Emerging Infections of International Public Health Importance





Module 3:  Public Health Response  
LECTURE 1 Readings

Workup on the Unknown—Epidemiologic Investigations

Ann Marie Kimball, MD, MPH

  1. Understand the work process for outbreak investigation
  2. Be able to define triggers and alerts in surveillance information
  3. Describe the role of descriptive and analytic epidemiology in investigating unknown outbreaks


Approaches to the unknown epidemic: clinical, epidemiologic, and laboratory synergies

All infections begin as an unknown. The domains of inquiry include, at least, clinical medicine, epidemiology, and the laboratory, and the synergies between these fields are important for detecting unknown agents. It is often with the alert clinician where we start the investigation of unknown (and known) agents; if there are few physicians, nurses, or other health care workers that a population can access, there is thus less reporting.

For known agents, there is a list of reportable conditions that physicians are required by law to report. However, when a physician gets a case with a reportable disease, he/she may not report the case. Our surveillance systems would not be very accurate if it were solely dependent on clinicians. Historically, physician reporting tends to be poor, and numerous studies have documented this. And, in fact, doctors are responsible for reporting, but it is usually the infection control nurse or epidemiologist in a hospital who are central in the reporting procedure. There is a lot of emphasis on pediatric reporting in Washington State, and a lot of that comes from children’s hospitals, which was how the large E. coli O157 outbreak was detected here. That outbreak was mitigated because of the clinical, epidemiological, and laboratory synergies.

Epidemiology involves the description of disease in populations, and the collaboration between epidemiology and laboratory science is essential for disease investigations. Part of the responsibility of the public health epidemiologist is to establish the relationships with not only clinicians but also with laboratorians in order to establish and maintain reliable, routine surveillance.

When approaching an epidemic, we need to ask the following questions:

  • How do we know it is infectious?
  • How do we know it is a cluster?
  • How do we know it is unusual? Could it be an epidemic or a usual occurrence?
  • What do we do to investigate it and what do we do to control it?
  • How do we know we have succeeded?

There have been numerous studies on international disease surveillance. One of the consistent findings is dependence on laboratory diagnosis. In countries where there is one physician for a very large number of people, there is generally less routine reporting. It may be that physicians or nurses are overwhelmed with their clinical load, that they do not have the law that outlines reportable conditions, that the existing laws are inadequate, or that health care workers will only report at certain times. And, if the clinicians receive no response or feedback, they will stop, which can be a serious problem in many developing countries.

Although clinical reporting continues to be a problem, the Internet has helped address some of the logistical difficulties. Routine reporting systems are improving and many agencies are turning to electronic systems. In the Republic of Korea, they have an Internet electronic disease reporting system, which is present in every health center. The providers at each health center input the data obtained on a daily basis and send it into the central Ministry of Health and Social Welfare through the web-based system.

Lastly, in an epidemic there is a higher prevalence of a disease or condition than expected in a population. Thus, there is a need to know the expected baseline level of the disease. However, for a novel, unknown agent, you do not know this expected level, as you have no previous experience in its epidemiology. There is also a need to watch the emotional content of the terminology to avoid unnecessary panic; i.e. a “cluster” may be used more often than “epidemic”.


Determining an infectious etiology

Criteria for “possible infectious etiology” can involve many systems. When there is an ill person with something he/she did not have two weeks ago, there are a number of questions to ask regarding basic etiologies, infectiousness, malignancy, environmental (poisoning), and the possibility of an acute onset of genetic or chronic illness. For example, rheumatoid arthritis can present with an acute condition that could appear to have a possible infectious etiology. Similarly, a clinician may think that a patient has a terrible infection when in fact he/she is having a diabetic crisis. Determining if a condition is due to an infectious agent depends on a rigorous medical algorithm and a diagnostic workup.

