The FAO has always been concerned with agricultural development and food security. Recent disease epidemics, in both developing and industrialised countries, have once again focussed attention on livestock disease and their potential to harm development. In the context of developing countries, disease epidemics do four things:
They reduce herds and flocks dramatically, which, in the case of pastoral peoples, is a major blow to food security and the ability to survive;
They cause trading partners to - quite understandably - put trade barriers in place in order to protect their own countries from infection. Where livestock or meat exporting countries are affected by epidemics, their "pariah" status can cost millions of dollars in terms of foreign exchange losses, and drive farmers and the local meat industry to the wall.
They are a deterrent to sustained livestock production.
They add significantly to the cost of livestock production through the necessity for the application of costly disease control measures.
When does a disease become important enough to warrant official intervention? Or to merit international attention? Much attention has been given to highlighting this issue in recent years. The International Office for Epizootics (OIE) has classified animal diseases into two “lists” - List A and List B in order to characterise their level of significance in terms of international trade.
The most important diseases are classified under LIST A. The definition of List A diseases is:
“Transmissible diseases which have the potential for very serious and rapid spread, irrespective of national borders, which are of serious socio-economic or public health consequence and which are of major importance in the international trade of animals and animal products”.
List A diseases are:
|Foot and mouth disease||Vesicular stomatitis|
|Swine vesicular disease||Rinderpest|
|Peste des petits ruminants||Contagious bovine pleuropneumonia|
|Lumpy skin disease||Rift Valley fever|
|Bluetongue||Sheep pox and goat pox|
|African horse sickness||African swine fever|
|Classical swine fever||Highly pathogenic avian influenza|
Of lesser importance are the LIST B diseases. Their definition runs as follows:
“Transmissible diseases which are considered to be of socio-economic and/or public health importance within countries and which are significant in the international trade of animals and animal products.”
This group includes such diseases as: Rabies, Heartwater, Tuberculosis, New and Old World Screw worm, Brucellosis, and many others.
Recent events such as the BSE epidemic in Europe and the outbreaks of Nipah virus in Malaysia shown that even “unclassified” diseases can have severe economic or trading implications, especially when there is a link to public health.
On taking office in January 1994, the new Director-General of FAO decided that the Organisation should be better focused in championing the goal of enhanced world food security and the fight against transboundary animal diseases and plant pests as outbreaks of such diseases or pests can result in food shortages, destabilise markets and trigger trade measures. A new programme with two sub-components was established: one to combat plant pests and diseases, and one to fight livestock diseases. These programmemes fell under the umbrella of EMPRES - Emergency Prevention Systems for transboundary diseases of animals and diseases and pests of plants.
This put livestock diseases something of a different light: transboundary diseases were now a specific target, and they are defined thus:
“Those diseases that are of significant economic, trade and/or food security importance for a considerable number of countries; which can easily spread to other countries and reach epidemic proportions; and where control/management, including exclusion, requires co-operation between several countries”.
EMPRES has classified transboundary animal diseases into three flexible categories. These are:
Epidemic diseases of strategic importance, namely rinderpest, foot-and-mouth disease and contagious bovine pleuropneumonia (CBPP) - these are accorded top priority by EMPRES at the global level. However, regions or countries can have a country-/region-specific set of strategic diseases, as well.
Diseases requiring tactical attention at the international/regional level, e.g. Rift valley fever, lumpy skin disease, Peste des Petits Ruminants (PPR), Newcastle disease, African swine fever (ASF) and classical swine fever
Emerging or evolving diseases, e.g. BSE, porcine reproductive and respiratory syndrome (PRRS)
The diseases in the first two groups - diseases of strategic and tactical significance - have a particular danger in that their occurrence can evolve into epidemics which may threaten populations in a region, and have dire potential consequences in terms of international trade. When is a disease occurrence an epidemic? This is notoriously difficult to define. One definition given is “The occurrence in a community or region of cases of an illness, specific health-related behaviour, or other health-related events clearly in excess of normal expectancy”. (J.M. Last)
Where a disease is unknown in an area or has been absent for a long time, only one or two cases may qualify as an epidemic and warrant immediate attention. Where a disease has been present at a fairly constant prevalence level for some time, a marked upswing in the number of cases seen may signal a change in status from endemic to epidemic and will require investigation.
Thus the primary role of surveillance is to detect these changes in status early enough to take action. It means having the ability to detect a new incursion, or changes in present status, and presents a challenge to veterinary services in countries around the world. Renewed attention is being given to Transboundary Animal Diseases (TADs), many of which have their greatest impact in those very countries where surveillance (for many reasons) may be weakest.
The key to success in handling animal disease epidemics is early detection. If a disease can be detected very early in the phase of epidemic development, the possibility exists that it can be arrested and eliminated before it actually inflicts damage. Early detection presupposes that there is a surveillance system in place that will bring infection to light when it is first seen. The country's veterinary authorities are then placed in the position of being able to manage the problem before it becomes uncontrollable, thus protecting the local livestock industry and ensuring food security for those closely dependent upon livestock.
That is why this manual is all about surveillance. Early detection enables early warning and an early reaction. Surveillance is the primary key to effective disease management.
For more information on early reaction, the reader is referred to the FAO publication, “Manual on the preparation of national animal disease emergency preparedness plans.”
