Technologies for Pandemic Influenza Surveillance and Response

1
Technologies for Pandemic InfluenzaSurveillance
and Response

• John PageResearch EngineerNorthrop Grumman
  Corporation
• John.page_at_ngc.com

2
Types of Applicable Technologies

• Geospatial Visualization
• Biosurveillance
• Disease Spread Modeling
• Predictive (and Risk) Modeling
• Natural Language Processing
• Resource Allocation/Supply Chain Management
• Outbreak Management
• Collaboration
• Situation Awareness and Decision Support

3
Geospatial Visualization
Avian H5N1 Outbreaks
Domestic Poultry Density
Migratory Bird Paths

• Correlation of events and environmental factors
• Accurate and timely Geocoding of events is
  crucial

4
Biosurveillance Systems

• Use automation to provide early warning
• Numerous Systems have been demonstrated
• Variety of potential data sources
• Variety of algorithms
• National, State and local levels
• Common Issues
• They require a lot of care and feeding
• Calibration, false alarms
• They require training and control data
• Algorithms vary in effectiveness for different
  threats
• They are no better than the data fed to them
• Emerging trend is a system of systems approach
• Common infrastructure
• Multiple detection algorithms in parallel

5
Disease Spread Modeling

• Transportation Models (Air and Land)
• Vector Models (West Nile, etc)
• Socially Based Models
• Micro Models model single individuals and
  communities
• Complex and time consuming to run
• Useful for determing local response policies
• Not readily scaled up for global predictions
• Macro Models model flow among populations
• Coarse-grained predictions but less
  computationally expensive
• MIDAS and Imperial College of London Models
  support what-ifs and other responses

6
Predictive and Risk Modeling

• Combine Linear Regression analysis and geospatial
  reasoning
• Use a set of training data to determine which
  factorsare positively and negatively correlated
  with events
• We developed a risk model based on reported
  avianoutbreaks in Thailand using
• Domestic Poultry Density
• Population Density
• Wildfowl Migration Routes
• Proximity to Wetlands
• Model effectively predicted the first avian
  outbreak in the UK.
• Next step is to apply approach to
  Avian-Humantransmissions (Particularly in
  Turkey, Indonesia)

Thai Training Data
UK Risk Model
NE US Risk Model
7
Natural Language Processing (NLP)

• Finding the needle in the haystack of
  unstructured data
• WHO, OIE, pandemicflu.gov, etc.
• Intelligence community has had this problem for
  years
• NLP can
• Help Identify data of interest
• Assist in georeferencing and
• time-stamping
• Web-scraping applicationscan extract data into
  computableformats
• Web-crawlers can increase timelyidentification

8
Resource Allocation/Supply Chain Management

• Planning and Distributing limited supplies of
  vaccine will be very problematic.
• Site Selection
• Workload Balancing
• Vaccine Transport andSecurity
• Supply Chain Management for Stockpiles

9
Outbreak Management

• Support initial containment and management of
  outbreaks
• Coordinate agencies and experts
• Identify and isolate potential exposures
• CDCs Outbreak ManagementSystem is a good
  example
• http//www.cdc.gov/phin/software-solutions/oms/ind
  ex.html

10
Collaboration

• Humans solve problems technology is only a tool
• Notification services allows experts to receive
  customized alerts
• Communities of interest can be created
  dynamically
• Workflow can support sophisticated interactions
  between users

11
Situation Awareness What is the Big Picture?
?
12
Global Disease Surveillance Platform

• A situation awareness platform for pandemic
  surveillance and response
• Developed as an Independent Research and
  Development project by Northrop Grumman in 2006
• Dr. Taha Kass-Hout is the lead investigator
• Supported the SARS response for CDC
• Taha.kass-hout_at_ngc.com
• Our aim is not to duplicate existing work, but to
  integrate it
• Open J2EE architecture

13
GDSP Architecture

• Ingests multiple sources of structured and
  unstructured data
• Alert and notification features for collaboration
• Incorporates feeds from biosurveillance systems
  and simulations
• Thin (web) and thick (touch table) display
  options.

14
GDSP Partners and Capabilities

• Free Text Searching
• Google
• NStein
• Northrop Grummans Text Trainer
• Biosurveillance
• RODS a Mature system that also supports over the
  counter sales as an indicator
• GIS and Visualization
• ESRI
• Google Maps
• Applied Minds, Inc.
• Disease Modeling
• Imperial College of London
• We are ALWAYS looking for more partners!

15
Platform Portal Application Prototype
RSS Feeds
ProMed Mail
User Alerts
OTC Data from RODS
16
Platform Portal Application Prototype
Local Outbreak Management
Responder Readiness
Global Status
17
Questions?
Taha A. Kass-Hout, MD, MS Chief
Scientist Northrop Grumman Corporation 3375 NE
Expressway, Koger Center/Harvard
Building Atlanta, GA 30341 678-530-3568 Taha.Kass-
Hout_at_ngc.com
Walton John Page, BS, BA Senior Health Research
Engineer Northrop Grumman Corporation 3975
Virginia Mallory Drive Chantilly, VA 20151 (703)
272-5901 John.Page_at_ngc.com