Epidemiological surveillance is the ongoing and systematic collection, analysis and interpretation of health-related data. The process involves describing and monitoring health events in populations of humans or animals, or, in the case of a zoonosis such as brucellosis, both. For example, typical questions to which a surveillance system might be asked to provide answers include:
How extensive is the infection, and when and where is it occurring?
Which species of Brucella are involved?
Which animal species are involved?
Is the prevalence and incidence (human or animal) decreasing, increasing or static?
When epidemics occur, what is the source, and how is the agent being transmitted?
What strategies should be adopted to control, prevent and ultimately eradicate the infection?
What are the knowledge, attitudes and practices of the populations affected?
What laboratory or field research needs to be undertaken?
Because surveillance systems vary widely in methodology, scope and objectives, that which is important in one system may be less important in another. For example, it may be necessary to compromise the sensitivity (ability to detect infection) of surveillance in order to achieve other attributes, such as simplicity and timeliness. However, the surveillance tests should have adequate sensitivity and specificity and should be properly standardized. Thus, surveillance systems should be flexible, with the motto “adapt not adopt”.
Traditionally, a distinction is made between passive and active surveillance. Passive surveillance (or monitoring) is the routine, mandated reports that a Health or Veterinary Department receives on brucellosis, while active surveillance is where specific efforts are made to supplement the passive data by use of directed investigations, surveys or epidemiological studies. Passive surveillance is generally less costly than active surveillance, but its sensitivity and specificity are generally unknown. Active surveillance is more specific and sensitive, and the performance of the system should be measurable. These two systems of data collection are not interchangeable, but both are necessary and should be integrated wherever possible.
Brucellosis poses a number of challenges in designing an effective surveillance programme. The infection is chronic in both humans and animals, symptomatology and incubation periods are variable, and laboratory confirmations are essential. The human links to the animal reservoirs may be ill-defined. In areas where the disease is of greatest importance, animal populations may be poorly identified, not enumerated or even inaccessible for long periods.
There are ten basic steps in designing and operating a coordinated human and animal brucellosis surveillance system, and these are discussed below.
Surveillance should always be outcome-oriented, and thus focused on events associated with the disease under surveillance. These include specific epidemiological indices such as the total number of cases, incidence and prevalence rates, and severity as measured by days hospitalized and economic impacts such as productive days of work lost for humans or reduced fertility in the case of animals. Indicators of surveillance may take several forms:
numerical, such as the number of known infected herds in an administrative unit;
ratios, such as the number of newly identified herds in a year compared to the same figure for the previous year; or
rates (percentages), where both a numerator and a denominator are available. For example, the herd incidence rate for a year would be the number of newly identified infected herds divided by the number of known (uninfected) herds at risk.
Ideally, rates are preferred, particularly if the population of herds at risk changes over time. Incidence rates (i.e. new cases) are more useful in general as they reflect better the dynamics of the disease or infection under surveillance, rather than do prevalence (i.e. all cases) rates.
Specific indicators of surveillance can be identified, such as:
Performance indicators. These are key, quantifiable and objective measures that indicate whether surveillance is effective. In other words, measures of a country’s capacity to detect disease or infection.
Diagnostic indicators. These are used to identify why one or more of the above performance indicators are below expectation, and suggest remedial actions.
Resource or workload indicators. These are used to measure process events or productivity, such as number of doses of vaccine issued or number of hours worked. These may not necessarily relate to the health outcome and also may not be very accurate. For example, animals missed in a vaccination programme may well be missed from both the denominator and the numerator.
All indicators should be periodically evaluated to ensure they are still appropriate for their original intended purpose. There is a tendency in some surveillance programmes to collect too much data on the chance it might be useful, so always distinguish between “need to know” and “nice to know” information, where “need to know” information is that which is critical to make the system function as designed.
For brucellosis, objectives could include:
For human disease, a specific set of symptoms and signs, together with laboratory tests, are needed to describe possible, probable or confirmed cases. With animals, isolation of Brucella species is used, with or without serological evidence. Whatever system is chosen, it should be both comprehensive and mutually exclusive. In other words, it must be possible to place every herd or every animal into a category and only one category. For animals, this may be positive, negative or uncertain. There should be a time limit on how long an animal could remain in the uncertain category. For herds and regions, specific definitions are critical to measuring progress. Categorization is obviously needed for computer-based recording systems. Abortion, while often an end result of Brucella infection, is an unreliable case definition for infection, as it can be multi-causal, although it may be a very useful sentinel event in the later stages of an eradication programme, justifying laboratory investigation.
