Figure 1

Figure 1. Number of cases of syndromic illness by time in a hypothetical bioterrorism attack and two pathways to establishing a diagnosis: syndromic surveillance coupled with public health investigation (upper pathway) and clinical...

Two pathways to establishing a diagnosis are described by the scenarios below and in Figure 1, using a single, clandestine dissemination of an anthrax aerosol as an example:

The early signs of inhalational anthrax include nonspecific symptoms that may persist for several days before the onset of more severe disease (12). Patients with prodromal illnesses seek outpatient care and are assigned nonspecific diagnoses such as “viral syndrome.” Data on patients fitting various syndromic criteria are transferred to the health department and tested for aberrant trends. This process “flags” that a statistical detection threshold has been exceeded. Epidemiologists conclude that a preliminary investigation is warranted and collect blood for culture from several patients. Within 18 hours, one culture yields a presumptive diagnosis of anthrax, prompting a full-scale response.

Some persons in whom inhalational anthrax develops will have short incubation periods and prodromes (12). Respiratory distress occurs in one such person, and he is hospitalized. Routine admission procedures include blood cultures. Within 18 hours, a presumptive diagnosis of anthrax is made. The patient’s physician informs the local health department, prompting a full-scale response.

Figure 2

Figure 2. Timeline to presumptive anthrax diagnosis, 11 patients with inhalational anthrax, 2001, United States. Abbreviations: Dx, diagnosis; OutPt, outpatient visit followed by discharge home; ER, emergency room visit followed by discharge home....

In practice, how a bioterrorism attack might be detected and diagnosed will probably be more complex. Published descriptions of 11 persons with inhalational anthrax in the United States in 2001 (13–19) provide some insight into this issue (Table 1 and Figure 2),¹ even though that epidemic was too small and geographically diffuse to be detectable by syndromic surveillance. For six patients with known dates of exposure, the median duration between exposure and symptom onset was 4 days (range 4–6 days). The median duration between onset and the initial healthcare visit was 3 days (20) (range 1–7 days), and the median duration between onset of symptoms and hospitalization was 4 days (range 3–7 days). Two of the 11 patients visited emergency departments and were sent home with diagnoses of gastroenteritis or viral syndrome 1 day before admission. In one patient, a blood culture obtained in the emergency room was read as positive for gram-positive bacilli the following day, which prompted recall of the patient. The culture was subsequently confirmed as positive for Bacillus anthracis. Two other patients were seen by primary care physicians and sent home with diagnoses of viral syndrome or bronchitis 2–3 days before admission, including one patient who was begun on empiric antibiotic therapy. For seven other patients, initial emergency room or hospital visits led directly to admission. In addition to the patient whose blood culture was obtained in an emergency room, seven others had not received prior antibiotic therapy, and B. anthracis was presumptively identified from blood within 24 hours of culture. One of these seven patients was the index patient, in whom B. anthracis was also recognized in cerebrospinal fluid within 7 hours of specimen collection. Three other patients had received antibiotics before blood cultures were taken (one as an outpatient and two at the time of hospital admission), requiring alternative diagnostic methods.

Despite the small number of patients, their experience offers four lessons for detecting an epidemic of inhalational anthrax. First, a key objective of syndromic surveillance is to detect early-stage disease, but fewer than half of these patients sought care before hospitalization was necessary, and the interval between such care and admission was relatively narrow (1–3 days). This finding suggests that syndromic surveillance data must be processed, analyzed, and acted upon quickly if such data are to provide a clue to diagnosis in advance of late-stage disease. Second, emergency room data are a common source for syndromic surveillance, but detecting an increase in visits coincident with hospital admission may not provide an early warning because the time needed to process surveillance data and investigate suspected cases would be at least as long as the time for admission blood cultures to be positive for B. anthracis. Blood cultures are likely to be routine for patients admitted with fever and severe respiratory illness, regardless of whether anthrax is considered as a diagnostic possibility, and B. anthracis grows readily in culture in the absence of prior antibiotic therapy, as observed in most of these patients. Thus, if emergency room data are to be useful in early detection of an anthrax epidemic, those data would need to be for visits that occur before hospital care is required—a pattern observed in only two patients. Third, the four patients who received early care and were discharged to their homes were assigned three different diagnoses, which suggests that syndromic surveillance systems must address the potential variability in how patients with the same infection may be diagnosed during the prodrome phase. Fourth, rapid diagnosis after hospitalization was possible only in those patients who had not received antibiotics before cultures were taken. This finding emphasizes the importance of judicious use of antibiotics in patients with nonspecific illness.

In addition to the specific attributes of individual bioterrorism agents, multiple considerations will shape the recognition of a bioterrorism-related epidemic. Five of these attributes follow.

How persons change locations after an exposure will affect whether disease occurs in a concentrated or wide area, and thus whether clustering is apparent to clinicians or detectable through syndromic surveillance at specific sites.

The attributes of the CDC category A bioterrorism agents that affect their detection, as well as the benefits of early detection, are summarized below, on the basis of potential bioterrorism-related epidemic profiles developed by experts (12,22,23,26–28). These profiles reflect current knowledge of these diseases; their epidemiology might differ if novel modes of dissemination or preparation were employed. Each disease has attributes that could increase or decrease the likelihood of early outbreak recognition through either clinical diagnosis or syndromic surveillance.

