The Emergence Of Operational Biosurveillance As A Professional Discipline

Parallels With The Birth And Operationalization Of Modern Meteorology

The most direct, and in a sense the most important, problem which our conscious knowledge of nature should enable us to solve is the anticipation of future events, so that we may arrange our present affairs in accordance with such anticipation. -Heinrich Hertz (1898)

The emergence of operational biosurveillance as a professional discipline and early warning service to the public bears remarkable similarity to the birth of modern meteorology. 

Meteorology as a practice and the concept of "forecasting" the weather was started by Vice Admiral Robert Fitzroy of the British Navy in the mid-1800s.  The promotion of meteorology as a professional discipline was to be his final career in life and suffered the hardship of every social pioneer seeking to change societal function.  He was ostracized by academicians, fought with competing individuals and organizations, struggled for funding and recognition by the British government, and as a lifelong sufferer of depression ultimately committed suicide under the strain.  The world owes the term "forecast" and the operationalization of weather forecasting to Fitzroy and his remarkable persistence.

In the early 1900s, Bjerknes, a physicist who was interested in applied theory, became intrigued with atmospheric science.  He used medical analogies to describe the state of meteorology as a professional discipline.  He declared in 1901 meteorology held prognosis as its highest ideal, and the basis of prognosis was diagnosis.

Bjerknes pushed for standardization of parameters (humidity, temperature, etc.) and declared the fate of the problem of diagnosis and prognosis (i.e., forecasting) of storms was linked to the availability of data.  He tapped into the newly emergent aerological community, who were enthusiastically using balloons to take upper air readings.  This community was comprised of individuals of a wide range of backgrounds, education, and expertise.  They were essentially hobbyists.  He pushed for standardization of measurements and developed indirect methods to approximate data gaps.

Fate provided Bjerknes with a means to promote near-real time data reporting across large tracts of land, courtesy of the emergence of the telegraphic network.  This enabled rapid, distributed sharing of information; distribution of standards in data collection; and rapid transmission to users of the information. 

Market demand emerged first via a national security concern: the advent of World War I and utility of weather information to inform bombing runs via zeppelin and balloons.  Later, the commercial markets picked up the slack after the war with the birth of commercial aviation and demand for improvements in crop and fishing production.

Here meteorology as a discipline was far more art than science.  But it did create the demand for scientific discovery to underpin what would eventually become a routine social commodity: weather data.

Carrying the analogy further, let's take a look at the birth of operational tornado forecasting during the 1940s in the US. 

In 1940, there was little difference in tornado forecasting and warnings (diagnosis and prognosis) as in 1870.  Two events would drive a change: World War II and the civilian spin-off of the military weather bureau.  The US' entry in WWII meant censoring domestic weather data due to concerns about the Germans using the information for bombing raid targeting.

Many of America's munitions plants were in tornado-prone areas, a serious worry for employees.  The creation of a civilian Weather Bureau improved employee morale.  By December 1942, a volunteer storm spotter service was in place in >100 munitions facilities.  At the Bluebonnet Ordnance Plant in McGregor, Texas, the spotter network alerted the plant 150 times, prompting evacuation 6 times.  Broader efficacy was limited due to poor telephone coverage...

In spring of 1943, the first experimental civilian tornado warning program was created in St. Louis, Kansas City, and Wichita.  Impressions of this initial capability were varied:

The position and direction of movement of tornadoes on eight occasions as reported to and broadcast by the Wichita Weather Bureau has shown that public reports on such storms can be reliable and that people can react in a rational manner when radio warnings of such storms are received. -FW Reichelderfer, Chief, Weather Bureau (1945)

A statement to the effect that a tornado is approaching makes news which is carried by the press to all parts of the country, creating unfavorable publicity for the city and the community.  We feel that the interests of all concerned may be served by refraining for the word tornado. -HM Van Auken, general manager of the Wichita Chamber of Commerce (1948)

The US Air Force was the first to initiate a formal tornado forecasting program.  The trigger for this program was an undetected tornado that touched down on Tinker Air Force Base, Oklahoma City on 3/20/1948:

• 8 injured
• $10M in damage to aircraft, destroying: 17 C-54s, 15 P-47s, and 2 B-29s
• Blew out the control tower

The Air Force undertook an investigation of the March 20 events and concluded, "due to the nature of the storm it was not forecastable given the present state of the art."

