In my M.Sc. eHealth degree I took a course where I dipped my toes into the murky depths of public health informatics. This article on syndromic surveillance at the London 2012 Olympics just touches on the panoply of data sources now available to track and predict disease outbreaks:
http://www.popsci.com/science/article/2012-07/how-it-works-worlds-largest-health-surveillance-system
How do you tell if a flu is dangerous enough to bring down the
Olympics? Map diseases in real-time, throughout the entire country
From a Public Health Perspective, the Olympics Can Be a Dangerous Place Athlete's at the 2009 Mexican National Olympics wear masks to protect themselves from an outbreak of swine flu.
Right now in London and various sites around the UK, more than half a
million international travelers are sharing stories, beers, doner
kebabs, close living quarters and--let’s be frank--the occasional
mattress. Roughly 17,000 athletes and officials from hundreds of
countries are packed into the Olympic Village alone, and that doesn’t
take into account the spectators--more than 8 million tickets will be
punched at the Games--who have piled on top of greater London’s nearly 8
million inhabitants. Culturally speaking, it’s a marvel that we can do
this and all get well enough along. Epidemiologically speaking, it’s a
nightmare scenario.
An international gateway city like London is certainly no stranger to
the comings and goings of large numbers of people, but when this many
people settle into close proximity over an extended period--eating the
same things and sharing the same spaces for several days or weeks at a
time--it’s the perfect recipe for pathogen outbreaks. A microorganism
that might be fairly benign in one part of the world can blossom among a
population with low immune resistance. Food-borne illnesses threaten at
every food stall. Something flu-like, that spreads through normal human
contact, could potentially clean house.
The key to maintaining the public health during something as massive
as the Olympic Games isn’t battling every individual bug but quickly
containing any pathogen that does get loose in the population, and to do
this the UK’s Health Protection Agency (HPA) has developed what is
being billed as “the world’s largest health surveillance system.” By
rapidly meshing many different streams of information pouring in from
hospitals, general practitioners, clinics, infirmaries, and health care
hotlines across the UK, the HPA has developed a way to measure and
monitor the public health in near-realtime, ensuring that any pathogen
outbreak is quickly identified, diagnosed, and contained before the
whole of London comes down with the sniffles--or worse.
While the HPA’s system is imperfect, stitched together from a mash-up
of pre-existing information channels (reports emailed and phoned in to
regional health authorities), and newer, digital data streams (online
systems that beam anonymized patient data from hospital databases to
authorities automatically), it's likely the most ambitious public health
monitoring system ever deployed. It’s not just a model for future
Olympic Games, but for the future in general; as global populations
continue to swell and our urban centers become both larger and more
connected, systems like this will help cities and states keep vigilant
watch on their people's collective health. Here’s how it works.
SYNDROMIC SURVEILLANCE
"In this kind of epidemiology, timeliness trumps complete data sets. The threat picture is constantly evolving."The
HPA has collected certain kinds of data from hospitals, clinicians, and
general practitioners for some time now, but to be truly useful the
data needs to be centralized as quickly as possible. To that end, the
HPA has automated as much of the data flow as was feasible for the 2012
games. Anonymized diagnostic data now pours in from emergency rooms and
hospital labs in realtime via automated systems--when nurses and doctors
enter a patient’s information into their in-house computer systems, an
anonymized report is automatically generated and sent to the HPA
immediately. The HPA has also asked doctors around the country to
include a notation in such reports for Olympics links, designating the
patient as an athlete or coach, a staffer, a spectator, or someone who
has otherwise been in or around an Olympic venue. Even in cases where
doctors cannot diagnose an infectious illness, the HPA has created a
means of cataloging the symptoms and defining as much as possible the
condition, allowing it to identify new or emerging infectious diseases
that haven't been seen before.
This near-realtime clinical data builds the foundation of what
becomes an evolving threat picture. The HPA models include data from
hospital labs as it streams in over the Web via a secure system called
CoSurv, but the models don't wait for lab work (which can take hours or
days to process) to begin painting that picture. Clinicians have been
put on alert for a spcific list of infectious diseases (authorities are
actually particularly concerned about food borne illnesses, as stomach
bugs can spread with impunity in close quarters like the Olympic
village--just look at historical examples of cruise ship outbreaks for
an example of this) as well as symptoms of chemical contamination. When
they see symptoms of any of these they report them to the HPA
immediately based on clinical (rather than lab) diagnosis. In this kind
of epidemiology, timeliness trumps complete data sets. The threat
picture is constantly evolving as more and better data comes in (from
the lab, for instance), but that the HPA is always working from the most
immediate information it can get.