Surveillance has as its main purpose, early detection of disease. The sooner a disease is found before it makes progress along the epidemic curve, the better. The developing world is full of examples of countries with devastated livestock agriculture and severe economic losses incurred as a result of having found out too late. When the perceived threat of livestock epidemics recedes into the background, and there are spending cuts to be made, official Veterinary Services are usually the first to suffer, with a concomitant loss of ability to detect disease.
Thankfully, there are also examples of countries that did detect the very first outbreaks of disease, and were able to mobilise forces to neutralise them before they spread. It is much easier to tackle a disease problem in a small corner of a country where it is only necessary to deal with a small animal population, than to get to grips with a developing epidemic that is spreading on many fronts.
So much for surveillance and early detection. Surveillance has other roles, as well. One of these is monitoring the spread of a disease in order to manage it effectively. Knowing how fast a disease is spreading, in which directions it is going and the size of the populations threatened are all key factors in resource mobilisation. One needs to know how much vaccine to purchase, how many staff to deploy and where they should be deployed, the length of the cold chain that will be involved, and so on. Even when a disease is not present, but is the subject of regular vaccination campaigns (as in buffer zones), good surveillance will give a good idea of where to vaccinate and how many doses of vaccine to take along.
Surveillance plays an important role in the monitoring of progress in control and eradication programmemes. It is important to have in idea of whether the programme is successful (in other words, whether disease incidence is being reduced) in order to assess the efficacy of the control mechanisms being used. In this sense, surveillance becomes even more crucial during the eradication phases of the OIE pathways for various diseases. In these phases, it becomes necessary to prove the absence of a disease rather than to detect its presence: here, carefully planned surveillance actions are of the utmost importance.
The word “surveillance” has been used by epidemiologists for some considerable length of time, often interchangeably with “monitoring”, and it is only recently that serious thought has been given to defining the two words.
Surveillance may be thought of as having a broad definition, in the sense of watching a population closely in order to see if a disease makes an incursion. The object of surveillance is early detection of disease. For the purposes of this manual, a definition of surveillance could then be given as:
“All regular activities aimed at ascertaining the health status of a given population with the aim of early detection and control of animal diseases of importance to national economies, food security and trade”.
Monitoring, on the other hand is a more specific activity/ies that will follow as part of an early reaction should surveillance activities indicate introduction of disease. It will focus more specifically on the identified disease in order to ascertain changes in prevalence level, rate and direction of spread. Monitoring can thus be defined as:
“All activities aimed at detecting changes in the epidemiological parameters of a specified disease”.
It should be pointed out that many of the techniques used to implement monitoring can be used in surveillance, and vice versa - and in fact, in practice, the distinction between the two often becomes blurred. As this is a practical book, the blurring will be noticeable in the text that follows. Readers will find much that is value in these pages, no matter whether they are anxiously waiting for a disease that they hope will never appear, or nervously following the progress of a disease they wish hadn't broken out. The distinction is more in the objectives than in the techniques applied.
Surveillance efforts, although as all-encompassing as possible are by their nature, are often not planned to be aiming at a particular confidence level in their execution, whereas monitoring is usually mathematically planned and aims to follow disease dynamics with a certain measure of precision. Readers should be aware that doors are open for bias and error in both monitoring and surveillance, and should consult reputable textbooks on epidemiology to ascertain sources of error and bias, and how these are best counteracted. Mention will be made of many of the pitfalls of the different activities in the text. As mentioned earlier, this is a practical “how to” manual on information gathering and management, and the reader will want to make use of other texts for a fuller background on many of the concepts introduced here.
Although perhaps not always mathematically precise, disease surveillance is not a haphazard action, but a meticulously planned and managed activity. It involves the deployment of personnel who will be moving in the field in a carefully programmed manner, using various methodologies to detect signs of livestock disease. This manual assumes that since the control of disease epidemics is in the greater public interest, it will be official veterinary staff who are involved in doing the work of surveillance for those diseases. Obviously, not all governments will have large armies of veterinary personnel; staffing levels will vary according to budgetary constraints and the importance of the livestock industry; this manual endeavours to deal with such differences.
National governments must realise that animal disease surveillance is a key function of their national veterinary services. Once that is an accepted principle, there is a need to agree on the need for a property structured and administered system. Having a properly implemented and utilised system will require co-operation from a number of stakeholders - the official veterinary service (management and field staff), other extension staff, private veterinarians, farmers and other organisations that might be operating on the ground, for example NGOs. Assuming at this point that it is the official veterinary service that is initiating the drive, it will have to make contact with these stakeholders, explain the intention, and enlist their support. It will also require transparency, as those who are involved in data supply and collection will want to see that the information they supply is actually put to use.
In developing countries, properly supervised sub-professional groups (veterinary assistants, auxiliaries and community animal health workers and the like) are often important elements in surveillance systems, and must be singled out for special training.
The next item on the agenda would be to agree on the objectives of such a system. A specimen set of objectives might be:
What would the priorities of such a system be? The first would be to identify the most important diseases in the country. Field staff (and farmers, for that matter) would need to be familiarised with these diseases, and at least know the basics about recognising them in order to be able to report their presence. There is obviously nothing wrong in reporting any disease through the system, no matter how insignificant it may seem; the problem comes in allocating resources for staff training - this is where decisions will have to be made about priority diseases. Time and training materials will have to be devoted to priority diseases.