Forecast information was passed to the Oklahoma Highway Patrol (OHP) and Red Cross (RC) to enable them to proactively prepare for disaster conditions.  As more forecasts were issued by the military bureau, the channel to the OHP and RC became inadequate, as it was clear the word was not getting to the public.  This resulted in the Amarillo Globe proclaiming on May 16, 1949,

The weather bureau, like most government bureaus, is so scared of making a mistake that it straddles the fence.  Its fear of error results in a string of continuing errors.  If some observer would take the bull by the horn and forecast 'possible' tornadoes he'd be fired if the tornadoes didn't bob up!  If there's any one department of the government that needs a complete over-hauling more than any other it's the US Weather Department.

In Feb 1951, The Air Force established the first center dedicated solely to predicting severe weather and tornadoes: the Air Weather Service Severe Weather Warning Center (SWWC) at Tinker Air Force Base.  That same year, the SWWC issue 156 tornado forecasts; they were accurate 102 times (65%)
From 3/25/49 to 5/31/50, they forecast tornadoes accurately 31/34 times (91%).

Timeliness was just as important as accuracy:

• 6 hrs advance notice was given 27% of the time
• 3-6 hrs advance notice was given 37% of the time
• 1-3 hrs advance notice was given 30% of the time
• Tornadoes appeared before advance notice 2% of the time
• Exact time of tornado appearance was unknown 4% of the time

The SWWC's forecasting came into question on 9/1/52 (Labor Day).

Few base personnel were on duty, and severe weather was forecasted- the prediction was ignored, however.  A tornado with 125 mph winds touched down at Strategic Air Command's Carswell Air Force Base:

• 107 B-36s were parked wing to wing, tied down preemptively with 3/8 inch cable
• 106 were damaged, and 1 was destroyed; more than half of the US strategic bombing capability was incapacitated

Tension between the military SWWC and the civilian Weather Bureau increased over time due to the SWWC's successful, but poorly communicated forecasts- the public was consistently the last to hear warnings of tornadoes.  This ended when Congress engaged in 1952 to ramp up funding to the Weather Bureau... the precursor of today's National Weather Service.

The Emergence of Biosurveillance As a Professional Discipline

Today we are dealing with the challenge of diagnosing in order to provide prognosis for infectious disease events, crises, and disasters.  Events that range from "windy" to a "Category 5 hurricane".

The challenge, like the early days of meteorology, has been to identify the appropriate parameters to detect and track these infectious disease "storms".  The rapid expansion of the Internet, like the telegraphic networks, has enabled rapid collection and sharing of data.  The proliferation of blogging communities devoted to detecting and tracking infectious disease events is remarkably similar to the hobbyist aerological community who took readings of the weather.  Standardization and coordination of these communities' effort into relevant and meaningful input is likewise a challenge.

Following the 2001 Amerithrax attacks and warning failures of SARS in 2002, national security interest likewise funded Project Argus and spawned conceptualization of the "Cheyenne Mountain" or "NORAD" for biological threat early warning: the National Biosurveillance Integration Center at the Department of Homeland Security. 

In biosurveillance we have struggled with the concept of event severity scales, analogous to the scales used in meteorology and the natural disaster community such as the Richter, Saffir-Simpson Hurricane, Beaufort Wind, and tornado Fujita (F-) Scales.  Our initial attempts to define different phases of infectious disease event evolution resulted in the creation of the Wilson-Collmann Scale (WCS), defined as:

• Stage 0- conditions favorable to trigger an epidemic
• Stage 1- uni-focal event
• Stage 2- multi-focal event
• Stage 3- infrastructure strain and depletion of local response capacity
• Stage 4- social collapse

The WCS was derived from a single anthropology study that reviewed over 200 ethnographic studies describing social disruption due to epidemics, and later augmented with a review of current open source reporting of outbreaks and epidemics.  Like heuristic models used in the natural disaster and meteorology communities, we were compelled to redesign this model several times until a more operationally valid approach was achieved.  The most profound development was the merging of disaster sociology into our tactical assessments, using the following operational definitions as a guide:

Infectious disease crisis: time-sensitive, non-routine organization-level decisions that may affect a local community’s activities of daily living. It is more common such decision-making falls within the roles and responsibility of a public health institution than a public or private hospital or individual healthcare provider.  This becomes a community level decision-making activity in countries where there is no public health capacity.

