Activate Your Data
and Healthcare
Ecosystem
Luc Chamberland
WW Business Development Executive
June 18, 2015
Solutions targeting holistic care management reduce
costs and deliver better quality outcomes
Wellness
Wellness
Costs
20% of people
receiving care
consume
80% of
expenditures
Costs increase along the continuum of care
Diagnosis and
Early Intervention
Care Outside the
Hospital
Disease Maintenance
Care for High cost/
High need
Population
Disease Mgmt.
Late Stage/
Co-Morbidity Mgmt.
Care Inside
Hospitals
The Healthcare Industry is dealing with data overload
The average person projected to generate over 1 million gigabytes of health-related data
Determinants of health
Exogenous
60%
Volume, Variety,
Velocity, Veracity
Genomics
30%
Volume
Clinical
10%
1100 Terabytes
Generated per lifetime
6
TB
Per lifetime
0.4 TB
Per lifetime
Variety
Source: "J.M. McGinnis et al., “The Case for More Active Policy Attention to Health Promotion,” Health
Affairs 21, no. 2 (2002):78–93
3
Healthcare has Mountains of Unstructured Content
Biggest blind spot still remains unstructured data
• How are you measuring and
reducing preventative
readmissions?
Does unlocking the unstructured data
help accelerate your transformation?
Physician notes and discharge summaries
• How are you providing
clinicians with targeted
diagnostic assistance?
Patient history, symptoms and non-symptoms
Pathology reports
Tweets, text messages and online forums
• Which patients are
following discharge
instructions?
• How are you using data to
predict intervention
program candidates?
• Would revealing insights
trapped in unstructured
information facilitate more
informed decision making?
Satisfaction surveys
Claims and case management data
Forms based data and comments
Emails and correspondence
Trusted reference journals including portals
Paper based records and documents
Over 80% of stored health
information is
unstructured*
Ecosystem Collaboration
Medical Research Centers
Clinical Research
Cohort Studies
Clinical Trials
Public Health
Life Sciences
Clinical Development
Clinical Trials
Medication Compliance
Pharmacies
Pandemic readiness
Vaccine inventory & distribution
Sanitation & public safety
e-prescribing
New services
Patient Education
Healthy Lifestyles and Diet
Living with Chronic Disease
Health Clubs
Health & Wellness Programs
Hospitals & Physicians
Electronic Medical Records
Health information exchange
Patient ID & eHealth
Patients /
Consumers
Clinical Decision Support
Medical Devices
Evidence-based Decisions
Clinical Analytics
Clinical Protocols
Consumer Relationships
Wellness and Care Services
Clinical Trials
Transaction Services
Retail Clinics
Consumer Services
Government
Healthcare Policy
Medical Research
Regulatory Compliance
Employers
Claims Processing
Banks: Health Savings
Accounts and Payments
Benefit Plan Design
Health & Wellness Programs
Private and Public Insurers
Patient Education
Disease Management
Fraud Prevention
Risk Management
6
6
Every organization is on its own
analytics journey
Advanced, Predictive
Prescriptive
Foundational
• What happened?
• When and where?
• How much?
BI Reporting
•
•
•
•
Dashboards
Clinical data repositories
Departmental data marts
Enterprise data warehouse
• What will happen?
• What will be the
impact?
Population Analytics
• Enterprise analytics
• Unstructured content
analytics
• Outcomes analytics
• Evidence-based medicine
• What are potential
scenarios?
• What is the best
course?
• How can we pre-empt
and mitigate the crisis?
