Warning of the 2009 H1N1 Influenza Pandemic: An Operational Perspective

Introduction

 

In the modern era, media remains an important source of infectious disease event reporting.  “Global epidemic intelligence” in the near-real time operational setting has, since the mid-1990s, relied on computer automated online media harvesting capabilities that may or may not include human supervision of translated output.  Examples include the Canadian Global Public Health Intelligence Network (GPHIN)¹, ProMED², and HealthMap³.  Media reports that reach the attention of the World Health Organization (WHO) and the WHO-facilitated Global Outbreak Alert and Response Network (GOARN) are evaluated against the International Health Regulations.  Should criteria be met, then verification that may include an epidemiological investigation or laboratory confirmation is facilitated through communication with GOARN partners that may include subject matter experts deploying to the site in question to obtain samples for laboratory confirmation and conduct an epidemiological investigation. ⁴

 

Those that monitor global media use interpretive techniques more akin to open source intelligence than classic epidemiology.  Key limitations to the use of media include diagnostic ambiguity in the information, experience and credibility of the media source, official government control of media outlets, and contextualization of the information within indigenous cultural baseline.  An ability to gather information across multiple data sources such as human ground networks is essential for signal interpretation, enhancement of credibility, and ultimately, ability to act on the information.  Key to this process is an ability to communicate within a trusted network of professionals to evaluate warning information of varying levels of ambiguity and credibility. ^5,6

 

Until recently, there was little recognition of the need for a professional discipline that utilizes near-real time multi-source information to detect, track, and issue notification of infectious disease events of potential international importance.  Here we present the efforts of multi-lingual analysts at the Veratect Corporation who participated in warning of unusual respiratory disease in April 2009 in Mexico, later discovered through laboratory confirmation to be the first recognized area in the world with rapidly transmitting novel swine influenza (H1N1), that evolved into the first pandemic since 1968, now known as pandemic (H1N1) 2009.

 

Methods

 

Veratect Corporation currently maintains two operations centers staffed with multi-lingual analysts who collect raw information in 28 native vernacular languages and are highly trained in culturally sensitive interpretation of reporting behavior and social responses to infectious disease events, crises, and disasters.  These analysts draw upon multiple sources of near-real time information that includes online media harvesting.  For signal validation purposes, Veratect is confidentially partnered with more than 14 organizations that provide direct ground observations in 238 countries.  The model of event detection and verification follows that of the WHO GOARN. ⁵

 

In the interpretation of raw information, analysts are trained on proprietary grounded taxonomies⁷ that map the socio-cultural context of an infectious disease event.   As analysts evaluate information in the native vernacular language, they assess the described event for whether it is associated with indicators of a routine occurrence of disease (an “event”); whether time-sensitive, non-routine organized decision making is occurring in response to disease where such decisions may affect a local community’s activities of daily living (a “crisis”) ^8-10; or whether perceived failure of crisis mode decision making has occurred with subsequent indicators of community disintegration (a “disaster”) ^11-16.  Analysts working with different cultures have become accustomed to locally-relevant nuances in reporting behavior, and changes in both the behavior in reporting as well as reported social reactions to an infectious disease event are passed internally through a warning notification protocol inspired by the meteorological and natural disaster communities. ⁶ 

 

An online web portal containing Veratect’s up to the minute global infectious disease event detection, tracking, and notification information was provided pro bono to the United States Centers for Disease Control and Prevention (CDC), the World Health Organization (WHO), and the Pan American Health Organization (PAHO, the WHO Regional Office of the Americas) several months prior to the worldwide acknowledgment of pandemic (H1N1) 2009 in Mexico.  Additionally, an email user alert system was created in collaboration with the Global Disease Detection Operations Center (GDDOC) at CDC that provided alert for 1) any infectious disease event assessed by the analyst to have evolved to a crisis or disaster that was 2) within 300 km of an international airport connected to the United States by direct, non-stop air traffic. This algorithm typically presents approximately 1 to 3% (2 to 6 event reports) of our daily operational output for the world to CDC in the form of an email push.

 

Results

 

The detection timeline and notification actions for the H1N1 crisis in Mexico are provided in Table 1 and online data supplement that accompanies this manuscript.  Documentation of the timeline begins with first report of “unusual” respiratory disease (as described by local communities in Mexico) and is carried through to the point of CDC and WHO’s first public announcement of the crisis.

Table 1.  Evolution of Veratect Corporation’s awareness and notification of the Mexico crisis.