Having identified priority diseases, the next step is to identify priority areas. It will probably also not be possible to direct resources equally throughout the country, and the areas where the identified diseases pose the greatest problem will have to be earmarked for the first resources in terms of training, staff, intensity of surveillance, etc.
It is likely that priorities will differ in different parts of a country, and that must be taken into account. While CBPP may be of importance in one province, trypanosomosis may be of greater importance in another. Planning of training and awareness creation will have to take such differences into account.
Beyond a country's borders, it may be that other diseases enjoy priority. In each region of the world, a different set of circumstances, with its own set of livestock diseases, exists. Being a responsible global citizen will require a country's veterinary service to be aware of regional priorities in terms if disease surveillance, and internal surveillance systems will also need to take account of these. Regional economic and political groupings often play a leading role in drawing up regional priority lists of animal diseases.
A final criterion in developing a priority list for surveillance is that of international disease reporting. Every country that is an OIE member will need to decide what OIE List A and B diseases are of local importance, how to detect these diseases, and how to report on them. Transboundary Diseases as defined by the FAO will also play a role in shaping surveillance systems. Defining priorities is therefore not purely a matter of local interest, but also a matter of sensitivity to international concerns. The greater the attention given to diseases that are transboundary in nature, the greater the opening for participation in international trade and economic advancement.
Targets must also be defined, as well as responsibilities. In the various priority areas, decisions will need to be made concerning staff development levels, and time-frames set. What will the frequency of surveillance inspections be in area? When will they start? If there is to be sero-surveillance, when will the first survey take place? Will there be regular serosurveys? Who will be responsible for each activity? Will computers be used? Do questionnaires need to be drawn up?
A proper management plan, defining priorities, targets, responsibilities, resource allocation and responsibilities must be drawn up and adhered to. The various elements that should be included in such a plan will be made clearer in the pages of this manual.
In drawing up a management plan, the normal information flow in a surveillance system will need to be taken into account, and every step in the flow properly monitored and controlled.
Information flow in Surveillance and
Livestock Disease Management
If information is to be collected, it will have to be managed, and - most important - used. Data collection with no clear purpose is a waste of time and resources. One thing that data managers will have to define very clearly, is what incident will trigger what action, and at what level. The data that should be collected is that data which will lead to action.
For example, the first-time suspicion of a Transboundary Disease in an area will require immediate action by the local veterinarian - perhaps in terms of quarantine or movement control, and certainly in terms of follow-up. At management level, it may require further decisions in terms of publicity and redeployment of resources.
Exactly what action is taken by who, and under what circumstances, must be spelled out as a part of the surveillance system. It will obviously differ from country to country, and even from region to region within countries, and this type of planning rests firmly in the domestic arena. But it is planning that must be done.
It was mentioned earlier that surveillance is a planned activity, involving a carefully laidout routine. The type and level of activity, the routine (frequency of activity in any area) are all determined by the country's veterinary management, in agreement with the various stakeholders, according to the criteria listed above.
Suffice it to say that whether government staff are combing an area in large numbers looking for a disease themselves, or whether a small team is conducting retrospective farmer interviews, the activity must be a planned activity that is worked into the annual activity plan.
A little extra time and resources spent on surveillance at the beginning will save a lot of time and money later on should a disease break out. Surveillance planning will include such things as deciding on how many staff members will be devoted to surveillance, the frequency of visits, distances to be travelled and transport requirements. Planning must be such that activities are evenly spread, not that one day is overburdened with work (creating the necessity for a “rush job”) while another day's activities are very light. This kind of information is necessary not only for execution of the activity itself, but also for budgetary planning. For an example, see below:
|Example of work programme|
Month: June Week: 1st week Name: Tobias Malinga Duty station: Kwela
|Date||Village(s)||No.inspection points||Estimated km travel||Cost ($)|
Having a fixed work plan means that the staff member knows exactly where he needs to be and when he needs to be there; and should supervisory staff need to monitor what he is doing, they know where to find him. Should the staff member need to be contacted in the field in an emergency, his whereabouts will be known. If the person is using a vehicle, the distance estimates in the work plan will also serve as a guide to indicate how closely the plan was adhered to.
Visits to the field must be planned in such a way that a single round trip will cover the biggest area possible. Surveillance teams (or individuals) should have areas demarcated for them, and should not overlap. Wherever possible, the same staff should be visiting the same areas every year - this enables the livestock owners to get to know the staff and build up trust in them, and enables the staff to get to know the area and the movements and habits of the people more intimately. Regular surveillance by the same staff member (preferably of the same ethnic group as the farmers being visited) builds mutual confidence and means that farmers will report diseases more readily.
Surveillance does not simply mean that a staff member “does the rounds” inspecting livestock or taking blood samples: it requires that the surveillance officer be involved in a constant process of informing and educating his farmers so that they will immediately recognise telltale disease signs and report their presence without delay. Extension activities could be incorporated into the surveillance plan, and certainly the official should be carrying a supply of information leaflets and/or posters giving details of diseases of local importance.