Care Optimization
• Streaming analytics
• Similarity analytics
• Personalized healthcare
• Consumer engagement
• Cognitive Computing
IBM integrated portfolio for Smarter Care
Coordination
Care identification
Care planning
Care delivery
Outcome evaluation
Care pathways
Operational reporting
Analytics and Cognitive Computing
Population analytics
Diagnostic support
Cognitive computing
Foundation
Data warehouse
and data models
“Single view”
customer EMPI
(MDM)
BI, reports and
dashboards
Portals, mobile
and collaboration
Remote monitoring
and medical device
connectivity
Paper and Fax
capture, conversion
and extraction
Comprehensive global consulting, technology, infrastructure and managed services
IBM Care Management
Other Data
Sources
EMR /
EHR
Care Workers
Generate
individualized care
plans
Analysts
Patient 360 View
Enterprise
Services
Claims
Multi-disciplinary
Care Team
Doctor’s notes
Case worker’s
notes
Analyzed
Unstructured Data
Unstructured
data
Social
Workers
Mental Health
Professionals
Comprehensive Care
Plan
Ingest and
Unify Data
Medical
Professionals
Provide Insight at
point of care
Standards Driven Integration Support
Support bidirectional
integration with EMRs and
other source systems
following health care
standards for data exchange
Leverage graphical mapping
tooling. connectors (nodes),
IHE, HL7, and Continua
schemas and development
pattern for easy integration
Unify and synchronize
fragmented clinical, social
and behavioral health
information to create Mary’s
personalized care plan
Care Workers Multi-disciplinary Social
Workers
Care Team
Mental Health
Professionals
Analysts
IBM Care Management
Medical
Professionals
Patient Centered, Team Based Care
Visualize biopsychosocial profile of the
client in 360 degree page
Receive referrals for leveraging
configurable workflow and automatically
create an outcome plan
Use intuitive and flexible outcome
planning interface to compose
comprehensive care plans for Mary
Collaborate across diverse
stakeholders efficiently coordinating
care, locating and referring care
providers and optimizing resources
360 Degree View - Visualization of
biopsychosocial profile
Electronic Medical Records adoption reaches $22.3B
adoption by 2015
HOWEVER
EMRs
Integrated Care
Single EMR
Multi-EMR integration
Structured
Provider-centric
Dynamic, ad-hoc
Patient-centric
Predefined terms
Non-standard terms
System of record
Support for future goals
Uniform care
Care Pathways
Personalized plans
Group decision making
Proactive Delivery
EMR records do not support the aspirations or the workflow of integrated care,
but are a complementary enabler for integrated care solutions
Integrated Care
+ Multiple provider
+ Patient engagement
+ Personalized care plans
+ Workflow & Collaboration
Electronic Medical
Records (EMR)
Provider-centric
Billing oriented
Mature reporting
Sources: EMR Adoption statistics Accenture 2014
1 Rudin, Bates 2014, 2 Bates 2010, 3 O’Malley et al 2010, 4 Graetz 2009, 5 Rudin 2014, 6 PWC 7 Cipriano et al
Improved outcomes
Catalan Institute of Health, Catalonia, Spain, collaborates across
clinicians and social care teams to cut costs and improve outcomes
Business problem:
Rising chronic disease in an aging population are consuming more
healthcare resources
Solution:
Targeted program for elderly aimed at improving adherence in
care programs, enhancing patient quality of life and satisfaction
with the healthcare system, and controlling costs
Outcome
Clinicians and social workers
coordinate care planning and delivery
with a comprehensive view of the
individual
Knowledge
Sharing of best practices and
holistic view of the patient enables
individualized care plans that
engage clinical and social providers
Collaboration
Unified view of care plan across
stakeholders increases effectiveness
and informs adjustments
Coordination
Resources responsible for referral
management and in home care
delivery can collaboratively and
quickly support incoming requests
Big Data & Analytics
South Florida Behavioral Health Network provides individual-centric
treatment through coordinated care management and analytics
30 – 50% decrease
anticipated in the probability of
re-arrest when integrated
behavioral health treatment starts
within 90 days
Automated alerts
support provider accountability
and crisis prevention
Provides insight
into treatment-and-cost
effectiveness and near-real time
visibility of provider activity
Solution components
•
•
•
•
•
•
•
•
Cúram Social Program Management
IBM® Cognos® Business Intelligence V10
IBM DB2® Advanced Enterprise Server Edition
IBM InfoSphere® Warehouse Enterprise Edition
IBM WebSphere® Application Server
IBM SPSS® Modeler
IBM Global Business Services® – Application
Innovation Services
IBM Alliance Partner Otsuka Pharmaceutical
Co. Ltd.