Infectious disease disaster: when crisis mode decision making by public health officials or institution fails to control the situation, either from an informational or response perspective and substantial social disruption associated with features of community disintegration occurs as a result.

The term "community disintegration" refers to the dissolution of a community as a social unit.  This is classically exemplified by the evacuation of a village due to an uncontrolled epidemic.  Ebola is the archetype disruptor capable of generating this kind of social response in undeveloped areas of Africa.

We discovered quickly that even Ebola, with its high potential to generate disaster conditions, is associated with a routine pattern of social recognition and reporting.  This is directly analogous to Bjerknes' attempts to define meteorological events through indirect methods.  We became sensitized over time to these indicators in a manner identical to a physician over time learning to quickly recognize the signs of a heart attack.  The process we went through was essentially learning how to "diagnose" an infectious disease "storm".  After learning when and where these patterns emerge in the world, we learned how to anticipate patterns and become adept at issuing early warnings for biological threats.

What to Notify About

Key issues we ran into in the early days were what infectious disease events are important to notify people when detected?  What are the respective priorities? 

When we discussed this question with CDC and USDA, the answers took days to work through because when you ask people to "rack and stack" notification priorities beginning with a "Red Alert" notification where you will wake the Director out of bed on a moment's notice all the way down through a list of hundreds of pathogens, one finds the task extremely difficult. 

The salient point, after years of doing this work, is where does the brutal cut-off exist?  We decided to work backwards from the standpoint of what infectious disease events can take a community offline from its Activities of Daily Living (ADLs)?  Those activities encompass the essential elements of activity that defines an integrated community (see Infectious Disease Events, Crises, and Disasters: An Operational Definition):

• work
• education of children
• religiously related activities
• organized leisure activities
• unorganized social play of children and adults
• voluntary activities for charitable or other purposes
• treatment of sickness, birth, death (healthcare)
• buying and selling of property
• buying consumable goods (food, etc.)
• saving and borrowing money
• maintenance of physical facilities (roads, sewers, water, light)
• protection from fire
• protection from criminal acts

All pathogens worldwide considered, there are very few that can so thoroughly disrupt a community's ADLs to the point it loses integration (i.e. community disintegration, see above for definition).  Those pathogens that do will be discussed in a later post.

Essentially, we were looking for the delineation between pathogens capable of generating crises and disasters along the definitional lines discussed above, where the emphasis is the acute and the non-routine occurrences of infectious disease.  Working backwards from that viewpoint takes you eventually to the key early warning indicators.  

Depending on the culture and location in the world, we recognized there was a baseline level of resiliency in whatever cultural protection was in place.  For the developed world, this would represent healthcare workers working within an established medical and public health infrastructure.  Daily, routine infectious diseases are handled at this level, and provision of warning about these diseases is not deemed 'relevant'.  The situation is different when encountering routine disease that now is associated with non-routine features such as staphylococcal aureus acquiring resistance to vancomycin (VRSA).  Warning of such events would be routinely handled by public health authorities.  Other well-known but uncommon disease requires periodic education and awareness campaigns such as rabies.

Then we found there were a subset of routine diseases that have seasonal profiles and place tremendous demand for focused periods of time on the medical infrastructure.  The seasonal case increases observed with these diseases, especially if overlapping at the same time and place, may cause serious strain on local medical resources.  Examples of this include the range of disruption generated by seasonal type A influenza that has not drifted for a year or two versus a new vaccine-drifted variant that appeared before vaccine could be made available.  Further complications may be observed when vaccine-drifted type A influenza is seen during a year when vaccine is unavailable and a "bad year" of Respiratory Syncytial Virus (RSV) is observed.  This is a situation where Pediatric Intensive Care Unit (PICU) bed excession may occur with subsequent urban grid-level strain.  Here is a case where warning of such imminent infrastructure challenges is relevant for what would otherwise be routine disease.  This is a warning challenge only partially handled by public health.