After the Four Corners hantavirus outbreak in the southeastern U.S., the CDC was struggling with how to pick up unknown outbreaks. In order to increase the capturing of such cases and have a better understanding of new agents, many factors have to come into play. Detection of outbreaks and new agents occur when there is collaboration between the medical experts, the epidemiologists, and the laboratorians. The political will to look into a situation with greater extent may also be necessary.

There are objective ways to look for potential infectious etiologies. Perkens et al. (1996) described the syndromes considered to be of possible infectious etiology.

[Figure:  Infectious Disease Related Syndromes]

The authors were attempting to prospectively collect data to detect novel, unknown infectious agents. From all the different syndromes that they looked at, fever was the underlying theme. If there is fever, there is potential for an infectious origin.


The Integrated Process for Public Health Disease Response

[Figure:  Integrated Process for Public Health Disease Response]

This is a systematic, integrated work process model for responding to an outbreak. It represents a public health disease response process, where we initiate with a surveillance system. If the surveillance is ongoing and there is no alert, the surveillance simply continues; one of the issues in programming public health is how much you spend on additional information and determining whether or not to continue an investigation. This is topical now in the state of Washington as we try to rebuild our mosquito survey from the time when we had little or no arboviral diseases. There were fewer reasons to follow mosquitoes in our state in the past, but now we need to follow them with the emergence of West Nile virus.

The same issues apply to mad cow disease and its impact. The major justification for continuing to conduct traditional surveillance was that there was no mad cow disease and therefore no threat to the meat supply. However, now with the imported mad cow disease case in this state, we are now doing catch-up in surveillance for the disease among the herds. There has been a ratcheting up of surveillance for neurologic diseases in cows.

If there is an alert, public health goes into initial response, epidemiologic investigation begins with the obtaining of case reports, and diagnostic testing occurs by collecting specimens. All these things often happen simultaneously. As a public health worker, you need to make some response, even if it is only to the people reporting to you. It is also often important to maintain good communication with the public, which is part of the initial response by public health agencies.

After obtaining the case reports, epidemiologists should compare and integrate case information. This process needs to be considered thoroughly. When WHO first started reporting HIV, they were reporting cumulative cases. By starting with an infection and recording cumulative cases, the recovered or fatal cases remain, and there is an ever-mounting toll of cases. Thus, even though cumulative cases are useful, they are not as useful when trying to focus your actions on control and intervention. These are some of the issues that need to be acknowledged when dealing with infectious disease surveillance, especially in international settings.

Creating a line listing is very straightforward. The access code for each individual case is recorded along with the data for the necessary variables. As you begin this process, it is important as an epidemiologist to think ahead to what variables are necessary. Who is a case in your outbreak? What is the case definition? Initially with HIV/AIDS there were three case definitions; depending on the region, cases were actually defined by a slightly different definition of HIV. Thus issues regarding consistency of data can arise.

The parameters around the case definition depend on information that you can capture easily, cheaply and in a timely way. It is good to define “who” specifically as possible, but such information cannot be obtained quickly. So if you put it in your case definition, it is not going to help you in the course of your investigation. After defining the case, epidemiologists conduct active case finding. Analytic epidemiology studies (to define the exposure risks) and descriptive epidemiology studies (to describe the outbreak) occur simultaneously. The initial response should be a specific public health intervention response that can be quantitatively measured.

You need to collect specimens from the cases detected from active case finding and begin to funnel those through the laboratory process for additional confirmation. Specimens must be properly collected and stored. The laboratory can only analyze the collected samples, so the epidemiologist and other investigation personnel have to collect enough specimens in a proper manner. Collecting and following up with the workup on the samples are an important part of epidemiologic investigations. I once did an evaluation of surveillance in Bahrain (in the Persian Gulf region) and the people there said, “We don’t have any problem with gonorrhea. We’ve never had any gonorrhea here.” There was a myth that no sexually transmitted diseases happen there because of the religious and cultural beliefs of that region. However, at the neonatal intensive care unit, I found out that they had collected in the previous year three specimens for endopthalmitis in the neonate for gonorrhea, and all three were positive. That was impressive because the ambient temperature was about 110 F and these specimens had to be transferred from the point of delivery to the room and then to the central laboratory. There were only three collected and all were positive; in fact, when we started looking more thoroughly we found that there were higher gonorrhea rates in Bahrain than there were in the U.S. It was a fascinating process of looking objectively at the information and finding that they have an important disease which they had never looked at before.