Laboratory confirmation of unusual respiratory disease in Mexico was not known until eighteen days into the timeline.  Alerting and notification was based primarily on judgment that asserted local community behavior was unusual both in reactions to the events themselves and how the events were reported by local media.  Specific examples of this include reference to “unusual” respiratory disease, use of the phrase “atypical pneumonia” and reports of an attempted diagnostic rule-out for SARS. “Atypical pneumonia” was a phrase used during the early days of the emergence of SARS in southern Asia in 2002 and 2003 and thus considered a key indicator of interest.

 

Notification protocols were escalated during the timeline, which reflected not only a change in the content of the disseminated warning information but the communication mode.  The first level was posting to an online website in a “pull” model, where viewing of the information depends on the user.  The second level was an automated user alert email that did not provide a listing of all infectious disease crises and disasters throughout the world but rather those within proximity to an international airport connected to the United States by direct, non-stop air traffic.  In this case, the email alert generated on April 16 and 17^th to CDC was due to event proximity to Oaxaca airport, which was located approximately 150 miles from Reforma and connected by direct, non-stop air traffic to Houston, Texas.

 

Notification by phone to CDC was initiated due to awareness of an international request for diagnostic assistance by Mexican authorities, along with reports of possible more widespread involvement of Mexico with “unusual” respiratory disease.  During the conversation, information was presented in terms of what was and was not known.  It was emphasized that while claims of SARS was unlikely, the level of non-routine social concern and disruption and knowledge of an international request for assistance demanded immediate attention.

 

Figure 1a displays the volume of respiratory disease event reporting by Veratect for Mexico and the United States.  Figure 1b displays the volume of respiratory disease event reports in Mexico where the events were in proximity to an airport connected to the United States by direct, non-stop air traffic.  Escalation in the notification protocol was activated before an overt increase in the volume of source reporting of the events in Mexico.  The April 6^th report, available on the Veratect web portal, was derived from 6 sources of information.  The April 16^th report, which was sent by email, was derived from 2 sources of information.  The April 20^th phone call was triggered by communication with a highly trusted individual.

 

[a]

[b]

 

Figure 1.  Respiratory disease reports issued by Veratect during emergence of the H1N1 pandemic.  Notification of unusual respiratory disease in Mexico began on April 6.  Figure 1a displays all Veratect respiratory and influenza reporting for Mexico and US.  Figure 1b displays Veratect reporting for Mexico filtered by proximity to international airports connected to the US by direct, non-stop air traffic; all respiratory disease and influenza reporting for the US also shown.

 

Steps were taken to disseminate information to other organizations and governments due to the realization of a potentially serious, rapidly evolving international public health crisis occurring in proximity to multiple international airports.  None of the domestic U.S. authorities Veratect contacted indicated awareness of these events at the time of notification and requested access to situational awareness.

 

At the suggestion of Colorado state officials, Veratect activated a Twitter account on April 24^th to distribute all real-time updates of the evolving crisis.  The intent was to freely provide situational awareness until official public health organizations had an opportunity to inform the public. By April 29, the Veratect Twitter account was the second most retweeted (reposted) source of information on Twitter.com under the #swineflu trend group behind CDC’s CDCEmergency Twitter account. ¹⁷  Figure 2 displays the increase of public followers of the Veratect Twitter feed, an indicator of public demand for information.  Veratect ceased distribution of information through its Twitter account at 2000 Eastern Standard Time on May 15, 2009.

Figure 2.  Daily public subscription to the Veratect Corporation Twitter account.  Reporting of the H1N1 crisis on the Veratect Twitter account began on April 24.

Discussion

 

A critical aspect of warning is communication within a framework of trust. ¹⁸  This was the single most important aspect of the warning process during the 2009 pandemic of H1N1, where an expression of concern, level of uncertainty, and ultimately, an adjustment of community sensitivity and orientation to a certain issue was achieved.  Communication of Canadian engagement in the Mexico crisis was an essential piece of information to drive sensitization of the US public health and healthcare communities.  Collegial communication with CDC on April 20^th occurred within this framework, and their response was an immediate attempt to coordinate with international colleagues to understand the nature of the crisis in Mexico. ²⁰

 

The evolution of reporting and resolution of uncertainty was similar to the time frame observed during the 1968 influenza pandemic.  In 1968, 28 days passed between initial social awareness of respiratory disease in Hong Kong and 20 days after local recognition of unusual event features, respectively, before laboratory confirmation of the presence of a genetically shifted strain of influenza. ²¹  In the case of the 2009 pandemic of H1N1, 18 days passed between local report of unusual respiratory disease and publicly declared laboratory confirmation of a novel H1N1 strain.  We propose poor reporting and healthcare infrastructure in the initial areas of transmission in rural Mexico, as well as delays in information sharing delayed diagnostic evaluation of the situation.

