1. How
can we access BioSense?
To obtain access,
BioSense users will need to sign-up on the CDC Secure Data
Network (SDN). We have requested each metropolitan reporting
area (MRA) to identify a BioSense Administrator who will be sent
instructions for applying for a digital certificate. Once the
certificate has been issued, the BioSense Administrator can
register other individuals in his/her organization via the SDN
enrollment site. If your area is not an initial MRA, contact
the BioSense Help Desk at
BioSenseHelp@cdc.gov to determine your appropriate BioSense
Administrator. |
2.
Are the components of BioSense "portable”?
BioSense is a
web-based application, and therefore can be accessed via the
Internet with a digital certificate obtained through the CDC
Secure Data Network. As a web-based system, it can be accessed
from anywhere, but the back-end application relies on a large
data warehouse, and was not designed to be portable. However,
BioSense is beginning to partner with local jurisdictions who
want to have local data analyzed through the BioSense system,
and is also interested in working with partners to make
components of the technology available in other settings. |
3.
Our agency is using early detection software now, should
we discontinue in lieu of BioSense?
No one system meets
all early detection needs at this point. We encourage you to
evaluate what systems best meet your early detection needs and
strongly advocate for the use of the Public Health Information
Network (PHIN) standards so that all early detection systems
will be able to work together and exchange data. |
4.
Is documentation available such as "on-line help" for
interpretation of results in displaying tables, charts or maps?
BioSense will have
an on-line help system, a user’s guide, and documentation on the
data sources, analytic methods and visualization techniques as
well as how data are used in the system. |
5.
Could we have BioSense implemented in our local region?
BioSense is an
on-line application that can be made available to any state or
local public health entity. The system is national in scope, but
presents regional views, so any local jurisdiction can have
access to specific analysis for their region. No additional work
is required to implement the system if data exists for that
region in the current data sources. Currently limited data are
available through BioSense. Regions need to be represented by
an adequate amount of data in order for the output to be
useful. A major focus for BioSense in ’04 and ’05 is to make
more data and data sources available. |
6.
Will this system be sensitive enough to detect a small
outbreak (e.g. anthrax 2001) from daily background events?
Different types of
outbreaks will be detected in different ways. For many
outbreaks, an astute clinician will be the best detector for
some time to come. In BioSense, we are trying to make
day-to-day fluctuations apparent for appropriately detecting
medium to large outbreaks, but also flagging specific diagnoses
relevant to public health. Regardless of the method of initial
detection, these BioSense approaches have value in subsequent
detection, quantification and localization of an event. |
7.
Is CDC providing resources for monitoring the analytic
results, or is CDC relying on public health jurisdictions within
these metropolitan areas to do the monitoring using this tool?
The CDC is staffing
the BioIntelligence Center (BIC) to analyze data and assist
local and state health departments with their analyses. In
addition to these activities, the staff will operate an
“analytical help desk” to assist state and local users in the
identification and investigation of potential signals found.
Because most all public health surveillance and investigations
occur at the state and local level, BioSense will be a useful
adjunct to these activities. Recognizing resource limitations
in public health nationally, a major thrust of BioSense is to
minimize the burden of data monitoring, analysis and
investigation. |
8.
How will thresholds for alerting be established?
BioSense currently
provides analytical algorithms to assist local public health
officials in their investigations of multiple data sources. More
refinement of these algorithms will be needed before alerting
can be reliably implemented for multiple data sources. Even
with that refinement, decisions for setting thresholds to alert
will remain with the public health users in their jurisdictions. |
9.
How does a state or local department pursue these
'contracts' for data acquisition and also partner in the
development of BioSense?
We will make
information about contracts available in the coming months.
There are opportunities for states and locals to partner in the
development of BioSense through the feedback in the use of the
system for monitoring, the analysis of the value within the data
streams, and in suggestions of useful analytic techniques.
State and local health departments interested in partnering with
BioSense should contact
BioSenseHelp@cdc.gov. |
10.
Are there published studies that demonstrate a
correlation between perturbations in surveillance for broad
syndrome categories and true outbreaks?
BioSense is not just
about syndromic categorized data, and in many circumstances, it
includes data down to the individual health event level. The
BioSense initiative and software system is intended to advance
near real-time surveillance capabilities and assist public
health departments in monitoring the health status of their
jurisdictions. If you’re interested in the syndromic components
of this activity, go to
http://www.cdc.gov/epo/dphsi/syndromic.htm for an extensive
research listing. An example reference is:
Mostashari F, Fine
A, Das D, Adams J, and Layton M. “Use of ambulance dispatch
data as an early warning system for communitywide influenza-like
illness, New York City.” J Urban Health. 2003; 80 (2,
suppl): i43-i49. |
11.
