Biology Phenomics/Populomics
DNA mRNA Protein Molecular Interaction Network
Biochemistry Physiology Clinical
Data Disease
Phenotype Populational
Data
Publications Clinical Guidelines
Biological Models
Knowledge
Information
Data
Acquisition/Collection
Organization
Storage
Curation
Security
Interfacing
Aggregation
Linkage
Semantic
Mapping
Presentation
Security
Pattern recognition
Artificial Intelligence
Forecasting
Scenario Analysis
Meaningfulness
Comparative
Effectiveness
Intellectual Property
Big Data solutions accelerate healthcare transformation
Evidence Based Medicine Evidence based Healthcare Models driven by “Health outcomes”. Analyzing Care experience requires inspecting structured and
unstructured data
Health Outcomes Provider & Staff shortage demands workforce
productivity & Efficiency improvements Knowing patients 360 implies looking at all their data to provide
optimal care
Patient Centered Care Knowing patient’s lifestyle and habits helps drive optimal outcomes
and is part of providing comprehensive care pre and post visit.
Disease Management
Conducting disease management and surveillance requires exhaustive
processing of structured and unstructured data to identify chronic and re-
emerging infectious diseases
© 2011 IBM Corporation
HIMC DWs Current Architecture Configuration
ET
L
Clarity
HIMC
Stage
EDW
Unity UW
Netezza
Security Groups and Views
HIMC
Land A2B
Other
Source
Systems
Users and reporting tools
(QlikView, SAS, BOE)
Informatica
I2B2 Research
Users
BMI – HIMC Innovative Technologies
Infrastructure Development
EHR
Other Clinical
Sources
D a t a Q u e r y
M a n u a l
BMI/HIMC Research
Users
i2b2
REDitor
Ontology
Clinical
Users
OnCore/RED
Cap
Other
Research
Sources
External to Clarity Internal to Clarity
Clinically
Administered
Medications
Charging
Payment
Collection
Coverage/
Authorization
Treatment /
ProcedureTests
Costing
Pre Access/
RegistrationVisits/
Appointments
Plan Care/
Orders
Care Giver
Performance
Disease
Management
Patient
Education
Retail
Pharmacy
Service Line
Utilization
Community
EducationPopulation HealthPatient Outcomes
Le
vel 1
Ro
llup
Le
vel 0
(A
tom
ic L
eve
l)
Financial
Health
Measures
Clinical
Effectiveness
ETL and Data Integration
External Data Master Data ADT TransactionsAppointments /
Visits
Tests/
PharmacyAccounts
Diagnosis
Patient
Tracking
Clinical Trials/
Research
Clinical, Quality
and Safety
Measurements
Quality and Safety
Improvement
In-Basket
Communication
Outcomes
Measurement
Improving Patient
ExperienceMarket Position
Panel
Management
Case
Management
Clinical Patient CareRevenue CycleRevenue Cycle
Registries/Data
Silos
Research
Collaborations
Safety
Events
Repetitive
Services
Patient
Resource
Needs
Patient
Survey
Patient
History
Patient
ConsultPCP &
Referrals
DME
`
Vitals
Immunization
Complaints
Patient Population Definitions
Resource
Utilization
Problem /
Medication
Best Practice
Advisories
Follow-up
Payroll
Budget
Risk
AssessmentsPeer Review
Operational
Efficiencies
Accounts
Activity Definitions
EDW
Q u e r y T o o l
A u t o m a t e d
Honest
Broker
Natural
Language
Processing
20
Current ETL Process for EDW & i2b2
Health Link
Chronicles
(Caché)
Health Link Cogito
Clarity
(Oracle)
UW/Unity DW
Staging
(Netezza)
i2b2 Landing
Zone
(SQL Server)
i2b2 CRC
“Clinical Research
Chart”
(SQL Server)
i2b2 METADATA
“Ontology Management
Cell”
(SQL Server)
i2b2 HIVE
(SQL Server)
i2b2 PM
“Project Management
Cell”
(SQL Server)
I2b2 WORKDATA
“Workplace Framework
Cell”
(SQL Server)
Informatica Epic ETL
Kettle