 

It could be argued that H1N1-positive cases identified in California and Texas was an earlier detection point in the pandemic.  However, H1N1 cases have been identified sporadically in the past for several years.^19,20  We propose the crucial missing piece of information was evidence of rapid, sustained community transmission as observed during the Mexico crisis to enable recognition of an international public health emergency.  Communication of biosurveillance information in a trusted social network of biosurveillance analysts facilitates an integrated situational awareness of public health threats.  Here the process entails bringing together disparate lines of evidence for a fusion of actionable information that enables decision-making.  Information pertaining to laboratory confirmation of H1N1 in California and Texas and that produced by operational biosurveillance analysts is derived from two different professional cultural domains.  Both of these domains are distinct components of an integrated warning system, which will be described below.

 

Collegial communication among professionals in operational biosurveillance occurs through various modes such as email and phone calls.  Conversations occur to share events of concern, collectively resolve discrepancies or ambiguity in the information, elicit different perspectives based on experience and discipline, and ultimately, arrive at a consensus about a particular event.  Such consensus can either be realization an event’s etiology and threat potential is well understood to be of minimal potential consequence, an event’s initial features warrant closer scrutiny and potential ground investigation, or a full escalation of warning notification to multiple parties is required.  The constant processing of indicators, communication of that information, and consultation with professional colleagues has resulted in the emergence of a novel analytic community comprised of different disciplines, experience, and professional perspectives that when combined has enabled realization of a new capability to process a disparate array of biosurveillance information.    This nascent community is comprised of United Nations organizations, Non-Governmental Organizations (NGOs), foreign governments, corporations, academia, and individuals.⁶ Sociologically, this is highly reminiscent of the birth of modern meteorology as a professional discipline.^22,23

 

There is a cultural difference between those that provide early warning and those that respond.  For example, in meteorology the National Weather Service focuses its efforts solely on the detection and warning function, whereas the Federal Emergency Management Agency (FEMA) is more concerned with what happens when and after a storm strikes.  The culture of early warning, and the people drawn to that kind of work, are very different from the response culture of FEMA.  We have found in our experience the culture and people who work in our environment are very different in orientation from the classic epidemiologist or public health officer.  It is similar to the difference in personality between an emergency room physician and a pathologist- both attract very different types of people to the job, but both specialties are essential for patient care. 

 

From a sociological perspective, components of an integrated warning system function together as a social network based on trust.¹⁸  This is true in the biosurveillance operational environment.   Figure 3 displays a schematic of an integrated warning system as it applies to current biosurveillance operations.  Human subject matter experts are required to interpret, process, and communicate signals from detection subsystems such as biosensors, syndromic surveillance systems, open source information, direct clinical observations, or laboratory diagnostic findings.  Fusion of that information, an assessment of credibility and potential threat represented, and consultation with biosurveillance community members are important components of the decision process to generate a given level of warning notification.  Issuance of a warning implies operational sensitization and orientation to the event in question, which requires ongoing monitoring, reassessment for the need to maintain a given state of alert, and communication with biosurveillance community members who themselves may also have a role in translating warnings into response action.

Figure 3.  Integrated biosurveillance warning system, adapted from Mileti and Sorensen (1989). ¹⁸

 

Here we have documented the first instance of using Twitter-based text messaging to warn of an international public health crisis.  The speed with which the pandemic crisis unfolded demanded a dramatically different method of disseminating warning information to a large body of disparate, multi-disciplinary communities.  The dramatic uptake of information by the public, an indicator of confirmation behavior and demand for information was a classic sociological hallmark of a crisis.¹⁸  Veratect analysts monitored Twitter accounts, websites, and public broadcasts by CDC, the U.S. Department of Health and Human Services and the World Health Organization in order to decide the appropriate time to cease the Veratect Twitter feed.  This decision to cease “Twittering” was a specific attempt to encourage public use of credible official sources and avoid confusion in a rapidly complicated information environment. 