Surveillance: local needs vs national needs
The battle cry is usually to “keep surveillance focussed”. Having a narrow focus for disease control activities will, however, often backfire. Diseases of importance to the national veterinary service are often not of importance at the grassroots level. Low profile rinderpest will almost certainly be of lesser importance to a rural community than concurrent endemic CBPP; ongoing surveillance for FMD in an area where the disease is normally absent (but perhaps occurs only every five or six years) will not gain a community's sympathy. Forcing an approach onto a community won't work, and will only cause resentment. Local communities need to feel that their needs are being catered for, and in order to gain co-operation, compromises will need to be made.
Shaping such compromises means listening to farmers, and taking other disease needs into account. It will certainly mean carrying out surveillance for a package of diseases rather than a single one. The logic for this concept is both economic and practical. The system put into place for surveillance of the critical target disease can easily capture data on other diseases in the same area. The sustainability of the system will be better secured as it will be viewed more positively at both the local and the national level. Furthermore, it should be remembered that even during a disease eradication process, there is a requirement for the surveillance system to demonstrate its capacity for those that are relevant for differential diagnosis.
Nevertheless within a broad-based national surveillance system, there will be a need from time to time for specific targeted surveillance programmes aimed at either specific and special awareness/early warning or in aid of special eradication programmes.
While targeting a broad range of diseases makes sense from the managemental and economic points of view, there is a lot to be said for surveillance "riding on the back" of a particular disease. Very often, it is a single disease (such as rinderpest or Foot-and-mouth disease) which causes a national scare and provides the impetus for putting a national surveillance network in place. Once resources are in place, moving from an emergency response to a proper surveillance system is only a very small step. Moving to wide-spectrum surveillance must ensure that impact scenarios can be constructed. Surveillance also means having the information necessary to predict what the consequences of other disease epidemics might be. Having secured the resources necessary to put a surveillance network in place, the same network must ensure its survival by justifying its economic importance on the national scene.
The same kind of reasoning applies to extension work. If surveillance personnel are equipped for extension (which they should be), then their extension materials should take account of local needs. Training must be provided on the recognition of a number of diseases, so that the network is "sensitive" to a varied repertoire of needs.
This is especially so where a network relies heavily on the reporting by farmers of disease outbreaks. If farmers feel confident about reporting diseases with which they are familiar, they will also feel easier about reporting occurrences that are foreign. The benefits work two ways: farmers feel that something is being done to assist them with their own identified needs, while the surveillance system will have a greater chance of detecting "exotic" diseases.
The flexibility to cope with ad-hoc needs
Much has been said above about having a planned surveillance schedule that runs for year after year. A routine programme is important, but it must not exclude ad-hoc needs. It should be possible to re-deploy staff to take care of new needs - for example, a need for heightened surveillance in a particular part of the country for a short while to guard against a threatening disease incursion. Reviewing surveillance programmes and their efficacy-and, if necessary, changing them to suit population shifts or disease patterns, is an important part of surveillance management.
Having planned a surveillance programme, what types of surveillance could be undertaken? Probably the most popular - and one of the easiest and cheapest - is visual surveillance. Two kinds of visual surveillance can be distinguished: direct and indirect. For both types, the veterinary official visits all (or as closely as possible to all) farmers in the area allocated to him/her. This is less of a random sampling exercise and more of a census. It is carried out on a regular basis, any number of times per year, depending on staffing, budgetary constraints and the prevailing disease situation.
In direct visual surveillance, the observer physically inspects all animals and records what he sees in terms of disease. Probably the best way of doing this with cattle is to first observe them from a fairly close distance and run them slowly through a race/crush if one is available. Individual animals can be selected for closer examination should they appear unhealthy. Small ruminants are probably best looked at from as close a distance as possible in a pen. Again, those appearing sick can be singled out for closer examination. The official's findings are recorded on a pre-printed data questionnaire. It may be possible for the officer to take samples for laboratory confirmation, or even to perform a post-mortem, depending on his/her mobility and other logistics.
This approach assumes that the veterinary official involved - who is usually a lay person - has had fairly comprehensive training in recognising diseases of local importance, as well as diseases of national or international importance which might not occur locally on a regular basis.
With indirect surveillance, it is assumed that the livestock inspection frequency is very low, and therefore, while the official visits all possible livestock owners, he/she relies on their recall to gain details of diseases that affected their flocks and/or herds. While some may scoff at this kind of surveillance, it is important to note that most herders have an accurate knowledge of local diseases, and are usually fairly close to target when it comes to making diagnoses. If they aren't always all that accurate clinically (usually because there may be different diseases in an area with similar clinical pictures) there will be certain syndromes which usually have fairly descriptive names in the local vernacular. These names at least - and the approximate numbers of animals which were affected - can be recorded for further investigation if necessary. Usually, by closely questioning farmers about the clinical signs seen, it is also possible to arrive at a tentative diagnosis. Again, it is necessary for the veterinary staff to have a good idea of clinical signs and post mortem lesions to be able get all possible information out of the farmer.
It is important when questioning a farmer not to ask leading questions, or to use a questionnaire with “pre-cooked” clinical signs listed on it (if at all possible). Herders might be led to believe that this is a quiz where “yes” is the right answer, and may well report disease signs that were never seen.
In practice, of course, the two types of visual surveillance are never really separated. Usually, the veterinary staff member will inspect all the animals present, record what he sees, and then ask the farmer about all diseases that were seen since the previous inspection. All observations - both direct and indirect, are then recorded.
