MEDgle is a tech company that addresses healthcare waste and inefficiency by providing a graph-based big health analytics platform. The platform aggregates data from electronic health records, medical literature, and other sources to create a "Graph of Medicine" with over 150 million data points. MEDgle uses this graph to power diagnostic, predictive, and prescriptive analytics through APIs and apps. These analytics can be applied at both the individual and population level to guide care delivery and improve outcomes.
2. MEDgle is a tech company, addressing healthcare’s problem of
over $300 Billion of waste, inefficiency and inaccuracy
spanning 4 Billion care decisions because Health Systems
and Payers cannot:
access and apply the best medical science within existing
and emerging workflows, hyper-personalized, and with
scale for individuals and populations?
(PWC, Gartner)
2
3. To solve this problem we took inspiration from
and looked at the world not as individual elements but
as rich complex inter-connected graphs
3
4. diagnostic, predictive, and prescriptive analytics
… built with data mining + machine learning + physician curation (20K+ hours)
… and created a graph-based big health analytics platform providing
Data-mined
knowledge
Reviewed
knowledge
Expert
Knowledge
API
feedback loops
&
inductive learning
4
Refined with
EHR Data
Sensor Data
3M+ accessible with
14M+ coming soon
7M+ pages
of text, textbooks, journal articles
100GB+
of public datasets and databases
Healthdata.gov
FDA.gov
PubMed
NextBio
Rxlist.com
CDC
Healthweb.org
Nice.org.uk
UptoDate
Cleveland Clinic
Mayo Clinic
NIH
WebMD
Discovery.com
MedicineNet
Medical Journals*
Drugs.com
Merck
Adam
Pedbase.org
Yahoo Health
Emedicine
Cecil 19th Edition
Diff. Diagnosis in Internal Medicine
Walter Siegenthaler 2007
Primary Care Medicine Alan Goroll
and many more
3M+ patient records (Q2’13)
7M-14M patient records (Q4’13)
5. And today, the Graph of Medicine already has
150M+ Data Points
40K+ Symptoms & Signs
4K Diagnoses
7K Procedures
7K Medications
across ages, genders,
durations, lifestyles
5
6. … and examples of the graph-based analytics are
Individual Acute
Analytics
6
Individual Chronic
Analytics
Population Chronic
Analytics
Real-time contextual info
at the point-of-care
(diagnostic+prescriptive)
Personalized Health
Forecast
(predictive+prescriptive)
Population wide health
forecasts and
prescriptive options
7. … built on a scale-out architecture to support web-scale growth!
MEDgle Graph of Medicine and API
Hadoop |Couchbase | ElasticSearch
Health
Stream
Store
Healthcare
Apps across
care
continuum
EHR
text data mining,
supervised learning
7
9. which turbo-charges a number of actionable use case workflows today
Nurse Call Center
Triage & Advice
Software
Home Triage
PHR & Patient
Apps
EHR
Emergency/UC/MD Office
Care Management
Clinical Research & Innovation
Quality Reporting
& Measurement
diagnostic,
predictive,
prescriptive
analytics
diagnostic,
prescriptive
analytics
diagnostic
analytics predictive
analytics
diagnostic
analytics
Population
Health Mgmt
10. Contact: Ash Damle – ash@medgle.com – 617.283.0226
and lets see a few in action!
hyper-personalized
Triage [Kelly – a call-center nurse in Kentucky]
Health Assessment [ Doug – a care manager in NJ]
graph-powered
population analytics [Mark – CMO at an emerging ACO]
11. Contact: Ash Damle – ash@medgle.com – 617.283.0226
Get access to the graph-based big health analytics engine
powering hyper-personalized care @ scale
hello@medgle.com
12. A walk through Fred’s Triage (an example of Iterative Diagnostic Analytics)
Initial Predicted Differential
Dx
(in the background)
Round 1 of Emergency Qs
(Prescriptive Analytics)
2nd Round Predicted DDx
(raw data + R1 answers )
Round 2 of Emergency Qs
(Prescriptive Analytics)
Final ESI + Triage DDx
(raw data + R1, R2 answers )
Post Triage Prescriptive analytics:
an example of personalized acute care options
Post Triage Predictive analytics
an example of assessing a patients current health
Fred: 36 yrs, male, current dx: hypertension (ICD9
401), 210lbs, 5’9”, family hx: CAD, current complaint:
cough 1 day
13. An example of MEDgle Acute Analytics for Fred (Part 1a)
Acute Predictive analytics:
an example of assessing a patients current health
Acute Prescriptive analytics:
an example of personalized acute care options
Predicted Differential Diagnoses, Triage level with negative
answers to emergency Qs, est acute costs, and more
All test scores < 2
stars may not be
of value to order
All test scores < 2 star may
not be of value to order
All test scores < 2 star OTC for
symptom relief
For a mild to
moderate cough
for only 1 day
with no other
presenting
symptoms and
negative to all
emergency
question, it may
not be necessary
for Fred to come
in immediately.