Is this national only or is there a plan for global
surveillance or for surveillance in U.S. territories such as
Puerto Rico?
The first priority
of BioSense is to cover major metropolitan areas in the United
States related to the BioWatch program. Despite this, many of
the data sources that have already been acquired have broader
coverage. It is our intent to phase in access for all states
and territories as appropriate data becomes available. |
12.
How are regions defined and how do we out find if data is being collected for our region?
Theoretically, a BioSense “region” can be defined for any grouping
of zip codes that a public health department has appropriate authorization to view. Realistically, such a region would need to be represented by an adequate amount of data in order for BioSense output to
be meaningful. Once we have all the initial BioWatch cities “on-line”, we will be able to work with a variety of other jurisdictions to allow access by region. |
13.
In what formats can you receive data: HL7, real-time, batch? How do you secure the data feeds? VPN?
BioSense strives to use PHIN standards (messages, terminology, transport, and security) whenever possible. Most data are securely transmitted on a daily basis.
Some of data sources are batched processed and some are real-time data feeds. |
14.
The data analyses presented and definitions for metropolitan areas require zip code level data;
some data are not "reportable conditions" in our state. Is BioSense a research activity, and does this
activity require provider by provider IRB review?
The reporting of data in BioSense is not a research activity;
it is public health practice that is consistent with he rules and regulations of HIPAA. There is no “reportable
conditions” clause in HIPAA that specifies what data are reportable. The data providers are either providing data
with a data use agreement under the limited data set rules of HIPAA or through authorizations for disclosure
pursuant to the public health provisions of the Privacy Rule. |
15.
Are there other data sources being considered for BioSense (i.e. Medicaid claims data, ER/Urgent care admissions, EMS dispatch data, poison center data, prescription drug sales, etc)?
BioSense data have been exclusively used and prioritized utilizing
several different criteria: timeliness, coverage, validity, healthcare related, coding characteristics,
electronic record reporting, dual-use potential and cost. Input on subsequent data sources will be received
from public health partners. |
16.
Can we be sure there are no political filters to the data?
BioSense data have been exclusively used and
prioritized utilizing several different criteria: timeliness, coverage, validity, healthcare related, coding
characteristics, electronic record reporting, dual-use potential and cost. Input on subsequent data sources
will be received from public health partners. |
17.
What data do you get from LRN labs? I thought all the data was syndromic - not lab - please clarify?
BioSense includes both diagnostic and pre-diagnostic
data sources. As part of the BioWatch initiative, lab results from the BioWatch collectors are be generated in
laboratory response network labs. CDC developed an application for those labs that generates an HL7 standard lab
result message for Category A agents. Hence, between BioWatch and BioSense there will be the capability to
deliver electronic lab results from LRN labs to public health jurisdiction for which that lab result was generated. |
18.
Why limit the ability to
see data only from a local area? What about agencies which cover multiple states or cross other jurisdictional
boundaries? (BioSense Help)
The BioSense Initiative has been working with state and local health
departments to identify appropriate data sharing guidelines. These public health partners were most comfortable
with data being shared in association with specific jurisdictional needs. The default policy for BioSense will
have metropolitan area data being shared with all local jurisdictions that participate in that metropolitan
geographic area. State health departments will have access to data related to their jurisdiction. As part of
the BioSense system, we have also agreed to facilitate modifications to these jurisdictional policies on the basis
of overlapping and contiguous metropolitan areas. Other agencies needing access to the BioSense system will receive
authorization from the appropriate jurisdictional BioSense Administrator. |
19.
How do we define "metropolitan area"?
Will only very large cities be able to access data specific to their jurisdiction?
Many of the BioSense data are compiled at the
zip code level. Because of this, we have initially defined metropolitan areas based on zip codes mapped to
a certain density of population. However, we are also implementing methods to modify that definition as
appropriate based on requests from participant jurisdictions. While the larger “metropolitan areas” have
been pre-defined within BioSense, many data are nation-wide, and there will be opportunities to define and
provide access to other appropriate jurisdictions as granted by the appropriate BioSense Administrator. |
20.
Can you define a "SMART"
score?
The ‘Smart Scores’ in BioSense are calculated using a
model-based approach that requires some historical data. A Generalized Linear Mixed Model is fit to those data to
develop a prediction or expectation for each zip-code, each day. The expectation is compared to what is observed
and an estimate is produced for how likely that observed count is - given the modeled data history. With the
multiple comparisons adjustment, it is constructed so that it may be interpreted as the number of times (days in
the case of BioSense) one would have to observe in order to expect a value that high. The algorithm for this was
provided by Ken Kleinman of Harvard Medical School, Harvard Pilgrim Health Care and Harvard Vanguard Medical
Associates. The value of the ‘Smart Score’ is based on a transformation of a calculated p-value for comparing a
prediction with an observed count using the formula: (p-value x NT)-1, where NT is “number of tests.”
