Syndromic Surveillance: Early Warning Systems for Monitoring Emerging Outbreaks of Health Events from Either Natural Causes or from Bioterrorists

Edward L. Melnick^*

Department of Statistics, New York University, 44 West 4th Street, New York City, NY 10012, USA

*Corresponding Author:

Edward L. Melnick
Department of Statistics
New York University
44 West 4th Street
New York City
NY 10012, USA
E-mail: emelnick@stern.nyu.edu

Received Date: September 27, 2012; Accepted Date: September 28, 2012; Published Date: September 29, 2012

Citation: Melnick EL (2012) Syndromic Surveillance: Early Warning Systems for Monitoring Emerging Outbreaks of Health Events from Either Natural Causes or from Bioterrorists. J Bioterr Biodef 3:e107. doi:10.4172/2157-2526.1000e107

Copyright: © 2012 Melnick EL. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Visit for more related articles at Journal of Bioterrorism & Biodefense

Keywords

Syndromic Surveillance; Emerging diseases; Bioterrorists; Functional data analysis

Introduction

Biodefense, as defined in Wikipedia, refers to short term local measures designed to restore biosecurity to a collection of people in a given area, who are subjected to biological warfare. While the US Department of Defense had originally focused on applications of vaccines to fight pathogens in the environment, its scope has expanded to include protection of water and food supplies.

For a successful biological defense there must be a robust surveillance system that can quickly detect the presence of an infectious agent that could have been introduced by nature or by a terrorist. For example, if an aggressive bacterium is detected early in an environment, providing an appropriate antibiotic might greatly reduce the morbidity rate in the population. Early detection systems are required to monitor the outbreak, spread, and trends of a disease, as well as reassure the public that an epidemic has not occurred.

Public health surveillance has been in effect for many years and is a major responsibility of the Center for Disease Control and Prevention (CDC), which has created databases and computer systems that track and monitor emerging outbreaks of illnesses. Unusual patterns in the data, such as clusters or spikes, are often indicators of changing behavior that might signal events that could have a strong impact on public health if the signal is observed before a disease erupts in the population. However, if the signal were only generated after the disease has been inculcated into the population; potential public health strategies that could have been effective in fighting the disease would no longer be useful. The captured data in the early years of the program were the records of individual patients (retrospective) and were used for tracking the projection of the disease in order to measure the public health services’ ability in controlling the disease. What was needed was an early warning system that allowed sufficient time to install strategies for containing diseases that could develop into epidemics. This translated into seeking data that were prospective in that they generated signals on what might happen in the future. The need for an early warning system lead to the development of syndromic surveillance, which involved collecting and analyzing data on health trends before a health problem had been identified and, therefore, before there was a need to focus on a specific disease.

Syndromic surveillance

The CDC defines syndromic surveillance, as the study of real time health related data that precede diagnosis and signals a sufficient probability that cause an outbreak of a disease which warrants further public health response. These data streams have also been used as input to statistical algorithms designed to detect bioterrorist attacks. Finally, the data can also be used to track the development of a disease as well as determine the effectiveness of medical strategies designed to destroy the biological agent causing the disease.

Surrogate data are used in syndromic studies based on actions people take when they are aware of early symptoms or take precautionary actions. Some of these actions include: the taking of over-the counter drugs and conducting internet searches (self medication), making calls to emergency rooms in the hospital and being absent from school and/ or the work environment (precautionary).

Because of the uncertain relationship between the syndrome and the targeted disease, diagnostic testing is the most important step before issuing an alert about a developing disease pattern. This is necessary since the alerts are based on data sources related to behavioral responses to the disease rather than to the actual disease in question. It is also important that the alerts are made sufficiently far in advance of the disease so that the public health officials have sufficient time to allocate the necessary resources for combating the disease pathogens. In today’s world, syndromic surveillance based strategies have been more effective for anticipating natural disease outbreaks than for detecting bioterrorist attacks, characterized by lack of data combined with the need to generate warning signals far in advance to allow appropriate action be taken to mitigate the attack. Obviously, generating false signals cannot be tolerated. In the short run it would create panic among the population and if too many false signals were generated they would be ignored. If, on the other hand, believing in a system that does not generate timely signals about an attack would weaken a defense system that had a false sense of confidence when believing it was protecting the country.

Although the concept ofsyndromic surveillance was developed to generate early warnings about the possible development of a disease process, its most important application might be the indication of a potential problem and the suggestion of further investigations into the possibility of a developing health hazard. An example where syndromic surveillance has had a positive impact is described by J Ginsberg et al. [1] in Nature where Google was used as a search engine for detecting an influenza epidemic. Essentially, the authors’ monitored health-seeking behavior from online web search inquires. Before using this data to detect early influenza activity, the authors measured the correlation between physician visits and Google searches to track influenza questions. Once the correlation was found to be quite large, the number of queries was used as a surrogate for influenza like symptoms that successfully indicated the forming of an influenza epidemic.