This category includes multiple infectious agents that range from having a relatively broad to narrow incubation period (e.g., Ebola, 2–21 days; yellow fever 3–6 days). These diseases present with nonspecific prodromes that may have an insidious or abrupt onset. In severe cases, the prodrome is followed by hypotension, shock, central nervous system dysfunction, and a bleeding diathesis. The differential diagnosis includes a variety of viral and bacterial diseases. Establishing the diagnosis depends on clinical suspicion and the results of specific tests that must be requested from CDC or the U. S. Army Medical Research Institute of Infectious Diseases. The value of postexposure prophylaxis with antiviral medications is uncertain, and (with the exception of yellow fever, for which a vaccine is available) response measures are limited to isolation and observation of exposed persons, treatment with ribavarin (if the virus is one that responds to that antiviral drug), and careful attention to infection control measures (23). Patients seen with symptoms during the prodromal phase may not clearly fit into a single syndrome category, but syndromic surveillance focused on the early signs of a febrile bleeding disorder would be more specific.

One of the biggest concerns about syndromic surveillance is its potentially low specificity, resulting in use of resources to investigate false alarms (6,10). Specificity for distinguishing bioterrorism-related epidemics from more ordinary illness may be low because the early symptoms of bioterrorism-related illness overlap with those of many common infections. Specificity for distinguishing any type of outbreak from random variations in illness trends may be low if statistical detection thresholds are reduced to enhance sensitivity and timeliness. The likelihood that a given alarm represents a bioterrorism event will be low, assuming that probability of such an event is low in a given locality. Approaches used to increase specificity include requiring that aberrant trends be sustained for at least 2 days or that aberrant trends be detected in multiple systems (2). Another approach to enhancing specificity would be to focus surveillance on the severe phases of disease, since the category A bioterrorism infections are more likely than many common infections to progress to life-threatening illness. For those diseases that are likely to progress rapidly, such as pneumonic plague, syndromic detection of severe disease (e.g., through emergency room visits, hospital admissions, or deaths) may be more feasible than detection aimed at early indicators before care is sought (e.g., purchases of over-the-counter medications) or when illness is less severe (e.g., primary care visits). Whether detection of syndromic late-stage disease offers an advantage over detection through clinical evaluation will depend on the attributes of the infections and diagnostic resources, as described above.

Predicting how the mix of relevant factors would combine in a given situation to affect the recognition of a bioterrorism-related epidemic is difficult, although mathematical models may provide further insight (5). The most important factors affecting early detection are likely to be the rate of accrual of new cases at the outset of an epidemic, geographic clustering, the selection of syndromic surveillance methods, and the likelihood of making a diagnosis quickly in clinical practice.

Ongoing efforts to strengthen the public health infrastructure (34,35) and to educate healthcare providers about bioterrorism diseases and reporting procedures should strengthen the ability to recognize bioterrorism outbreaks. For example, in New Jersey in 2001, reporting of two early cases of cutaneous anthrax was delayed until publicity about other anthrax cases prompted physicians to consider the diagnosis and notify the health department, suggesting that opportunities for earlier use of postexposure prophylaxis were missed (36). In addition, while the importance of new diagnostic tools, including rapid tests, should be emphasized (37), the essential role of existing diagnostic techniques should not be overlooked. Clinical suspicion is critical, and a key prompt for arousing clinical suspicion may be the microscopic examination of a routinely collected specimen, as occurred in the index case of the 2001 anthrax outbreak, when a Gram stain of the cerebrospinal fluid led to the diagnosis (15). However, as recently highlighted by the Institute of Medicine, the use of basic diagnostic tests has decreased because of efforts to reduce the costs of care, the increasing use of empiric broad-spectrum antibiotic therapy, and federal laboratory regulations, such as the Clinical Laboratory Improvement Amendments of 1988, which have discouraged laboratory evaluation in some clinical settings (38).

While we have focused on the role of syndromic surveillance in detecting a bioterrorism-related epidemic, other uses of syndromic surveillance include detecting naturally occurring epidemics, providing reassurance that epidemics are not occurring when threats or rumors arise, and tracking bioterrorism-related epidemics regardless of the mode of detection (4,6,10). Syndromic surveillance is intended to enhance, rather than replace, traditional approaches to epidemic detection. Evaluation of syndromic surveillance to consider the spectrum of potential uses is essential. A certain level of false alarms, as the result of either syndromic surveillance or calls from clinicians, will be necessary to ensure that opportunities for detection are not missed. Efforts to enhance the predictive value of syndromic surveillance will be offset by costs in timeliness and sensitivity, and defining the right balance in practice, particularly in the absence of an accurate assessment of bioterrorism risk, will be essential.

Two committees of the National Academies have recommended more careful evaluation of the usefulness of syndromic surveillance before it is more widely implemented (5,38). Because the epidemiologic characteristics of different bioterrorism agents may vary in ways that affect the detection of epidemics, these evaluations should address the epidemiology of specific bioterrorism agents. Efforts to detect bioterrorism epidemics at an early stage should not only address the development of innovative new surveillance mechanisms but also strengthen resources for diagnosis and enhance relationships between clinicians and public health agencies—relationships that will ensure that clinicians notify public health authorities if they suspect or diagnose a possible bioterrorism-related disease.