 

In a prior report, we emphasized rapid recognition of foreign infectious disease threats and proactive communication of that information was key to mitigating the potential damage sustained by communities that may be connected to that threat by air traffic, cultural diaspora, or commerce. ²¹  Detection and warning activities in the early days of the pandemic (H1N1) 2009 dramatically emphasized this point.  Aside from voluminous road and foot traffic at the border of the United States and Mexico, substantial connectivity via direct, non-stop air traffic between the two countries was noted.  Canada issued a notification to their public on April 22^nd without waiting for laboratory confirmation of samples from Mexico.  This decision was based predominantly on media based reporting of the situation in Mexico and the request by Mexican authorities to Canada for diagnostic assistance (D. Skowronski, personnal communication).  As mentioned in our prior paper, Canada’s proactive use of early warning information for public health crises and disasters represents a potential new standard. ²¹ 

 

The World Health Organization recently noted “the 2009 influenza pandemic has spread internationally with unprecedented speed. In past pandemics, influenza viruses have needed more than six months to spread as widely as the new H1N1 virus has spread in less than six weeks.” ²⁴  When it comes to unusual respiratory disease events, the global community cannot wait for laboratory confirmation to issue public notifications due to the speed with which pathogens are able to translocate.

 

The definition of “early warning” implies time-sensitive dissemination of actionable information.  The “early” component, by its very nature, is associated with a degree of false positives and sensitivity that requires balance with the receiving community of users in terms of their tolerance levels for frequency of notification and operational resources to verify information within their own network.  The “warning” component implies some form of analytic process has occurred to bring the report in question to the attention of a resource-constrained user community. This balance, from an operational viewpoint, is best achieved with trained professionals serving as intermediates between raw information and decision makers.

 

Conclusion

 

We have suggested a robust analytic discipline is required to provide the appropriate cultural context and experience when interpreting the complex interaction between disease hazards and socioeconomic vulnerability and ultimately produce warning information.  In the case of the current pandemic, warning notifications were escalated based on a low volume of source information assessed by experienced analysts.  This is analogous to the comparison of the “astute clinician” versus an automated syndromic surveillance system, where the clinician is a more reliable source of relevant warning information. ^25-27  As with the evolution of medicine and meteorology as professional disciplines, improvements in performance over time is expected with more experience in this new professional domain as well as encouragement of evidence-based practice and scientific research.

 

Acknowledgments

This work was supported through internal funding by the Office of the Chief Scientist, Veratect Corporation.  The authors wish to acknowledge the assistance of Mr. Kyle Geiger and the analysts who participated in the operational functions described in this paper.

 

References

 

1. Mawudeku A, Blench M (2006) Global Public Health Intelligence Network (GPHIN). 7th Conference of the Association for Machine Translation in the Americas; 8–12 August 2006; Cambridge, Massachusetts, United States of America; .Available: http://www.mt-archive.info/MTS-2005-Mawudeku.pdf. Accessed 3 March 2009.

2. Madoff LC, Woodall JP (2005) The internet and the global monitoring of emerging diseases: Lessons from the first 10 years of ProMED-mail. Arch Med Res 36: 724–730.

3. Freifeld CC, Mandl KD, Reis BY, Brownstein JS (2008) HealthMap: Global infectious disease monitoring through automated classification and visualization of internet media reports. J Am Med Inform Assoc 15: 150–157.

4. Formenty P, Roth C, Gonzalez-Martin F, Grein T, Ryan M, Drury P, Kindhauser MK, Rodier G.  Emergent pathogens, international surveillance and international health regulations (2005). Médecine et Maladies Infectieuses 2006 Jan;36(1):9-15. Epub 2005 Nov 23.

5. World Health Organization (2009).  A guide to establishing event-based surveillance.  World Health Organization Regional Office for the Western Pacific Web site.  http://www.wpro.who.int/NR/rdonlyres/92E766DB-DF19-4F4F-90FD-C80597C0F34F/0/eventbasedsurv.pdf.  Accessed July 28, 2009.

6. Wilson JM (2008) Statement by James M. Wilson V, MD, Chief Technical Officer and Chief Scientist, Veratect Corporation Before the House Homeland Security Committee, July 16, 2008.  Available: http://homeland.house.gov/SiteDocuments/20080716143655-63876.pdf.  Accessed July 28, 2009.

7. Bailey KD (1989).  Taxonomy and disaster: prospects and problems.  International Journal of Mass Emergencies and Disasters 7(3):419-431.