With an early warning system, treatment of a disease begins much earlier, thus minimizing health consequences. It allows for further levels of protection to the population by making vaccines available, isolating infected individuals, and creating situational awareness for health delivery officials so they canbetter manage their responses and identify clusters of unusual diseases such as those resulting from dangerous pathogens that could have been introduced by terrorists.

Problems occur with syndromic surveillance strategies when the analysts miss relatively low correlated syndromes and non-health factors with high correlations to the disease process provide false signals. Examples are in the use of absenteeism as a syndrome for a disease when the absenteeism actually resulted from holidays or inclement weather, or the use of over-the-counter drugs as the syndromic agent when the sales were actually driven by a pharmaceutical company’s promotional advertisement.

In 2003 the BioSense system was introduced by the CDC to receive and analyze clinical data in real time. The goal was to provide: 1. Awareness for suspected illnesses; 2.Confirmation or refutation of the existence of a health issue; and 3. Monitoring the parameters and spread of a disease. In October 2011, Jeffrey Runge [2], reporting for Politico, stated that the government’s management of biological warfare was fragmented among a number of government agencies including the Department of Homeland Security, Department of Health and Human Services, and the intelligence agencies, which impeded a coherent strategy against a biological attack. In July 2012 President Obama issued the National Strategy for Biosurveillance. In an accompanying letter sent with the report was a statement that the US “…must be prepared for the full range of threats…terrorist attacks involving a biological agent, the spread of infectious disease, and food borne illness …to identify threats as early as possible”. The letter went further by calling for a “… coordinated approach that brings together Federal, State, local, and tribal governments; the private sector; nongovernmental organizations; and international partners”. The plan was designed to take advantage of existing efforts and focus on a set of four guiding principles.

Strategies for analyzing syndromic surveillance data

Traditional methods for signaling a change in the disease process were based on changes in levels of the data that exceeded a threshold (e.g., 3 standard deviations) from a pre-determined baseline, which is adjusted based on known cycles within the data (e.g., seasonal variation). The BioSense system, for example, has the capacity to produce maps reflecting spatio-temporal characteristics in the data. These characteristics include: discontinuities, jumps, clusters, and the development of new clusters that might signal new outbreaks of a disease.

Finding clusters with in spatio-temporal data are most commonly discovered using Kulldorff’s scan statistic [3]. The statistic is the total number of reported cases that occur in a cylinder where the base (circle) is a geographical region and the height corresponds to time.

A newly proposed strategy for studying spatio-temporal data is based on methodologies designed for performing Functional Data Analysis (FDA). The goal is to generate signals denoting changes in the disease process not based on the number of reported cases, but by noting acceleration in the number of reported cases. This technology has been successfully used to detect activations in brain activity [4].

The application of FDA for disease detection begins with the collection of syndromic data (over-the-counter sales) at the basic level (pharmacy). The data collected over a given time period is then transformed into a functional space with smooth curves, surfaces, hypersurfaces, etc. (Fourier transforms, wavelets) where the basic elements are 3 dimensional cubes (regions on a map represented on a Cartesian plane, and the amount of activity at a period of time). A goal of the analysis is to study the acceleration of the curves at specific locations and at specific time periods by first describing the curves with differential equations (see Ramsay and Silverman [5] for details). It is believed that a change in the curves describing the syndromic data will generate signals of change earlier than waiting for the number of cases to exceed a given threshold. Although the theory has been developed, the concept is now being tested with both simulated and live data to determine robustness of the derivative as a signal for change in the medically related data.

Conclusion

There is a long tradition of governments tracking incident rates of various diseases beginning with the bubonic plague in Europe. After the terrorist attacks in the US on 11 September 2011 and the anthrax tainted mails sent to government officials, the world became focused on the possibilities of bioterrorism and how to contain the threat. First attempts to understand the problem were based on government data that were used to track new and spreading diseases. Since these data were not useful for controlling the disease process newer methods were introduced to warn health officials about emerging diseases. It was within this climate that syndromic surveillance data were introduced. Although the data has potential value for generating early signals, the problem has not been solved especially for diseases developed by bioterrorists. This paper gives an overview of the current state of monitoring diseases and some ideas that might improve the process.

References

1. Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski MS, et al. (2009) Detecting influenza epidemics using search engine query data. Nature 457: 1012-1014.
2. Runge J (2011) Re-examining Our Bio-defense. Politico.
3. Kulldorf M (1997) A Spatial Scan Statistic. Communications in Statistics - Theory and Methods 26: 1481-1496.
4. Tian TS (2010) Functional data analysis in brain imaging studies. Front Psychol 1: 35.
5. Ramsay J, Silverman BW (2005) Functional Data Analysis. (2ndedn), Springer, New York.