Sentinel surveillance systems and the need for early detection

This is an example of sentinel influenza surveillance.

[Figure:  Sentinel influenza surveillance]

The figure shows the national summary of the percentage of visits for influenza-like illnesses (ILI), reported by sentinel providers, for the year 2003-2004. CDC identifies certain providers who see a large volume of patients in their practice and the agency will ask the providers to report to the CDC on a regular basis. Sentinel providers are used for other diseases and in other countries as well; for example, in Japan, they use sentinel providers for their measles surveillance. Sentinel surveillance has both advantages and disadvantages.

  • Advantages: 
    • can obtain basic data and assess temporal trends
    • often more convenient and/or inexpensive compared to traditional, comprehensive surveillance systems
    • potential for early detection
  • Disadvantages: 
    • potentially biased or poor reliability of data and lack of generalizability
      (due to sites selected)
    • provider and/or patient features associated with outcomes measured may change over time
    • lack of true numerator cases
    • unknown denominator
      (i.e. the underlying at-risk population - incidence rate cannot be calculated)

Sentinel surveillance can be problematic--you do not have the denominator data and you never know how many cases you really have--but it does give you a certain indicator if you keep the same providers in the recording pool and their practice and patient demographics do not change over the years.

This figure shows the isolation of the different types of human influenza viruses, namely the A and B types, from WHO/NREVSS (National Respiratory and Enteric Virus Surveillance System) collaborating laboratories.

[Figure:  Defining the "Unexpected"]

From the laboratory perspective, the isolation and reporting of a novel influenza virus from a patient is a sentinel event in itself. The laboratory would presumably call the CDC or WHO if they detect a novel influenza virus, especially if it appears similar to dangerous strains, such as H5N1, which was observed in Hong Kong in 1997.

This figure shows the pneumonia and influenza mortality detected from sentinel influenza surveillance.

[Figure:  Sentinel influenza surveillance, pneumonia and influenza mortality]

A computer program can create the threshold of the usual, background reporting level of pneumonia and influenza mortality per week, based on data collected from many years. CDC uses this epidemic threshold to assess an influenza season. However, because there is only the percentage of deaths due to pneumonia and influenza, the data do not help in implementing influenza prevention activities, such as the use of vaccinations. Essentially, this is only closing the barn door after the cows are already out. But at least you know how many cows have gotten out. That is the utility of the mortality data. The news media is always asking, “How many people have died from influenza? Is it more than usual?” CDC and public health agencies need to be able to answer these questions.

This is a figure presented by DARPA (Defense Advanced Research Projects Agency).

[Figure:  Need for early detection]

They assessed the role of early detection in an anthrax outbreak. Compared to traditional disease detection, early detection (involving data mining of emergency room visits and other methods to detect unusual occurrences) pushes the disease detection curve back for a gain of 2 days, which can be significant in terms of treatment. Thus DARPA has been a big investor in early warning systems in disease surveillance for anthrax and for other potential bioterrrorism agents.



In summary, all infections begin as an unknown, and it is important to know how to approach unknown agents. An effective approach involves the synergistic teamwork of clinical, epidemiologic, and laboratory fields. Surveillance is the initial component in the recognition and response process of an infectious agent, which entails specific actions by numerous agencies, often simultaneously. Surveillance systems are valuable for numerous reasons, from describing the severity of a particular influenza season to the early detection of a potential bioterrorism agent.


  Go to  Readings


  Go to Top

UW Home © 2004 University of Washington Department of Health Services
Box 357660, Seattle, WA 98195-7660