The case of the non-routine (from the perspective of a local community) is a difficult one.  Depending on how lethal, how fast the agent spreads, who (i.e., children) it affects, and whether medical countermeasures are available will determine the level of social anxiety when event becomes manifest.  It becomes challenging for the biosurveillance analyst to decide the timing of the warning (see below).  Obvious examples would include the introduction of West Nile virus, especially during the first year of emergence when the phenomenon was first recognized.  This is a challenge that is not handled effectively by current public health surveillance systems.

The extreme cases of successful biological weapon deployments represent the grand challenge of national security and will be discussed in later posts. 

In summary, the purpose of timely warning provided by operational biosurveillance is to identify changes in relative risk as opposed to absolute risk.  Absolute risk is already well understood by medical and public health personnel.  It is the change in relative risk that is challenging to quantify.

Critical Information Requirements (CIRs) are those requirements that set the detection, analysis, and notification priorities - otherwise known as the "op flow" of a biosurveillance cell.  Below is an example of general daily CIRs for global monitoring:

• intentional or accidental release of biological agent or a biowarfare attack
• "unusual" or "atypical" respiratory disease
• "hemorrhagic pneumonia"
• suspected pandemic H1N1 vaccine drift
• efficient human to human transmission of non-H1N1/non-H3N2 influenza viruses
• suspected SARS
• suspected smallpox
  
• Et cetera

The above are examples of events deemed worthy of immediate notification for further scrutiny and potential rapid decision-making to issue an emergency warning to our partners.  The current CIRs for our operational support for the Haiti response effort are provided here.

Of course, the devil of operations is making sure your operational information requirements marry up with your audience's expectations.  We constantly state our operational focus and caveats to manage expectations and remind the audience our perspective.

Who to Notify

One of the greatest challenges we faced with every effort we have been engaged in was who to notify.  One of the problems of solely notifying the federal government was assurance the information would be rapidly shared with state and local authorities, medical personnel, other relevant response personnel, and the public.  These same challenges were observed with the majority of the professional associations we have dealt with over the years.  Moving directly to state or local officials introduced the same issues of official bureaucratic processing and time delays, as well as the difference in operational perspective.  Many local officials and infection control practitioners told us they were busy dealing with MRSA and did not have the bandwidth to discuss events that may or may not impact them.  This was the classic low probability / high consequence warning challenge.  

The greatest synergy we found was with the medical professionals themselves who appreciated situational awareness so long as it was relevant and timely. 

We designed analytic methodologies to understand disease translocation lines via the air traffic grid and potential for a translocating disease to trigger an epidemic at the receiving site.  We employed this system to proactively notify healthcare providers in the southern states during the emergence of pandemic H1N1 in 2009.  The key observation was one city's exposure to a potential translocation site was not equivalent to another city's exposure.  Analogous to one city's weather being different than another city's weather.

The Future

We are seeing the future now with the Haiti response effort, where engagement of the public vis a vie "crowd sourcing" represents an interesting new development.  Facilitated by APAN and Crisis Mappers, we are seeing unprecedented synergism of multiple organizations ranging from NGO to government to for-profit.  Gaps still remain and challenge our ability to rapidly detect and verify events of concern, but the rapid expansion of social media has enabled a decoupling of hierarchical official reporting systems for more agile expression of situational awareness.

Background Reading

1.  http://www.amazon.com/Scanning-Skies-History-Tornado-Forecasting/dp/0806133023/ref=sr_1_1?ie=UTF8&s=books&qid=1230610448&sr=8-1

2.  http://www.amazon.com/Appropriating-Weather-Bjerknes-Construction-Meteorology/dp/0801481600

3.  http://www.amazon.com/Storm-Warning-Origins-Weather-Forecast/dp/0750932155/ref=sr_1_33?ie=UTF8&s=books&qid=1265135345&sr=8-33