Business challenge: People with mental illness who rely on publicly funded
medical care are among the most vulnerable, often ending up incarcerated
instead of receiving needed treatment. Even within this mental healthcare
provider network, without a systematic view, treatment and follow-up care could
still be disjointed, leading to preventable crises and incarceration.
The smarter solution: The network is combining coordinated care management
and healthcare analytics to help deliver more consistent, harmonized patient care.
Analytics personalize follow-up referrals by matching the individual’s unique
needs to provider characteristics such as specialty, treatment options, location
and languages spoken. Automated alerts and provider accountability help prevent
individuals from falling through the cracks and ending up in crisis.
[W]e look proactively for creative solutions to coordinate care for our patients…
[such as this] innovative approach to improving efficiencies within our…system.
Applying Natural Language Processing
•
•
•
•
•
•
Accurately identify and extract facts from text including negation
“55%” = LVEF
“Patient does not show signs” = Negative Symptom
Accurately interpret and assign values to ambiguous statements
“around 55%” = LVEF
“Shows slightly elevated levels” = if condition A = 10%, if condition B = 20%
Infer meaning from non-contextual content
“Cut back from two packs to one per day” = Smoker
Find inconsistencies between data sets
Cleanse, enhance and normalize raw data
“Myocardia infarction” and “heart attack” = equal same thing
Correct misspellings and abbreviations through NLP
Enhance or augment by assigning correct RxNorm, SNOMED, ICD or other codes /
terminology.
“Broken femur” (diagnosis) -> 821.00 (ICD9)
Preserve and structure facts and concepts from contextual content.
–
Augment structured data in clinical systems (EMRs)
A 42-year old white male
presents for a physical. He
recently had a right
hemicolectomy invasive
grade 2 (of 4) adenocarcinoma
in the ilocecal valve was found
and excised. At the same time
he had an appendectomy.
The appendix showed no
diagnostic abnormality.
15
Patient
Procedure
Procedure
Age: 42
Gender: Male
Race: White
hemicolectomy
diagnosis: invasive
adenocarcinoma
anatomical site: ileocecal valve
grade: 2 (of 4)
appendectomy
diagnosis: normal
anatomical site: appendix
#ibmiod
Care Management delivers out of the box value for content analytics
Problems
– Result of a series of interim annotations that
identify diseases, symptoms, and disorders
– Normalize to standard terms and standard
coding systems including SNOMED CT, ICD9,
ICD10, HCC, CCS
– Capture timeframes of the problem
– determine if past or current problem
– Determine confidence (Positive, Negative,
Rule Out)
Procedures
– Identify compound procedures
– Normalize to standard terms and standard
coding systems including SNOMED CT, CCS,
CPT
– Capture timeframes of the procedure
Medications
– Series of interim annotations that identify
drugs, administrations, measurements
– Normalize to standard terms RxNorm
Cancer Diagnosis
– Attributes: Name, Date, Modality, Grade
(Scale, Value), Behavior, Site, Measurement
•16
Allergies
Drug allergies, generic allergies e.g. food
Demographic and Social
• Patient Age
• Living Arrangement
• Employment status
• Smoking status
• Alcohol use
Compliance & Noncompliance
• Patient's history of medication
compliance with directions such as "take
all doses, even if you feel better earlier“
• Noncompliance - Patient's history of
medication noncompliance with
directions.
Labs results
Type of lab test performed, unit of measure,
result value
Ejection Fraction – in support of CHF use cases
100+ dictionaries, 800+ parsing rules
Care Management Analytics Use Cases
Regulatory Measures
Clinical and Research
• Quality measure gaps
– Meaningful use gaps
– PQRS
– HEDIS
• Risk-adjusted scoring
•
•
•
•
•
Payers
EMR enrichment
Post-discharge follow-up
Screen research subjects
Identify risk factors
Detect adverse events
Providers
Life Sciences
17
UNC Healthcare
Improve reporting and post-discharge communication of adverse events
Business problem:
Some of the data required to calculate PQRS measures are locked in clinical notes.