For the HPA, that paints a pretty decent picture of what’s happening
in hospital wards, but to fill in the holes in their canvas officials
there wanted more data from alternative sources--general practitioners,
walk-in clinics, or people calling into national health hotlines--that
aren’t plugged into the same system. This data isn’t quite so
automated--generally doctors or hotline operators have to phone or email
it in to the HPA--but it adds another layer geographically-relevant
symptomatic data to the HPA’s ongoing assessment.
IF THERE'S AN OUTBREAK
Say a clinic in East London begins showing a spike in incidences of
extreme stomach pain and nausea that is considered above normal levels
for this kind of illness. This symptom data passes through the HPA's
regional office first, and a flag goes up. Perhaps it's an anomaly, but
now it's on the authorities' radar. This data travels on to the HPA's
West Midlands office where it is fed into an algorithm-based computing
system that quickly extrapolates the overall public health picture of
the UK at that moment. Here, the system notes that a spike in similar
symptoms has been reported at Weymouth and Portland on the South Coast,
where the sailing events are held. The system makes connections between
the two; a potential outbreak of some kind of stomach-pain inducing
pathogen is brewing. Health authorities are now on alert.
As one day turns over to the next, more data comes streaming in from
both regions. In East London, it turns out several members of a large
tour group dining at the hotel restaurant got ahold of some salmonella.
On the south coast, a bout of stomach flu is circulating among some
locals unrelated to the Olympics--a blessing, since Olympic athletes,
staff, and spectators are doing a lot of traveling around. The cases are
unrelated and crisis is averted, but the important thing is that the
HPA and other health authorities were able to connect the dots almost
immediately and take precautions. By centralizing symptom and diagnostic
data and running it through these algorithms, the HPA can not only
monitor the entire UK at once, but can identify trends, outliers, and
abnormalities in the public well being with unprecedented speed.
Computationally speaking, this is tough work. A very small and mostly
harmless incidence of a rare pathogen might raise an immediate
statistical flag while a potentially hazardous symptomatic change in a
known pathogen might fall within the statistical “safe zone” while
actually representing the greater threat. The algorithms try to mitigate
for these kinds of statistical problems, and as algorithms do, they
will get better via time and testing. For now, they provide the best
rolling picture of an entire nation’s realtime health that authorities
have ever seen.
The Health Protection Agency is Monitoring the UK's Health Around the Clock for the Olympics: Courtesy Health Protection Agency
BETTER DATA, BETTER HEALTH
While the HPA’s system is far from flawless--relying on general
physicians and nurses to phone in symptom reports not only opens the
door to under-reporting but also pulls them away from their primary
jobs--it is an ambitious attempt to leverage the power of Big Data into
better public health. It’s a fairly strong argument for the digitization
of the medical field in general, and with more information fed
automatically into the loop, future systems built on this model could be
powerful tools for threat prediction and preventative care.
A
system where nationwide, anonymized symptom data flowed freely and
automatically to centralized computing centers like that at the HPA’s
West Midlands office could revolutionize the ability of authorities to
rapidly respond to emerging threats. But the HPA isn’t even tapping some
of the richest data streams available--those provided freely by
citizens themselves. We’ve seen how systems like
Google Flu Trends
can accurately predict incidences of flu outbreak in a given area
simply by trolling search terms for indicators of flu activity. And just
this week we learned about
an algorithm that scanned geotagged tweets in NYC to accurately predict which users were about to get sick up to eight days before they even showed symptoms.
That kind of data is everywhere, and it’s ideal for taking the pulse
of large groups of people--the same kind of large groups currently
tweeting so much from London that Olympic organizers have asked them to
tone it down
(it’s disrupting television coverage apparently). That’s why the UK
plans to leave large parts of its health surveillance system in place
after the Games are overwith (it will likely scale back the frequency of
some kinds of reporting, though it will keep the technology in place to
ramp it back up during an emergency). As global populations continue to
swell and our metropolises become bigger, denser, more diverse, and
better connected, epidemiological situations like that presented by the
Olympics will become less the exception and more the norm. The ability
to quickly crunch disparate data streams into a perfect picture of
public health will be the difference between staying out in front of
emerging biological threats and constantly trying to chase them down
from behind.