Sero-surveillance, while it may give a more “objective” view of the disease situation in an area (it measures antibodies rather than a lay person's conception of a disease), has its limitations. Among them are:
Nevertheless, sero-surveillance is still more “objective” than visual surveillance, and for many diseases it can be used to get an idea of current prevalence and geographic distribution. The costs and other problems involved should not be a discouragement to use sero-surveillance. A well-planned random survey will carry benefits that should far outweigh its costs. In the case of rinderpest, sero-surveillance can be used for antigen detection (for investigating wild virus activity or proving absence of infection), while seromonitoring is used for post-vaccination antibody detection.
Planning a serological survey
First, the objective and the tests to be used must be clearly stated. If surveillance is for Rinderpest, what is being sought? Evidence of exposure to disease, or evidence of vaccination? Where? In what species? Is this part of an eradication process? If so, is it only going to be a zonal process or a national one? Once the objectives have been decided and the area/s to be surveyed have been delineated, other decisions have to be made.
What will the sampling units be? The primary sampling units will, in most cases, be herds. Herds, in developing countries, need definition, and these definitions will vary by geography, ethnic group and farming system. It might be a group of animals cared for by an individual, or it might be a group of animals owned by a collection of individuals. The secondary sampling units will then be individual animals. If the individual is important (eg. when trying to link clinical signs to antigen presence or an antibody titre), then identification of the individual becomes important - if not, one needs not be too concerned with secondary sampling units.
Once the type of sampling unit is known, and the survey area is defined, the next step is to draw up a sampling frame from which the psu's will be chosen. The sampling frame would be, for example, a list of all villages in the survey area, together with the livestock populations of each village.
The primary sampling units are then chosen by random sampling. Procedure for random sampling will not be discussed here (but a short summary is given in an appendix), as they are described in epidemiology textbooks, as well as FAO and OIE publications (eg. “Recommended surveillance procedures for disease and serological surveillance as part of the Global Rinderpest Eradication Programme (GREP)” IAEA/FAO, 1994). What is important is to determine the confidence limits (usually 95%) and the prevalence level which is desirable to detect beforehand, as this will, in turn, determine the size of the sample to be taken. How to calculate sample sizes is covered in many authoritative texts; it will not be given here.
Nomadic herds present a special problem. Very often, instead of doing a random village selection, one can do random selection of grid blocks from a map, enter the areas selected, and then sample a percentage of the animals found inside the block. This technique has been well described elsewhere.
Once the area to be covered, the size of the samples and the whereabouts of the herds to be sampled have all been worked out, the next (and more down-to-earth) stage of planning is reached. This will involve the following:
Something often forgotten by field staff planners of serosurveys is to make all necessary arrangements with the laboratory analysing the samples. The laboratory should have the technical ability to perform the tests required, and should be informed of the arrival of the samples well in advance in case extra resources need to be reserved for the job.
Most ordinary surveillance routinely carried out falls into the category of passive surveillance. In this case, there are routine programmes that run - usually partly directly visual, or indirect, relying on farmer interviews and notification - basically to survey the landscape for livestock diseases and to detect and changes in status. This is probably the most important, and is a key element in early warning. The word “passive” should be seen as a characterisation of technique and not a sign of lowered importance of the work done.
Much has been made of the concepts of “passive” and “active” surveillance. Passive surveillance is usually thought of as regular - and perhaps infrequent - visits to an area by veterinary staff to assess the local animal situation and determine livestock populations. It would include voluntary disease reporting by farmers, traders and perhaps other individuals such as private veterinarians.
“Active” surveillance entails frequent and intensive efforts to establish the presence of disease in an area. Examples:
Any activity which is frequent, intensive and aims at establishing the presence or absence of a specific disease, could be described as “active” surveillance.
Once the presence of a disease is confirmed, and similar techniques are then used to follow trends in its development, this would (at least in terms of current terminology) be called “monitoring”.
There is no doubt that active surveillance activities can be expensive and time-consuming. There are benefits, however, that in the long run will outweigh the costs. In the first instance, beginning active surveillance (at least for diseases such as rinderpest and CBPP) means that vaccination has ceased, and huge amounts spent on blanket vaccination campaigns will be saved. Secondly, there are trade benefits to be gained - eventual proof of disease absence will allow the opening-up of hitherto untapped markets.
Abattoirs and slaughter slabs
These are a valuable source of data, particularly when it comes to diseases which present laboratory diagnostic difficulties, such as CBPP. It has rightly been said that abattoirs are “the post mortem halls of the nation” and are a goldmine of information. Export abattoirs are always closely monitored by veterinary officials, and instituting a reporting system to extract information from them would present no problem. Smaller abattoirs and slaughter-slabs present something of a problem in that there may not be enough official staff to keep them under surveillance. Possible solutions to this would include the following:
The kinds of information required here would be such things as species, origin of animals slaughtered, lesions seen, condition suspected. During the eradication phases of a disease pathway, surveillance would have to be organised such that:
* it was random
* it covered the required number of animals at a 95% confidence level to give reasonable assurance of the absence of disease.
Rapid Appraisals go by many names (Rapid Rural Appraisal, Participatory Rural Appraisal, Particpatory Epidemiology, Sondeo Method, etc). The basic idea of the Rapid Appraisal is collect information in data-sparse areas, places that are marginalised, remote, inhabited by nomads, inhospitable, infrastructure-poor - whatever adjectives one might choose for neglected areas, whatever the reason for the neglect might be.