8. Rosenthal, Uriel, ‘t Hart, Paul, Charles, Michael T. 1989  “The World of Crises and Crisis Management.” in Coping with Crises: The Management of Disasters, Riots and Terrorism, edited by Uriel Rosenthal, Paul ‘t Hart, and Michael T. Charles.  Springfield: Charles C. Thomas.

9. Gilbert, Claude. 1998.  “Studying Disaster: Changes in the Main Conceptual Tools,”  What is a Disaster?  Perspectives on the Question,  edited by Henry L. Quarantelli.  New York: Routledge.

10. Barton, Allen H. 1969. Communities in Disasters: A Sociological Analysis of Collective Stress Situations.  New York: Doubleday.

11. Dombrowsky, Wolf R. 1998. “Again and Again: Is a Disaster What We Call a “Disaster”?” in What is a Disaster?  Perspectives on the Question,  edited by Henry L. Quarantelli.  New York: Routledge.

12. Horlick-Jones, Tom. 1995. “Modern disasters as outrage and betrayal.” International Journal of Mass Emergencies and Disasters 13:305-316.

13. Kreps, Gary A. 1998. “Disaster: Systemic Event and Social Catalyst.”  What is a Disaster?  Perspectives on the Question,  edited by Henry L. Quarantelli.  New York: Routledge.

14. Oliver-Smith A. 1998. “Disasters, Social Change, and Adaptive Systems.”   What is a Disaster?  Perspectives on the Question,  edited by Henry L. Quarantelli.  New York: Routledge.

15. Stallings, Robert A. 1998.  “Disaster and the theory of social order.” in  What is a Disaster?  Perspectives on the Question,  edited by Henry L. Quarantelli.  New York: Routledge.

16. Coleman, James S. 1966. “Community Disorganization.” in Contemporary Social Problems.  edited by Robert K. Merton and Robert A. Nisbet.  New York: Harcourt, Brace & World, Inc.

17.  Twopular.  Trend Details: #swineflu. http://twopular.com/trend/_swineflu.  Accessed on July 28, 2009.

18. Mileti, Dennis and John Sorensen.  Communication of Emergency Public Warnings, Oak Ridge, TN: Oak Ridge National Laboratory.  ORNL-6609, 1990.

19. Centers for Disease Control and Prevention (2009). Update: Swine Influenza A (H1N1) Infections --- California and Texas, April 2009.  MMWR Dispatch 58(Dispatch):1-3.  April 24, 2009.

20. Centers for Disease Control and Prevention (2009). Swine Influenza A (H1N1) Infection in Two Children --- Southern California, March--April 2009.  MMWR Dispatch 58(Dispatch):1-3.  April 21, 2009.

21. Wilson JM, Iannarone M, Wang C (2009).  Media Reporting of the Emergence of the 1968 Influenza Pandemic in Hong Kong: Implications for Modern-day Situational Awareness.  Disaster Med Public Health Prep. 2009 May 11. [Epub ahead of print]

22.  Friedman RM (1993).  Appropriating the Weather: Vilhelm Bjerknes and the construction of a modern meteorology.  Ithaca: Cornell University Press.

23.  Halford P (2004).  Storm Warning: The Origins of the Weather Forecast.  Gloucestershire: Sutton Publishing Limited.

24.  World Health Organization.  Pandemic (H1N1) 2009 Briefing Note 3 (Revised).  World Health Organization Web site. http://www.who.int/csr/disease/swineflu/notes/h1n1_surveillance_20090710/en/index.html.  Accessed on July 28, 2009.

25. Lewis MD, Pavlin JA, Mansfield JL, O'Brien S, Boomsma LG, Elbert Y, Kelley PW.  Disease outbreak detection system using syndromic data in the greater Washington DC area.  Am J Prev Med. 2002 Oct;23(3):180-6.

26. Espino JU, Wagner MM.  Accuracy of ICD-9-coded chief complaints and diagnoses for the detection of acute respiratory illness.  Proc AMIA Symp. 2001:164-8.

27. Mandl KD, Overhage JM, Wagner MM, Lober WB, Sebastiani P, Mostashari F, Pavlin JA, Gesteland PH, Treadwell T, Koski E, Hutwagner L, Buckeridge DL, Aller RD, Grannis S.  Implementing syndromic surveillance: a practical guide informed by the early experience.  J Am Med Inform Assoc. 2004 Mar-Apr;11(2):141-50. Epub 2003 Nov 21.