|
What is ODIN?
Background
▪ Data Import
▪ Secure Data Store
▪ Statistical Analysis
▪
Database Design & Security Features
▪ Data Visualization
▪ Notifications
▪ Summary
The purpose of this document is to familiarize the reader
with the Outbreak Detection Information Network (ODIN) by
describing the overall purpose of the system and then further
defining the individual components that make up the system.
BACKGROUND
Development of the ODIN technical system is part of a
research project, funded by the Department of Defense. The
intent of the project is to evaluate the effectiveness of
combining civilian health indicator information with military
health indicator information to enable detection and early
response to bio-terrorism events and general health trends such
as avian flu and SARS.
ODIN will implement the emerging science of syndromic
surveillance by using the information gathered in initial health
encounters, such as emergency room visits, to predict disease
outbreaks days to weeks ahead of clinical certainty from
laboratory results.
Syndromic surveillance is not yet a fully accepted scientific
method, thus, one of the purposes of the ODIN project is to
provide additional information and experience to validate it
across a wide selection of methodologies. ODIN is unique in the
fact that it is not geo-politically bound, but is designed with
the capability to provide visibility across geographic
boundaries.
While being funded as a research project, one of the major
goals of the project is to produce a product that will have
extensive functionality for the public health community and
provide a long term toolset for epidemiological activities.
top
DATA IMPORT
The first key component of the ODIN system is the dynamic
data import engine. Data from a wide variety of sources is
required for ODIN to operate. The initial data sets will consist
of hospital emergency room encounter information. Additional
data sources, such as over the counter drug sales, lab results,
and others, are planned for the future.
Since the data feeds come from a wide variety of sources, the
ODIN system has been designed to accept data in virtually any
format. Advanced information processing, analysis and quality
control techniques have been used to enable the system to accept
data in multiple formats using a variety of data transfer
technologies. Initially, the ODIN system will operate on a
single server platform. However, due to the large amount of
statistical pre-processing that is performed on the incoming raw
data and as the number of data sources grows, it will
necessitate modifying ODIN to be capable of distributing the
processing load across multiple servers that are either
centrally located or distributed geographically.
top
SECURE DATA STORE
A second component of the ODIN system is the secure data
store. The secure data store is at the heart of the ODIN
system’s ability to use statistically analyzed information from
the past to detect and warn of anomalous trends in health
indicator data. While one of the primary functions of ODIN is to
generate an alert based on current trends in health indicator
data, the systems ability to analyze different periods in time
make it an extremely valuable tool in analyzing disease outbreak
information from the past. Analyzing patterns that exist in past
health indicator data will allow epidemiologists to better
interpret the patterns being detected in the present. The ODIN
system will provide an invaluable tool in researching and
understanding the past as a tool for predicting the future.
top
STATISTICAL ANALYSIS
The purpose of the syndromic surveillance statistical
analysis engine is to reliably identify trends and anomalies in
the incoming data, while minimizing the number of false
indications of a problem. This ability is key to the successful
adoption of any syndromic surveillance system.
Currently, a single “best” method of performing this
statistical determination has yet to be developed, and a great
deal of work is being done in the field to develop meaningful
analysis techniques. While other syndromic surveillance systems
are built around a specific statistical analysis algorithm, the
ODIN system implements a series of software interfaces that
allows multiple statistical analysis algorithms to be used
simultaneously. This allows ODIN users to compare the
information analyzed using various syndromic coders (the
intelligent agents that analyze the English text ‘chief
complaint’ and convert it to one of a set of syndromes) and
different statistical analysis techniques (two variations of
CuSum and Smart Score are being implemented in the system
prototype). By being easily extensible, the ODIN system supports
the ability to test and analyze new systems of statistical
analysis without the need for re-coding the ODIN application,
thus meeting the goal of providing a platform for meaningful
research as well as a production tool for public health
professionals.
top
DATABASE DESIGN & SECURITY FEATURES
ODIN is being designed to be “database neutral”, and it will
operate on either a Microsoft SQL Server or Oracle database
engine. The database is designed using state of the art
techniques to support Very Large Data Bases (VLDB) including
physical data partitioning, temporal partitioning, and advanced
indexing techniques. Security is a key consideration in any data
system containing health data, and the ODIN system implements a
complete end to end security model. The ODIN system supports
local named user accounts as well as working with external
trusted security management agencies. ODIN allows organizations
to control the security of their own data, allowing or denying
access according to their own specific policies. All external
data transmissions are encrypted, and no sensitive information
is transmitted via the e-mail or SMS messaging systems.
top
DATA VISUALIZATION
Any system containing or analyzing a large amount of data
requires a meaningful way to allow the users to visualize the
results. While the ODIN system is certainly capable of producing
massive amounts of tabular information, such a presentation is
not conducive to presenting the information buried in the
numbers.
ODIN provides a number of graphical interfaces which allow
users to see ranked summarizations of the statistical analyses,
allowing them to quickly identify trends that may be meaningful.
A large set of interactive data filters allows the user to
select sub-sets or super-sets of the displayed data as well as
narrowing down specific parameters. At almost every point in the
ODIN system, the user can double-click on specific information
to see the underlying detail information that makes up the
displayed information. Users can quickly go from statistically
summarized counts to the detail records submitted by the various
facilities. An alert generation capability allows each user to
set up customizable thresholds which cause the visual
representation of the data to change when exceeded, providing
for a quick visual indication of a potential problem.
Rounding out the visualization capabilities are interactive
graphs, charts and maps. ODIN includes the ability to animate
maps over time, further enhancing its ability to visually
present complex information in an intuitive, understandable
manner.
top
NOTIFICATIONS
In addition to the ability to interact with the ODIN system
in real-time via the world-wide-web, system users can set up
user configurable notifications which are automatically
generated by the system as it operates. There are two types of
notifications.
- System Event Notifications are sent to users to let them
know about significant system infrastructure events that
have (or have not) taken place. For example a user may
request a notification when all statistical processing for a
day is complete or a notification when their facility’s data
was not received by a certain deadline.
- Alert Notifications are generated by the statistical
analysis engines to indicate that a statistical result
exceeds a user specified threshold – for example the
observed incidence of a specific syndrome exceeds the
statistically expected value by 150%. ODIN is capable of
sending notifications to user’s e-mails, cell phones and
pagers.
top
SUMMARY
Gaining an understanding of the five major areas of Data
Import, Secure Data Store, Statistical Analysis Engine,
Visualization, and Notifications provides a good understanding
of the purpose and functionality contained in the ODIN system.
As a whole, this software constitutes a powerful, flexible and
leading edge research and production system designed to improve
public health operations in the military and civilian sectors.
There are a large number of other components and sub-systems
which make up the ODIN system, however, they relate to the
internal operation and security functionality of the system.
top
|