Patient can come
in if his
symptoms
worsen.
Fred: 36 yrs, male, current dx: hypertension (ICD9
401), 210lbs, 5’9”, family hx: CAD, current complaint: cough for 1
day
13
14. An example of MEDgle Acute Analytics for Fred (Part 1b)
Acute Predictive analytics:
an example of assessing a patients current health
Acute Prescriptive analytics:
an example of personalized acute care options
Fred: 36 yrs, male, current dx: hypertension (ICD9 401), 210lbs, 5’9”, family
hx: CAD, change that to having a cough and fever for 3 weeks
Now, with a cough and fever for 3 weeks, the predicted
differential diagnoses and their ordering are different
Some test > 2
stars would be
of value to order
Again, small shift in symptoms and duration results in different
differential diagnoses, estimated cost, labs, and more. MEDgle’s
analytics are highly personalized to the individual and situation!
Due to the non-
resolution of
symptoms within
the expected
time, it is
suggested that
Fred come in
within 24 hours.
Note that the
standard deviation
for est. acute cost
is larger than in
the previous
example as more
tests may be
needed.
Some scores are >2 stars
May make sense to consider
Some scores are >2 stars
May make sense to consider
14
15. An example of MEDgle Chronic Analytics for Fred (Part 2a)
Chronic Predictive analytics:
an example of assessing a patient’s future health
Chronic Prescriptive analytics:
an e.g. of personalized chronic care options and preventative
measures
Fred: 36 yrs, male, current dx: hypertension (ICD9
401), 210lbs, 5’9”, family hx: CAD, chronic analysis [HTN, BMI: 31
(obese), fhx: CAD]
Beyond generating a highly personal
chronic risk profile, MEDgle
generates a health FICO score
(FitScore), est. yearly cost, “health
age” ,and projects his health over the
next 10 to 15 years with his current
weight and with a weight loss of
16lbs.
Translating the predictive
analytics into prescriptive
analytics, MEDgle calculates
Fred’s top improvement
areas, and symptoms to watch
out for, and more.
Combining guidelines with
probabilistic analysis, MEDgle
provides nuanced monitoring
calendars and therapy options
as a starting point for a highly
personal care plan.
These analytics can be
aggregated over a population to
understand what programs
would be most impactful, what
cross-population data points
would be most meaningful, and
more.
15
16. An example of MEDgle Chronic Analytics for Fred (Part 2a-1)
Forecast in detail comparing trajectories of Fred’s current weight (orange line) to a weight loss
of 16 lbs (green line) with no other change in variables
Fred: 36 yrs, male, current dx: hypertension (ICD9
401), 210lbs, 5’9”, family hx: CAD, chronic analysis [HTN, BMI: 31
(obese), fhx: CAD]
16
17. An example of combining Acute & Chronic Analytics for Fred (Part 2d)
Real-time Acute Predictive Analytics with a Chronic Predictive context
an example of continuous health risk assessment combining MEDgle’s Analytics Platform and sensors
Fred: 36 yrs, male, current dx: hypertension (ICD9
401), 210lbs, 5’9”, family hx: CAD, chronic analysis [HTN, BMI: 31
(obese), fhx: CAD]
MEDgle’s real-time Analytics Platform is able to synthesize incoming sensor data
with contextual EHR data to provide a continuous health risk assessment. If at
some point, a triage is indicated to make sure Fred is ok, his care provider team
can be messaged or a nurse call center can reach out to him.
Combining the information, in the
background MEDgle is calculating
what underlying causes are relevant
for Fred’s specific background and
sensor inputs.
*10
17
18. raw data: dx,cpt,andrxehr&claims for 200 people for 2011
Predictive analytics: Prescriptive analytics:
Est Health Cost/yr By Zip
By FitScore (Risk Strata
AF)
By FitScore (∑ Est. Cost/yr)
Top population improvement areas to minimize future costs
An example of MEDgle Analytics for a Population
High At Risk for Diagnoses
Top 26 At Risk for Diagnoses: Automated Top At-Risk Disease
Registries
Top valued diagnostic monitoring tests to conduct to
gather key data to improve prediction accuracy
18
19. raw data: ehr + claims for 10k people for 2009-2012 (Predictive Analytics)
By FitScore (R
AF)
By FitScore (∑ Est. Cost/yr)
An example of MEDgle Population Analytics (Part 2)
Top 26 At Risk for Diagnoses: Automated Top At-Risk Disease
Registries
20. raw data: ehr + claims for 10k people for 2009-2012 (Prescriptive
Analytics)
Gaps of Care vsFitscore
An example of MEDgle Population Analytics (Part 3)
High At Risk Diagnosis vs Opportunity
Gaps of Care vs Age Range
Previous Diagnossvs Opportunity
21. Contact: Ash Damle – ash@medgle.com – 617.283.0226
a graph-based big health analytics platform,
enabling hyper-personalized care @ scale