Biology Reference
In-Depth Information
large number of candidates could have been applied, the confidence level for
the event is reduced.
15.2.5 HealthMap (Harvard-MiT)
Since beginning operation in September 2006, the HealthMap project pro-
vides a free and openly available automated surveillance tool run from the
Children's Hospital Informatics Program at the Harvard Medical School
Division of Health Sciences and Technologies (Brownstein et al. 2008; Freifeld
et al. 2008). The system combines both an automatic and a manual detection
component with an emphasis on providing an integrated view of disease
outbreak information. Since 2007, HealthMap has been closely associated
with ProMED-mail. HealthMap benefits from ProMED reports, and, in turn,
ProMED makes HealthMap output available to subscribers via report sub-
missions on the mailing list.
Currently, the HealthMap system monitors news in six major world lan-
guages (English, Chinese, French, Portuguese, Russian, and Spanish) and
has approximately 170 human, animal, and plant diseases under surveil-
lance. Additionally, HealthMap monitors conditions for early warning indi-
cators of outbreaks, such as poisoning, environmental disaster, and conflict.
Terminology is contained within large structured dictionaries that have been
derived from a variety of publicly available sources. Dictionary lists them-
selves are integrated into a large number of language-dependent patterns for
determining the name and location of the outbreak in actual news texts.
The automatic component within HealthMap performs several levels of
analyses on each document. The first stage identifies country and disease
using patterns described above. The next stage utilizes Bayesian methods
to classify news reports into the categories of breaking news, warning, con-
text, old news, and nondisease news. The final stage detects and removes
redundant information. Alert detection focuses on high-impact news using
statistics that determine the recency of an event as well as the reliability of
the news source. Dissemination is made through a variety of formats avail-
able from HealthMap's Web page, including biogeographic maps, links to
external sources, and mailing lists.
15.3 Development of BioCaster
We now turn to a structured discussion of the technical characteristics of
the BioCaster system, which began operation in 2006. BioCaster was an early
adopter of high-throughput cluster technology to minimize the time between
data sourcing and distribution to users. News is gathered on a 30-minute
cycle, which can be shortened as necessary during health emergencies.
