In case we needed any reminding, new infectious disease threats keep emerging, as evidenced by Middle East Respiratory Syndrome (MERS), chikungunya in the western hemisphere, and Ebola in West Africa. As surveillance professionals, we are likely to be looked to for data to support detection and response efforts, but also for advice as to where strategic investments in surveillance should be made in the course of an epidemic to support the response. Deciding on the optimal surveillance strategy requires the advice of a seasoned surveillance practitioner, but also an understanding of the goals of the outbreak response and of the control measures being put into place. The system needs to generate the information that is needed for decision-making by those leading the response, in time to be useful.
There are several important tasks for the surveillance function — all of course in support of more effective prevention and control:
— recognize the introduction of an infectious agent into a new population as early as possible.
— for diseases with a case-by-case public health response, identify every case promptly so that control measures can be taken
— identify the population groups at greatest risk of infection, of disease, and of severe disease or death
— monitor population impact of severe as well as mild disease
— monitor the geographic distribution, size, trajectory and end of the outbreak, epidemic or pandemic that may result from an introduction
— help to determine whether control measures are being effective
Some kinds of important questions probably can’t be answered through surveillance, but instead would be answered through focused case and contact investigations during outbreaks, or by planned cohort, cross-sectional or case-control studies. For example, when in the natural course of infection and illness are people infectious to others? How many asymptomatic infections are occurring?
The tools we have at hand to accomplish surveillance goals are diverse, and the right mix of these tools will vary depending on the disease, the epidemiologic situation, and the stage of the event:
— individual case reports of suspected cases from clinicians, both through the reportable disease mechanism and as clinician calls to public health authorities about unusual or alarming cases of disease
— laboratory reports from clinical laboratories, including reference laboratories, and from public health laboratories, either of increased detection of the condition of interest or of inability to characterize certain infections
— notifications of or queries about apparent disease outbreaks, by any of a wide variety of reporters: physicians, school nurses, child care center operators, organizers of group events, news media reporters, etc.
— increased chatter or mention of certain symptoms or diseases on blogs, internet news sites, social media sites, etc
— numbers of deaths recorded with certain causes of death mentioned on the death certificate
— numbers of visits to sentinel practices with a syndrome suggesting the presence of the emerging pathogen
— number of hospital admissions, or ICU admissions, with a syndrome suggesting the presence of the emerging pathogen, or with a suggestive admitting diagnosis
— numbers of visits to emergency departments, urgent care centers and other sites participating in syndromic surveillance, either with a syndrome suggesting the presence of the emerging pathogen or with mention of the name of the pathogen in free-text chief complaint or diagnostic fields.
For all these tools, it is easiest to recognize cases of an emerging infection if the disease caused by the infectious agent has clinical characteristics that make it distinctive, or if there is a specific laboratory test. Detection is also aided, early in an event, if a specific unusual travel history or other exposure is associated with likelihood of illness. The more generic the symptoms are — which is especially likely early in the illness — the harder it will be to detect likely cases by presenting illness alone, and either a specific exposure history or laboratory testing will be necessary to detect likely cases. Even with fairly generic symptoms, however — as is the case with the syndrome of influenza-like illness — tallying of healthcare visits by syndrome in comparison to a baseline can be helpful in monitoring the size, scope and direction of an established epidemic. Alternatively, surveillance can be focused on fully-developed disease which is more clinically distinctive, but this has a cost in sensitivity and timeliness of case and outbreak detection.
The underlying point of this little essay is that the optimal surveillance strategy for an emerging infection will depend on the disease's clinical and epidemiologic characteristics, the current stage of the outbreak, the control strategy that needs to be supported, and the relative costs of missing true significant events and of investigating large numbers of unimportant events. The optimal strategy will depend on the desired balance among sensitivity, positive predictive value, and timeliness (whether for cases or for outbreaks). These three are always in tension with each other. Optimizing one of these attributes leads to compromises in at least one of the others, unless there is significant system change. Those in charge of the response to a threat from an emerging pathogen, with a given set of surveillance systems available to them, will need to decide what the desired balance is between high sensitivity (with its cost of false alarms), high positive predictive value (with its costs of decreased sensitivity and slower detection) and increased timeliness (with its costs in both decreased positive predictive value and decreased sensitivity).
Improvements in key characteristics of surveillance systems can decrease the danger from these tradeoffs. For example, implementing electronic laboratory reporting can improve timeliness of case detection with little or no cost in sensitivity or positive predictive value, as can implementing electronic (as opposed to manual) syndromic surveillance. New diagnostic tests that can be performed at the bedside, or even in the field by EMTs, can in theory improve all three parameters, depending on how reliable they are under bedside or field conditions.
Richard Hopkins
ISDS Board of Directors' President