Always carefully review what systems are in use already to see if some, or all, can be adapted for Brucella surveillance. For example, if visits are being made to herds or livestock markets for routine vaccinations, it may be possible to draw blood samples for Brucella serology at the same time.
The following questions should always be answered in designing a surveillance programme:
A flow chart should be built up to include each step in the process. Consultation with an expert in computerized database development is strongly recommended at this stage.
There will always be unforeseen problems, especially in any new system, so a pilot test is always required. For example, pre-test questionnaires, forms and computer programs. Major errors can result in participants rapidly losing confidence. Veterinary data collection poses real challenges, especially where owners may be suspicious or uncooperative. Ingenuity and some incentives might well be needed.
The directors of medical and veterinary laboratories should always be involved in the planning stages, as their workloads will be increased. Identify current and future resources for both regional and central laboratories, including training, equipment, reagents and supplies. All tests should be documented by Standard Operating Procedures, and include quality control programmes. Many countries have now automated some laboratory tests, with computerized output. If field tests are to be used, adequate training should be provided, documented, and participants tested, together with regular proficiency testing, including the use of check samples.
Whether paper- or computer-based, errors can always occur. The person with primary responsibility for the surveillance data base should, in conjunction with the epidemiologists, develop a routine of checking for errors, say in 10% of case reports, including missing data, so that major errors can be avoided. Check digits can be incorporated in the records or crucial data can be entered twice in a row, thus confirming its value.
Exploratory data analysis involves using techniques to make the overall dataset more understandable. This may include using visual displays to summarize the main features of the data, simplify their distribution, and clarify the analyses to be undertaken, including evaluating the influence of outliers on the analysis. A wide range of computerized graphics and mapping techniques are now available to produce useful summaries of datasets.
The real art of conducting surveillance lies in interpreting what the data appear to show in relation to the known epidemiological features of brucellosis. By proceeding from the simple to the more complex, including comparisons with historical data, surveillance provides the basis for appropriate actions. A key issue, however, is to know the inherent limitations of the data and being clear in describing them. Be prepared to question constantly. For example, if there have been new cases in an area previously considered Brucella free, has the case definition changed? Why is one district reporting many infected herds while an adjoining district reports none?
While epidemiological studies to identify risk factors are usually designed to collect data separately, it is possible to carry out simple-case control studies using, say, high incidence versus low or zero incidence areas to identify potential herd or flock risk factors that could be followed up by more intensive studies. For example, are large herds at greater risk of being infected than small herds? At this stage, statistical assistance would be advisable.
Obviously, new surveillance information, conclusions and recommendations soon become redundant unless distributed promptly to those with a need to know. Also, unless the providers of the data are kept informed, they may well lose enthusiasm. Therefore a regular reporting system should be developed, whether it be a simple newsletter, posted, faxed or sent electronically to the district level, or a more complex set of analyses for decision-makers. Media such as newspapers, radio, television and websites can be used for public information, and especially for livestock producers.
Ideally, an evaluation of a surveillance system should be undertaken at regular intervals by an independent individual or group, preferably with experience of brucellosis epidemiology. Those charged with responsibility for the system should be asked to document the following components:
simplicity,
flexibility,
acceptability,
sensitivity,
predictive value,
positive results,
representativeness, and
- timeliness.
In summary, surveillance programmes must fulfil at least three major requirements to be considered effective.
Sensitivity. They must be able to detect a high percentage of field events compatible with the symptomatology and epidemiology of brucellosis. Low sensitivity is a clear reflection of under-reporting.
Specificity. After investigation, an effective surveillance system must be able to provide a definitive diagnosis for a high percentage of brucellosis-compatible field events.
Timeliness. Current information must be provided in a timely fashion to enable prompt field responses to situations identified.
The needs must be met of both those conducting surveillance and those utilizing surveillance data in real-world settings. However, there is no perfect surveillance system and trade-offs must be made between sensitivity and simplicity. Each system is unique, and requires a balancing of the efforts and resources put into the system. These general guidelines are intended to make any system more objective, explicit, uniform and simple. Accurate and timely surveillance does not necessarily ensure that the right decisions are made, but it should reduce the chances of wrong decisions. Remember, if the information remains unused, organizing a surveillance system is a waste of resources: time, staff and money.