I2b2 Staging
Snapshot
(Netezza)
Informatica
Unified Medical Language System
ICD SNOMED LOINC RxNorm
Diagnosis
Environment
Medication
Family history
Symptom
Billing Codes
Dispensed medications
Laboratory measurements
External to Clarity Internal to Clarity
Clinically
Administered
Medications
Charging
Payment
Collection
Coverage/
Authorization
Treatment /
ProcedureTests
Costing
Pre Access/
RegistrationVisits/
Appointments
Plan Care/
Orders
Care Giver
Performance
Disease
Management
Patient
Education
Retail
Pharmacy
Service Line
Utilization
Community
EducationPopulation HealthPatient Outcomes
Le
vel 1
Ro
llup
Le
vel 0
(A
tom
ic L
eve
l)
Financial
Health
Measures
Clinical
Effectiveness
ETL and Data Integration
External Data Master Data ADT TransactionsAppointments /
Visits
Tests/
PharmacyAccounts
Diagnosis
Patient
Tracking
Clinical Trials/
Research
Clinical, Quality
and Safety
Measurements
Quality and Safety
Improvement
In-Basket
Communication
Outcomes
Measurement
Improving Patient
ExperienceMarket Position
Panel
Management
Case
Management
Clinical Patient CareRevenue CycleRevenue Cycle
Registries/Data
Silos
Research
Collaborations
Safety
Events
Repetitive
Services
Patient
Resource
Needs
Patient
Survey
Patient
History
Patient
ConsultPCP &
Referrals
DME
`
Vitals
Immunization
Complaints
Patient Population Definitions
Resource
Utilization
Problem /
Medication
Best Practice
Advisories
Follow-up
Payroll
Budget
Risk
AssessmentsPeer Review
Operational
Efficiencies
Accounts
Activity Definitions
EDW
NLP using MedLEE pipeline
22
Research data
needs i2b2
DATA
Human Operated
Data Querying
Automatic Querying
Data queries – A critical resource
EDW
EHR NLP
Hon
est
B
roker
Redesigning the healthcare
model
The learning healthcare system (IOM)
Evidence Based Medicine
Meaningful use (Health Information Technology for Economic and Clinical Health - HITECH), define clinical quality measures and added electronic privacy (HIPAA extensions)
Comparative Effectiveness and Accountable Care Organizations (Patient Protection and Affordable care Act)
Launching PCORnet, the National Patient-
Centered Clinical Research Network PCORnet Kickoff Meeting
January 21st, 2014
25
Useful Links
http://www.ahrq.gov http://www.pcori.org https://www.healthcatalyst.com/
PCORnet: the National Patient-Centered Clinical
Research Network
The goal of PCORI’s National Patient-Centered Clinical Research Network Program is to improve the nation’s capacity to conduct CER efficiently, by creating a large, highly representative, national patient-centered clinical research network for conducting clinical outcomes
research.
The vision is to support a learning US healthcare system, which would allow for large-scale research to be conducted with enhanced accuracy and efficiency.
27
CDRN Highlights
28
• Networks of academic medical centers, hospitals and physician
practices
• Networks of non-profit integrated health systems
• Networks of low income clinics
• Networks leveraging AHRQ investments and NIH investments
(CTSAs)
• Inclusion of Health Information Exchanges
• Wide geographical spread
• Inclusion of underserved populations
• Range from 1M covered lives to 28M
29
• Variety of stakeholders in participating organizations and in
leadership team: patients, advocacy groups, physician organizations,
academic centers, PBRNs etc.
• Strong understanding of patient engagement
• Significant range of conditions and diseases
• Variety in populations represented (including pediatrics,
underserved populations etc.)