Rapid Appraisals provide a means for data collection which would otherwise not be there: the data may be limited in scope, inaccurate or biased - but it is data, in certain vast areas of the developing world where information is virtually non-existent, any data are better than none. Such data may not be statistically valid, but will at least provide an indication of what is happening “out there” - and where it is happening. It is based on rural livestock owners' impressions, but their impressions are often very accurate.
Many possible “tools” are available for appraisal work, but the end result is usually an unstructured/semi-structured interview during which the interviewer tries to capture some form of data, at least in a semi-structured form. The essential ingredients are (a) to know exactly what information is required, (b) how to capture it fairly informally, and (c) how to structure it so that it can finally be computerised (if necessary) and analysed.
In summary, Rapid Appraisals make use of the following tools to gather information:
When using appraisal techniques to gather disease data, one must bear in mind the various sources of error and bias:
Despite the fact that Rapid Appraisals are fraught with problems that amount to a statistician's nightmare, they are essential in areas where staff are thin on the ground and frequent and intensive surveillance is not possible. They provide useful information on the disease situation in the area, the farming systems, the distribution of people and livestock, and on trading routes - where they run, how and when they are used. Trading routes are notoriously difficult to document, and Rapid Appraisal tools provide a means of recording these routes (which need updating from time to time).
Another important use of Appraisals relates to the provision of baseline economic data for the calculation of losses due to disease impact. A number of appraisals, using interviews combined with direct observation in an area, will give an idea of livestock production. Cash values can later be imputed to these estimated production parameters in order to calculate the economic cost of a disease epidemic should it strike the area.
The role of Laboratories in surveillance programmes
A final word here about the role of laboratories in veterinary surveillance. Diagnoses (or tentative diagnoses) can be made in many ways:
Often, laboratories are thought of as having relevance only in serological surveys. However, it is imperative that laboratory backup be obtained for as many diagnoses as possible. It is essential in cases where an epidemic disease is suspected for the first time in an area that samples be taken for confirmation of the diagnosis. Where the diagnosis is uncertain, repeated follow-ups, with laboratory sampling, must be made in an effort to either confirm or exclude the disease. Where it is known that a certain disease has become endemic, confirmation of each individual focus becomes unnecessary, but 10-20% of cases must always be confirmed to ensure that the epidemiological picture has not changed and that another disease, different in epidemiology but similar in appearance, has entered.
For these reasons, the presence of a strong laboratory diagnostic service, subservient to the official veterinary service, and answerable to it, is absolutely essential. Laboratory services are an essential backup to what is being done by field service staff.
It is also necessary - and this is often forgotten - that field staff (including veterinarians) be regularly briefed on the kind of samples needed for the various diseases threatening in an area, and that they are also familiar with the requirements for preserving, packaging and transporting such samples.
Maintaining laboratory norms
Laboratory testing needs standardisation so that, for example, the results of serosurveys analysed by different labs are comparable. This means that laboratories need to belong to networks where the same reagents and methods are used in the same test; where experience and expertise is shared; and where use is made of reference laboratories. A chain of OIE and FAO reference laboratories has been established for this purpose. The FAO/IAEA Joint Division has also been established to assist with standardisation of tests, and for quality assurance. It is imperative that national veterinary laboratories make use of these services.
Surveillance often presents itself as a thorny issue in developing countries because it is seen as a costly operation necessitating an enormous army of surveillance personnel on the ground. This need not be so, and judicious deployment of resources can often achieve what is needed without great expense.
Critical point identification
The first step in setting up a “low cost” surveillance system involves identifying critical points or critical surveillance areas. These would include:
The bulk of veterinary resources should then be deployed at these critical points, with high frequency surveillance designed to move staff amongst such points with relative rapidity for whatever type of surveillance is deemed appropriate - visual (for detecting clinical/pathological signs), detection of antibody, and detection of the causative agent. Such surveillance is fairly structured but not sufficiently randomised for movement along an OIE pathway. It would, however, qualify a country for entrance to a pathway if it gained sufficient evidence of clinical disease absence - the point is that the work would have to be restructured along more “scientific” lines in order to move further along an OIE pathway.
As an aside, it must be mentioned that strategic resource deployment may clash with equity goals - politicians may want to see a more “even” distribution of resources across the country. Careful explanation will have to be given for unequal resource deployment, with the assurance that once the disease problem is cleared up, personnel and equipment will once again be redistributed.
The frequency of surveillance at these critical points is a matter of common sense and would have to be determined by the perceived risk of each point, with the higher risk points receiving the most frequent attention. Frequency of surveillance will, on the one hand, be determined by the frequency of population turnover (eg. along trade routes) and by the incubation period of the main disease feared at the time. In a relatively static livestock population at high risk to foot-and-mouth disease, for example, it would not make much sense for surveillance to be more frequent than fortnightly. On the other hand, financial constraints will also be a major determinant of frequency. Surveillance needs to be an intelligent trade-off between field realities and budgetary limitations.
All other parts of the country would be deemed to be non-critical areas where surveillance could consist of relatively infrequent visits by field personnel (perhaps once or twice per year) or annual Rapid Appraisals relying heavily on group interviews. Useful information can also be gathered from other existing networks - for example NGO workers, crop extension officers who may happen to be in the area, consultants, etc.
