96 PD Predictive Modeling: Now What?
Moderator:
Kara L. Clark, FSA, MAAA
Presenters:
Philip Fiero
Syed Muzayan Mehmud, ASA, FCA, MAAA Prashant Ratnakar Nayak, ASA, MAAA
Advanced Predictive Modelling
Phil Fiero, Vice President, Predilytics Inc. June 2014
TM
Predilytics is a
healthcare analytics
company that
generates insight from
big data
to:
• Improve quality of care
• Coordinate care
• Attract and retain members
• Manage reimbursement and shared savings
• Reduce costs
We use the latest
machine learning technology
and
computer science to identify and
predict
opportunities at
both the
population and individual member level.
This approach
enables use of our expansive
non-clinical
data
on over
225M lives
and our customers’ structured
and
unstructured clinical and financial data
to optimize the
power and economics of predictive modeling at the
individual level.
Who is Predilytics?
Predilytics serves health, services, and risk bearing entities:
• Health plans • Health systems • Providers • Health services • Medical device Manufacturers At both the member
• Use all available data to improve population and individual health
– 95%1 of the “data wake” we all leave annually is not in the healthcare
system
– Individual behavior is best predicted by socio-economic and lifestyle characteristics and consumer activities, not typically found in EMR and Claims Data
• Machine learning and advance computer science are required to convert massive amounts of data into actionable insights, by
optimizing identification of targeted events at the actionable cohort
• Individual insight, population impact by deploying interventions with highest probability of success
The ideas that drive new analytic
approaches. . .
SOURCE: IDC; US Bureau of Labor Statistics; McKinsey Global Institute analysis, May 2011 Big data: The next frontier for innovation, competition, and productivity
Rx Mental
Health
Admissions
Provider Office Visits Labs Required Highly recommended Recommend
Data
Inputs
Member Enrollment Details Claim History EMR/Clinical Including Notes CMS Files MOR/MMR HRA Call Center Logs & Details External Data Census Voting Consumer SocialExpansive use of data:
Demographic, administrative, operational, clinical
Incorporate data in any format, structured or unstructured
Approach maximizes data intake to drive highest order
Unstructured data mining and linguistic analysis
Can provide more accurate and predictive model results
•
Claims and membership data often represent the majority of model
input data
•
However, specific words and word pairs in the comment fields can
increase the predictive lift of the models (natural language engine)
•
Examples of data with free text that can be mined:
–
HRA data
–
Sales force notes
EXAMPLE:
Presence of the words: “SON” or “DAUGHTER” maps to the concept
of family involvement and changing situation
ANALYSIS:
When a son (or daughter or other family member) becomes involved, it
may be an early indication that the parent is experiencing health issues – it can
also be an early flag for disenrollment or exploring health plan changes
–
Clinical visits
–
Call center notes
Illustrative external data sources:
Public, consumer, financial, social media
Matched holistic view of over 220M people
Public Healthcare
• Medicare, Medicaid • Population Stats
• Healthcare Providers, Cost, Quality • AHRQ, NIH, CDC
• Health Outcomes
Consumer
• Consumer Behavior / Purchasing • Ethnicity
• Social Security / Death Records • Voter Registration • Legal / Regulatory Financial • Consumer spending • Credit risk • Public records
• Real estate indicators
Social Media
• Facebook Activity • Foursquare Check-in • Twitter Activity
Machine Learning background
•
Machine learning is a technology in which
software evaluates a data set
and
combinations of data sets
millions of times
to learn and predict data
relationships
•
Machine learning is capable of exploring
more data, faster
and more
thoroughly
than traditional statistical techniques
• Predictive patterns in the data are discovered and retained
• The software builds on previous learnings and highly predictive equations evolve
• Genetic Algorithms (GAs) are a form of machine learning that are highly effective in spotting subtle patterns in data sets. GA modeling technology and the output are transparent and more actionable
• Traditional modeling relies on statistical analyses of data, in particular
various forms of regression, which carry with it certain limitations that are not found in iterative – based learning models
• These patented algorithms have been
consistently
used in the
financial
services and marketing
industries for enhanced business success
Genetic Algorithms (GA)
125
models per
generation
in 10
seconds
10,000
generations
performed
1.25 Million
equations
evaluated with
learning
past
to next generation
Low
Fitness Accuracy Scale
High
Model 7 Model 8 Model 9 Model 10 Model 11 Model 12
Generation Two
Model 13 Model 14 Model 15 Model 16 Model 17 Model 18
Generation (n)
Model (n)
Current Served Populations
• Historical experience indicates 1/3 of
population at risk of not recertifying
• With predictive analytics “at-risk” individuals
can be identified increase probability of failure to recertify to 90% likelihood
• Improve business performance by
appropriately allocating resources to targeted cohort
Failure to recertify risk
Applying analytics to allocate resources
New Populations
• Integration of consumer behavior, social
claiming can identify risk in unknown populations
Sample Outputs:
Creating member target lists
Predilytics Prospective Member List
Operationalize established models to Analytic
Warehouse. Design, review, and develop member extracts.
Generation, validation, and delivery of member extracts in the pre-determined
format.
Integrate directly into client operations systems and processes
Sample Outputs:
Key model drivers
Purpose: Identify the top clinical and non-clinical drivers at a member level in order to support intervention operations. This methodology allows end users (providers, case managers, etc.) of to understand key risk drivers in a comprehensive and actionable way.
• Key drivers may include, but are not limited to, existing/pre-existing conditions, demographic, consumer, utilization, and financial attributes at a member/risk level.
• This information facilitates identification of the appropriate intervention at a member level, as well as provides an area of focus for those at the point-of-care.
Member CHF Hosp.
Risk Decile Key Driver 1 Key Driver 2 Key Driver 3 Person A .92 1 2-IP-HCC85-CHF GAP-CCSPx44-Lipid
Profile
GAP-GC3- Beta-Adrenergic Blockers
Person B .81 2 1-EM-HCC22-Morbid
Obesity Poor LDL-C Control GAP-Annual EM Visit
Person C .97 1 History-CABG GAP-GC3- Ace inhibitors
2-Visits-Cardiologist-Last 30 days
Big Data
Healthcare Analytics Machine
Learning
Delivering patented machine learning healthcare data analytics to generate meaningful insight to solve healthcare industry challenges
