Improving Health Outcomes
Using Big Data
Approach
and
Methodology
to
Achieve
Success
Ahmed Ghouri, M.D.
Founder and CEO,
Anvita Health
March 2013
“Big Data” is Not a Destination
New Understanding and New Interpretations are the Purpose of
Big Data.
Insights that Avoid:
•
Misdiagnoses and delayed diagnoses, leading to complications
•
Use of invasive measures before conservative measures
•
Unsafe prescribing leading to adverse drug events
•
Lack of preventative services, leading to disease formation, worsening, and
complications
•
Use of expensive tertiary and emergency services vs. affordable PCP and home-
based care
The Holy Grail of Medicine
•
What is the Best Treatment for a patient - based on thousands to millions
of observations of Like-Patients
Clinical decision support, MU
MRA
analysis
Quality reporting Consumer facing CDSReal
Real
‐
‐
time
time
and
and
Batch
Batch
Condition management ADE and Adherence opportunities Humana Large BCBS plan Top 2 national claims
clearinghouse Top 3 PBM
Top 5 Disease management
company
Humana
Harvard teaching hospital
Humana Inc.
Large BCBS plan
Top 5 Disease management
company
Available for PHRs
Cost savings Opportunities, ePA
Top 2 lab company
(laboratory benefits
management)
Humana
Top 3 BCBS plan Nationwide EMR, eRx Top 3 HIE
Harvard teaching hospital
Anvita: One Engine, Many Capabilities
Where Insights Need to Be Distributed
300+ EMRs (tablets and PCs)
HIEs dashboards
Consumer Portals
Eligibility Portals
Nursing Applications
Revenue Cycle Management Apps
Handheld Devices…
Single Patient Real Time CDS
Arbitrary number of clinical questions
Population Analyses
Arbitrary number of outcomes analyses
What to Do Right Now
Real-time, continuous guidance
Service
Oriented
Architecture
Any HIT Application
Anvita Delivers Interpretations via an
Infrastructure-Independent Platform
Data Volume Overload
(must use next-generation
query tools)
Data Type Overload
(must know how to reconcile
different dictionaries)
Keys to Interpretation of Big Data
Clinical Data
Points in a Single Patient
EMR & Claims: 10,000+
transactions per member
Gene Sequencing:
3 Billion data points
per member
Continuous
Bio-sensors
(1 meg / min)
Sophisticated
Semantic Search
(Symbolic
Reasoning) will be
essential due to
combinatorics:
(1) data explosion
(2) knowledge
explosion and
(3) terminological
explosion
2013
2015
2017
3
5
4
Billions
Data Volume Overload
Data Type Overload
Tower of Babel – More data becomes less usable
Pathways
Gene Directory
Proteins
Small Molecules
Sequences
Sequences
Inheritance
To
Compute
Meaning Transforms
Big Data Requires Transformation
NOT Just Storage and Real-time Retrieval
Data Transformation
Changes Everything
“Which patients have rapidly
worsening kidney function that is
correctable?”