An essential item in any surveillance system is farmer awareness. Training local livestock owners in disease recognition and encouraging them to report the presence of any suspicious clinical signs is a very cost-effective means of improving the quality of disease surveillance, both in critical and non-critical areas. There may even be a possibility of a small incentive to be provided for evidence leading to the discovery of a disease - eg. a fee to be paid should a farmer submit part of a diseased lung for CBPP examination.
Data from private veterinarians is an important item not to be forgotten. Capture of this may be via questionnaires sent to them regularly; a legal requirement for them to report certain diseases to the authorities; and by making the use of an official government questionnaire obligatory when sending samples to the laboratory (in this way, data from laboratory submissions will enter the system automatically).
Ultimately, the exact type of surveillance adopted by a country is its own decision, based on disease risk and available resources. What is important is the issue of transparency. It is incumbent on each country to make the precise mechanics of its surveillance system known to neighbours and trading partners. This includes the identification of critical areas and non-critical areas, and the types of surveillance operational in each. Such transparency builds confidence, facilitates mutual risk analysis, and in the long run, will promote investment and trade.
Much will be said about information systems and information management in this manual - for a very simple reason:
once information has been gathered, something has to be done with it.
Three things must happen to information: firstly, it must be managed, controlled and quality-checked; secondly, it must be analysed in order to become more understandable, and thirdly, it must be acted upon. These three points will be emphasised repeatedly throughout this manual. For information to be an analysable and eventually useful for decision-making, it needs careful management and quality control.
What is an information system?
An Information System is the collection of data, people, procedures, hardware, software, files, and information required to accomplish an organised set of functions.
Basically, the constituents of an information system are:
A storage/retrieval/analysis system - which in this day and age would usually consist of
A feedback delivery system
An information system is nothing more than a large communication cycle, involving information transmission and reception. In any communication cycle, one person (the communicator) transmits information via a medium (the spoken word, the written word) to a recipient. In order to ensure that the concept transferred to the recipient's mind is what was originally in the communicator's understanding, and also to motivate the communicator's further participation in the interaction, there has to be feedback. During feedback, the recipient will communicate with the original transmitter of information in order to receive clarification, or to take action.
The diagram below illustrates the essential elements of the communication cycle.
Elements of the Communication Cycle
A breakdown in any one of these elements will cause the breakdown of the entire cycle and communication will sooner or later come to a halt. This should be borne in mind when planning, and later when maintaining, the flow of information through a system. There needs to be a clearly defined flow of information with defined inputs, outputs and feedback. Suppliers of information need to know that the products of their labour are being put to good use, or their supply will dry up.
Structure and flow of veterinary information systems will be dealt with in greater detail later, suffice it to say at this stage that information is usually garnered from farmers by some kind of field worker, who will pass it up a chain to a central computer system, from where feedback will be transmitted back to the field.
Managing the system
Provision must be made for the position of system manager to run the system. Such a person will have to carry out the following functions:
Monitor data flow into the system - and take follow-up action when flow slows down, or when inflow becomes abnormally large. Both could indicate abnormalities in the system which need to be remedied. A “drying-up” of data coming in could indicate a lack of motivation of field staff for a number of reasons (including poor feedback), or logistical problems among data gathering staff in the field. A “surge” of flow could mean that there is an upswing in disease incidence, or livestock population, or rampant data fabrication, or that the system had been functioning suboptimally before the surge, and it is now coming to an equilibrium.
Check data quality - various parameters need checking. They vary from simple spellings to handwriting, looking for fabricated data, checking internal logic (for example, when a veterinarian reports zero deaths in an area, but in the same report describes post-mortem lesions). This might also mean cross-checking with data from other sources, providing direct feedback to suppliers of data and requesting them to verify what has already been submitted. The manager must be motivated enough - and have enough “clout” in the hierarchy - to take action when deficiencies come to light.
Carrying out data analysis and ensuring that analysed information reaches decision-makers. This is an ongoing function, and disease data will need ongoing temporal and spatial analysis to look for trends in terms of increasing or decreasing incidence, spread of disease, etc. Vaccination data will need following-up to assess percentage coverage in vaccination campaigns. Again, the manager must not just be an observer. Disturbing trends demand remedial action, and no manager can afford to be a spectator.
Ensuring feedback to the field. Field staff need to receive reports indicating the outcomes of trends in data, and what action is being taken to respond to these trends. This will encourage them to continue with their work, as they will see that what they do is constructive and helpful. It will also give them the opportunity to respond, and voice their opinions as to whether analyses made are reasonable, and whether actions taken are justifiable. In short, it gives them a stake in the system. At the same time, feedback needs to be given to livestock owners to encourage their further co-operation with field staff. This could take the form of information dissemination through pamphlets, posters or farmers' days. Once again, it effectively gives them a stake in the system, as well. It is the job of the information manager to ensure that such feedback is made, and that it reaches grassroot levels.
At country level, it is the national epidemiologist who is best placed to deal with overall information management. In large countries, some of his functions could be delegated to lower levels, but the overall responsibility for maintaining the integrity of the system should obviously - for simple reasons of accountability - be in the hands of one person.
Performance indicators - measuring efficacy of surveillance
Various parameters can be used to measure the efficacy of surveillance. These should be agreed upon during system formulation, and reviewed from time to time.