(Because of the caveats to interpretation in using probabilistic inference in an empirical setting, direct
reporting of p-values is avoided.) |
21.
Are you presenting the
"Change" from previous or the count?
‘Counts’ are presented in the time series
line graphs and the different detection algorithms use changes from previous day’s history as part of
their calculation. |
22.
What is the utility for areas with very
few zip codes or for datasets that do not include zip code?
BioSense will present some information to characterize
a zip code such as population density, but the interpretation of BioSense results is dependant upon the knowledge of
state and local experts that understand the nuances of areas within their jurisdiction. Currently, all BioSense data
sources have indicators that can be mapped to the zip code level. |
23.
How much baseline data are required for
the SMART scores?
Baselines can be calculated with a few months of
data to take ‘day-of-the-week’ into consideration but require a couple of years to minimally account for
‘seasonality’ or ‘month’. The SMART scores are currently implemented as an intermediate surveillance algorithm
using several months worth of data. Several years of back-log data are available for some sources. When this data
has been loaded into the data mart, the SMART scores will use additional terms in building the models to account
for the seasonality. |
24.
Do you have a plan to consider forecasting?
The calculated CuSum values are measured in numbers
of standard deviations from a calculated baseline. The Smart Scores are derived from comparisons with a modeled
expectation based on historical data. The ‘expectation’ is recalculated every day and values are forecasted to
compare with the new day’s observed counts. In other words, we are forecasting for comparison with what is
realized. We do plan to consider any forecasting refinements that will further inform the detection capabilities
of the system. |
25.
What are the 11 syndrome categories and
how are the syndromes defined?
Some data in BioSense are categorized as pre-diagnostic
or syndromic. A multi-agency working group was established to identify and define candidate syndrome groups.
Eleven syndromes and corresponding code sets were selected based on a systematic selection process. Definitions
for each syndrome group were created by consensus; and after an exhaustive search through all possible candidates,
individual ICD-9-CM codes were selected. For a complete list and more detailed information on the syndrome
categories visit
http://www.bt.cdc.gov/surveillance/syndromedef/index.asp. |
26.
How do you use 11 syndromes for OTC pharmaceutical data?
Select over-the-counter pharmaceuticals are grouped
into general product categories such as “pediatric cold remedies”, and these product categories are then mapped
to the11 standard syndrome categories based upon the symptoms the product is intended to treat. |
27.
How co-linear are the CuSUM and SMART score
algorithms?
Although we haven’t done any rigorous co-linearity
diagnostics, the two sets of scores over time and place do demonstrate substantial co-linearity, in general. |
28.
We're hearing about MOST and NCEDD at our
local health department. How would BioSense be used in concert with these systems or would it replace them?
There are a number of early detection activities
which may be supportive of or supported by BioSense in the near future. These systems may provide opportunities
for enhancing the data needs of early detection as well as establishing partnerships in pursuing methods for
supplying useful public health data. Substantial work still needs to be done in identifying and optimizing the
best algorithms for early detection. At this point, it is critical for all systems to use appropriate data and
technology standards so data can be shared with all participating public health jurisdictions. |
29.
How will the Outbreak Management System (OMS)
integrate with BioSense?
Part of the BioSense Initiative is to advance early
detection data standards that will work with outbreak management data standards under the broad umbrella of the
Public Health Information Network (PHIN). It is critical for systems like BioSense to have the ability to share
early detection data with other systems (like the Outbreak Management System) to ensure that early detection
leads to outbreak management and appropriate countermeasure administration. We plan to have data from BioSense
be exchangeable with OMS. |
30.
Is the USPS BDS system incorporated in
BioSense accessories?
The United States Postal Service (USPS) Biohazard
Detection System (BDS) does not currently feed data into BioSense. |
31.
What is the Federal Health Architecture?
The Federal Health Architecture (FHA), a Department
of Health and Human Services (DHHS) initiative, is an architecture for federal agencies. It includes “a multi-departmental
business and technical architecture to facilitate identification of collaborative business opportunities that leverage
existing efforts and investments; development of performance measures and outcome strategies; adoption of technical and data
standards; and development of specifications for how to implement those standards.“ For more information on FHA
visit
http://www.hhs.gov/fedhealtharch/index.html. |
32.
Will you be using some form of OLAP software to
analyze data?
To prepare the data for presentation, they are
converted into cubes within SAS for quicker processing. BioSense utilizes the analytical components of the
SAS software along with analytical techniques such as SMART Scores which are programmed specifically for the
analysis of BioSense data. |