Also, the need to reduce hospital readmissions is a major challenge and expensive
for most healthcare providers in terms of financial penalties and unreimbursed care.
Improve patient health with better follow-up of post-discharge instructions and further
tests and treatment plans after the leaving hospital.
10%+
Quality improvement of PQRS
measures
Proactive care communication
Ensure relevant, accurate and timely
communication across transitions of
care by automatically generating
reminders and alerts to inform the
care team
Reduce readmission
by extracting predictors of risk from
clinical notes
Discharge instructions consist of many pages of free-text notes and can be difficult
for patients and care managers to decipher creating the potential to miss valuable
information such as medications, diagnosis and follow on appointments.
Solution:
Care Management leverages unstructured data from discharge instructions, in the
form of reminders and alerts, to better enable post-discharge healthcare providers
and empower those responsible for patient centered care.
Hospital staff can now use the solution to analyze unstructured text for key discharge
terminology using natural language processing to determine the context of the
content, extracts any relevant data from discharge summaries, doctors’ notes, UKG
reports and other unstructured discharge related content, and converts it into
structured data. This structured data is then used to generate alerts and reports for
patients’ primary care doctors and other caregivers. Clearer data and better
communication between health professionals helps ensure that patients keep their
follow-up appointments and complete their post-discharge treatment. Not only can
patients stay healthier, but the hospital can also save millions of dollars on costly
hospital readmissions.
https://www.youtube.com/watch?v=LQTXQsAnq7s
Risk-Adjust CMS Payments by Finding Comorbidities to Influence
HCC Scores
19
Payer leverages IBM Care Management, helping them achieve
quality measures and save more than $2.5M annually
Business Challenge
–
–
–
–
Three million members across 21 states.
Payer’s mission: improve the health of the community through health insurance solutions for
the under-insured and uninsured.
Challenge: quality of the information provided by its existing HEDIS (Healthcare
Effectiveness Data and Information Set) reporting system. There were significant gaps,
which resulted in limited insight around the effectiveness of the care plan.
The key contributor to this issue was data trapped in clinical and physician notes that were
unstructured. Payer did have a concession plan that involved a third party manually
reviewing each member’s chart, but this was not only time consuming and costly, it was not
delivering effective results.
IBM Solution
–
–
–
Parse unstructured clinical notes to improve the quality of medical records
Reduce annual labor costs by $2.5M USD by eliminating the need for manual analysis of
charts by a third party company
Improve distribution and response time to enterprise-wide HEDIS rate calculations
20
Care Management analytics and Epic integration
What’s new?
• Care providers are adopting electronic medical records but traditional
doctors’ notes still play an important role in tracking & managing patients
• Q1 2014: Integration testing with the Epic EMR 2014 release & IBM
Advanced Care Insights for Natural Language Processing (NLP) has been
successfully completed, solidifying leadership of both companies in their
respective markets
What value does this provide?
• Traditionally a manual process, IBM’s software can analyze doctors’ notes
& transform them into a format that can be readily uploaded into the
patient record, including automatically adding industry standard diagnosis
& treatment codes
• Allows doctors to accurately capture information from unstructured text in
real-time, to improve patient outcomes & simplify administrative processes
What does this mean?