POC
Self-Reported Data
HRA
Biometrics, Labs, PBMs, Claims
Fill in Data Gaps
Crosswalk
Lab Value
Standardization
Signal to Noise
Data Veracity
Examples of Important Data Transformations
Rosetta Stone operations, Information Gap Resolution, and
Signal-to-Noise Analysis
A TRUE, COMPUTABLE PATIENT VIEW CPT [2D Echocardiogram] + RxClass [Diuretic] + CPT [Lab: B-Natriuretic Peptide] + (ICD-9 [Edema or Shortness of Breath])INFERS
ICD 428.0 Congestive Heart Failure
Patient is Allergic to Cephalosporin FDB 477-1 Medispan 027 Multum 002 SnoMed 51779009 RxNorm 2235 HRA 4242
∑
Simultaneous Variability in Multiple Dimensions
has made
the problem almost intractable
Quantity Type Variability
(pick 1 of 3)
•
Moles (number of molecules) – specific to analyte
•
Charge (e.g., milliequivalents) – specific to analyte
•
Mass (e.g., grams)
Mathematical Factor Variability
(pick 1 of 20)
•
Milli, Micro, Femto, Pico, etc – sender choice
•
Ratio of quantities (numerator and denominator choices
arbitrary to lab system)
Representative LOINC Code
(pick 1 of 10)
•
Arbitrary Choice of code to transmit/store
Lab Value Standardization in Real-time
17861-6 9 milligrams per deciliter 2000-8 4.5 MEQ/L
17861-6 90 milligrams per liter 2000-8 4.5 mceq/mL
17861-6 90 micrograms per milliliter 2000-8 4.5 ueq/mL
17861-6 9 mg/dL 35246-8 9 milligrams per dL
17861-6 90 mg/L 35246-8 90 milligrams per liter
17861-6 90 mcg/mL 35246-8 90 micrograms per mL
17861-6 90 ug/mL 35246-8 9 mg/dL
1996-8 2.25 millimoles per liter 35246-8 90 milligrams per liter
1996-8 2.25 micromoles per milliliter 35246-8 90 mcg/mL
1996-8 2.25 mmol/L 35246-8 90 ug/mL
1996-8 2.25 mM 35246-8 2.25 millimoles per liter
1996-8 2.25 mcmol/mL 35246-8 2.25 micromoles per milliliter
1996-8 2.25 umol/mL 35246-8 2.25 mmol/L
1996-8 4.5 milliequivalents per liter 35246-8 2.25 mM
1996-8 4.5 microequivalents per milliliter 35246-8 2.25 mcmol/mL
1996-8 4.5 meq/L 35246-8 2.25 umol/mL
1996-8 4.5 mEq/L 35246-8 4.5 milliequivalents per liter
1996-8 4.5 MEQ/L 35246-8 4.5 microequivalents per milliliter
1996-8 4.5 mceq/mL 35246-8 4.5 meq/L
1996-8 4.5 ueq/mL 35246-8 4.5 mEq/L
2000-8 2.25 millimoles per liter 35246-8 4.5 MEQ/L
2000-8 2.25 micromoles per milliliter 35246-8 4.5 mceq/mL
2000-8 2.25 mmol/L 35246-8 4.5 ueq/mL
2000-8 2.25 mM 49765-1 9 milligrams per deciliter
2000-8 2.25 mcmol/mL 49765-1 90 milligrams per liter
2000-8 2.25 umol/mL 49765-1 90 micrograms per milliliter
2000-8 4.5 milliequivalents per liter 49765-1 9 mg/dL
2000-8 4.5 microequivalents per milliliter 49765-1 90 mg/L
Business Intelligence
will Not work on this Data
Business Intelligence
will Not work on this Data
Single Lab Result – Identical Variants
Serum Calcium = 9 mg/dL
Variable Copay of Medications Using EMR Data
-Allows for personalized increase/decrease of copay based on efficacy and safety of a medication.
Increases the use of drugs of choice for conditions (e.g., chemotherapy with tumor response genes
and receptors, antibiotics with highest efficacy for specific infections, etc) and decreases use of
dangerous drugs (e.g., Beer’s criteria in elderly) by financial versus educational means alone. Key
value is evidence-based adherence to best practices resulting in lower total medical costs and
superior outcomes. Secondary value is market differentiation and leadership by innovation.
Comparative Efficacy of Medications Using EMR Data –
POC –
Identifying, ranking, and scoring of medication choices using patient data. Enables direct, head-to-
head comparison of treatment options that are scored independently by safety, efficacy, and cost
at the Point of Care, at the moment of decision making. Value – drives competition for outcomes by
providing transparency into safety, efficacy, and price head to head. Reduces inappropriate
decisions, increases cost-effective and safe use of medications, leading to medical cost avoidance
and superior outcomes
.
Couples Incentives with
Evidence-Based Medicine
derived from Big Data
vs. passive physician
education
Signal‐to‐Noise –
Information Overload
Realtime
Biological
Waveforms
in
Context
Comparative
Efficacy of
Medications
US Patent 2010
Personalized
Head-to-Head
Comparisons
Real-time Population Analyses
(who needs to be contacted)
Real-time, Single Patient CDS
(what to do Right Now)