• 50% rare diseases
• Significant range in the maturity of the group in terms of data
available
• Several have capacity to work with biospecimens
PPRN Highlights
Multiple Networks Sharing Infrastructure
Each organization can participate in multiple networks
Each network controls its governance and coordination
Networks share infrastructure, data curation, analytics, lessons, security, software development
Health
Plan 2
Health
Plan 1
Health
Plan 5
Health
Plan 4
Health
Plan 7 Hospital 1
Health
Plan 3
Health
Plan 6
Health
Plan 8
Hospital 3 Health
Plan 9
Hospital 2
Hospital 4
Hospital 6
Hospital 5
Outpatient
clinic 1
Outpatient
clinic 3
Patient
network 1
Patient
network 3
Patient
network 2
Outpatient
clinic 2
31
Our Clinical Research System is Not Generating
the Evidence we Need to Support Practice!
High % of decisions not supported by evidence
Poor health status of US population
Great disparities
Questions about reliability of the system growing
Current clinical research system is great except:
Too slow
Too expensive
Unreliable
Doesn’t answer the questions important to patients
Unattractive to providers and administrators in the system
39
Another view . . . Toward Personalized Medicine
From Personalized Healthcare 2010, IBM Executive Brief, 2003.
40
• No “one size fits all” drug
• Most drugs work for 30%
to 70% of patients
• Multiple factors determine
drug responses
• Phamacogenetics is
essential for individualized
therapy
Drug Response
Genetic factors
Environmental factors
Physiological factors Demographic factors
Drug-drug interactions
Non- Responders
Responders Toxic
responders
The Right Therapy and the Right Dose for the Right Patient at the Right Time
Variation in Drug Response
Baby Tariq Jamieson Toronto, 2005
Morphine found in baby: 70 ng/mL
Mom took Tylenol 3
Tylenol 3 supposed to be safe for breast feeding
Tylenol 3 contains codeine
Mom is a ultra-rapid metabolizer
Codeine and breast feeding
* Safe concentration in baby: 10 to 12 ng/mL
Mom Baby
Normal metabolizer
Usually 0-0.2 ng/mL
Ultra-rapid metabolizer
Tariq Jamieson: 70 ng/mL
FDA Warning
1 in 3 Patients in 2011 would Benefit from a Pharmacogenomic Test
0
50000
100000
150000
200000
250000
300000
350000
400000
450000
2005 2006 2007 2008 2009 2010 2011
Nu
mb
er
of
Pat
ien
ts
Year
Number of drugs >=1
Number of drugs >=2
Number of drugs >=3
Total Num of Patients
Economic Impacts
• Cost of Preventable adverse Events
– Warfarin-related bleeding: $11,542 *
– Breast cancer recurrence: $40,461*
– Myocardial infarction or stroke: $30,000
* Numbers from Schildcrout et al (2012) Clinical Pharmacology & Therapeutics
Estimated Economic Impact: $11.3M
Medication Adverse event
Num of Patients at Marshfield Clinc (2011)
Preventable Events by Pharmacogenomics Economic Impact
Warfarin Bleeding 16034 605 $ 6,984,088
Clopidogrel
Myocardial infarction or stroke 8967 111 $ 3,330,000
Tamoxifen Breast cancer recurrence 883 25 $ 992,418
Total 25884 741 $ 11,306,507
There is much promise in the process . . .
50
Source: Genzyme Genetics, as presented in Allison, Malorye,
“Is Personalized Medicine Finally Arriving?”, Nature Biotechnology,
Vo.l 26, No. 5, May 2008, p 517.
htt
ps:
//w
ww
.hea
lth
cata
lyst
.co
m
htt
ps:
//w
ww
.hea
lth
cata
lyst
.co
m
Not good!
https://www.healthcatalyst.com
Good!
https://www.healthcatalyst.com
https://www.healthcatalyst.com
https://www.healthcatalyst.com
https://www.healthcatalyst.com