• Empowers 297 Health Systems that have adopted Epic to capture
actionable insight from IBM’s NLP capabilities – the same technology
utilized in the revolutionary Watson cognitive system
Clinician Use Case of Epic-NLP Integration
Step 1: Clinician Enters New Encounter Note in Text Field
Clinician Use Case of Epic-NLP Integration
Step 2: View additional medical problems that are recognized via NLP
Clinician Use Case of Epic-NLP Integration
Result: appropriate condition codes are generated
Case Study: Readmission predictors at Seton
The Data We Thought Would Be Useful … Wasn’t
•
113 candidate predictors from structured and unstructured data sources
•
Structured data was less reliable then unstructured data – increased the reliance on unstructured data
New Insights Uncovered by Combining Content and Predictive Analytics
•
LVEF and Smoking are significant indicators of CHF but not readmissions
•
Assisted Living and Drug and Alcohol Abuse emerged as key predictors (only found in unstructured data)
•
Many predictors are found in “History” notations and observations
Top 18 Indicators
1. Jugular Venous Distention Indicator
2. Paid by Medicaid Indicator
3. Immunity Disorder Disease Indicator
4. Cardiac Rehab Admit Diagnosis with CHF Indicator
5. Lack of Emotion Support Indicator
6. Self COPD Moderate Limit Health History Indicator
7. With Genitourinary System and Endocrine Disorders
8. Heart Failure History
9. High BNP Indicator
10. Low Hemoglobin Indicator
11. Low Sodium Level Indicator
12. Assisted Living (from ICA Extract)
13. High Cholesterol History
14. Presence of Blood Diseases in Diagnosis History
15. High Blood Pressure Health History
16. Self Alcohol / Drug Use Indicator (Cerner + ICA)
17. Heart Attack History
25
18. Heart Disease History
Predictor Analysis
% Encounters
Structured Data
% Encounters
Unstructured
Data
Ejection Fraction (LVEF)
2%
74%
Smoking Indicator
35%
(65% Accurate)
81%
(95% Accurate)
Living Arrangements
<1%
73%
(100% Accurate)
Drug and Alcohol
Abuse
16%
81%
Assisted Living
0%
13%
Model Accuracy, precision and recall
Retrieved by model
Recall
Precision
Fraction of relevant
Instances retrieved
Fraction of retrieved
Instances that are
relevant
False
negatives
False
positives
Relevent
Not relevent
Similarity analytics supports data-driven decisions based on
comparisons to a meaningful cohort
Physicians have limited time and resources
to focus on complex care dilemmas, yet many
patients have multiple conditions
Clinical trials and health research typically
focus on single diseases
Treatment guidelines are usually developed
with “standardized” reference data
Care delivery tends to be ad hoc in nature;
care guidelines are not followed 40 percent of
the time
83 percent of Medicaid
patients have at least
one chronic condition
(almost 25 percent have at
least five comorbidities)2
1
83%
Why not augment caredelivery guidelines with
population-specific
insights—including those
derived from unstructured
data—to enhance
decision making?
5
Medicare patients with
5 or more chronic
conditions accounted
for 76 percent of all
Medicare expenditures3
76%
1 RAND Health, Projection of Chronic Illness Prevalence and Cost Inflation, October 2000.
2 Health Affairs, The rise in spending among Medicare beneficiaries: the role of chronic disease prevalence and changes in treatment intensity, K.E. Thorpe and D.H. Howard,
August 22, 2006.
How Similarity Analytics Work, Part 1
For this patient …
• Analyze longitudinal
data to develop
profile across
30,000+ possible
points of comparison
• Determine the
individual risk factors
for this patient based
on the desired
outcome
28
• Create an outcomes
based personalized
How Similarity Analytics Work, Part 2
Based on this personalized profile …
• Find the most similar patients (or dynamic cohort) from entire population
• Analyze the attributes and outcomes for this cohort (across 30,000+ dimensions)
• Predict the probability of the desired outcome for patient in question
• Suggest a personalized care plan based on the unique needs of this patient
Historical Observation Window
This Patient’s Longitudinal Data
Prediction Window
Predicted Outcome For This Patient
Dynamic Cohort Longitudinal Data with Outcomes
29
Desired
Outcomes
Treatment Efficacy
Identifies the outcomes of drug treatments prescribed to groups of similar patients
30
Thank you
Luc Chamberland
[email protected]
31