Such parameters should be monitored by the epidemiologist/information manager on a regular basis. Following trends every month is probably the best, and the epidemiologist should work out a simple monthly schedule to cater for routine activities such as data checking, reporting and analysis, and monitoring performance indicators. Examples of indicators that could be used are
* number of reports submitted/1000 head of livestock/district/month
* number of individual livestock inspections/staff member/month
* percentage of observed disease incidents for which laboratory samples were submitted
* percentage of suspected cases actually confirmed for any particular disease
* time lag from sample submission to final laboratory diagnosis
The Joint FAO/IAEA Division and EMPRES has proposed guidelines for use in Rinderpest surveillance for the GREP; these guidelines could be adapted to many situations. The information manager will need to establish the set of performance indicators best suited to his situation, and ensure that these are acted upon when the situation so demands.
To some extent, the outputs derived from an information system will determine what sort of inputs are required. In other words, when designing an information system, the first thing that must happen is that the system's future users must discuss their needs with a system designer. What sort of outputs are needed? To quote a trivial example, if we would like to know each month which diseases occurred, which species were affected, and how many animals were affected, it presupposes that the basic data will contain details such as the animal species affected, how many affected, the reporting date and the name of the disease suspected.
A good starting point would be to look at any existing manual system (often regular “monthly returns” that come to a supervisor and gather dust), and ascertain the inputs and outputs of that system. The new system in the process of conception might be able to build, at least to some extent, on what is already established. It will also show very clearly what kind of information is not used, and what could be improved upon to make it more useful.
Practicalities are important here. A natural tendency, while an information system is being designed, is to add all sorts of data requirements without any clear idea of whether they will be needed, and if they are ever needed, exactly how they will be used. Input requirements should be kept to an absolute minimum, remembering that the more complex the requirements, the greater the chances of error entering the system. The watchword is, “when in doubt, chuck it out”.
For visual or clinical surveillance, for example, the following data items for each observed outbreak would be sufficient:
|No. of cases*||No. of deaths*|
|No. of animals at risk||No. of animals examined|
|History and/or clinical signs||Clinical examination/field tests|
|Tentative diagnosis*||Age category most affected|
|Any treatment given||Sex category most affected|
|Any post-mortem lesions seen||Any samples to a laboratory|
|Laboratory Results if available||Farming system|
|Name of reporting officer|
|Category of reporting officer|
* indicates minimum acceptable information
For sero-surveillance inputs (aggregation of these data at village/district level is satisfactory for a national database; information about each individual animal would be unnecessary) the following are reasonable guidelines:
|Locality and date||Species|
|Farming system||Disease history of herd/flock|
|Vaccination history (past yr)||Age/sex category/ies bled|
|Vacutainer serial numbers||Disease being monitored|
|No. positive||No. suspicious|
Data requirements from abattoirs or slaughter slabs should include the following:
|No. of animals in consignment (where applicable)|
|Origin of animal/s (where known)|
|Lesions seen||Condition diagnosed|
|Age most affected||Sex most affected|
|Samples sent to lab||Laboratory results if available|
|Category of reporting officer|
Being practical about data recording is important. The inputs required will determine the exact variables being recorded, and with increasing database complexity, management and analysis become increasingly difficult. This is particularly so where the database is a manual one - ie. not computerised, but it applies for computer-based systems as well. It will also influence the design and complexity of the questionnaire to be used. Again, the question will arise, what exactly is to happen to the variables being recorded? What will this information be used for? By whom? When? Is it really necessary?
“Information for the record” which will be used for analysis and as a basis for decision-making regarding disease control strategy must be distinguished from “information for immediate action” which will be communicated to the nearest veterinary officer for follow-up. For example, when recording data in a visual surveillance system, if disease was introduced into the area, there is little point in trying to use a computer database to record where the introduced animals came from, and what their destination was after they had infected animals already in the area. Apart from the fact that such information is notoriously difficult to computerise (and difficult to extract from a computer and even more difficult to use later), it is not that necessary for posterity. It is information that must be dealt with in a different way, as it obviously demands action in the form of following up the animal movement, finding the culprit animals and destroying or quarantining them - in other words, recent livestock movements must be reported to, and followed up by, the local veterinarian. This can then be dealt with in a narrative from the local veterinarian once he has finally dealt with the issue.
The matter of livestock movement in general is something that is tackled via a Rapid Appraisal rather than by a questionnaire specifically destined for a database, and which can be mapped rather than left to rot in a database.
Exactly how data are to be recorded and where is also an issue that needs decisions. Certain information may be necessary, but where and how is it to be recorded? Trying to put everything into a computer just because computers are fairly easy to come by, does not make sense, and such issues are a valid part of system creation.
A small aside is perhaps necessary at this point. While potential stakeholders in a computer system may be well-meaning, and have very genuine information needs, care must be taken to ensure that the system remains simple, with a direct information flow, a manageable number of inputs, and simple, clear outputs. The system designer must take a strong lead in the initial planning and be prepared to veto ideas that may detract from the efficiency of the system. The old adage about “the camel being a horse that was designed by a committee” can certainly be true when applied to information systems that are designed to accommodate too many “needs”!
Detail has been given above as to what sort of variables should be recorded. In large hierachical information systems, it should be remembered that such detail is probably only necessary at the lowest level of input and analysis. In a very large country, with a livestock population of many millions, it is unlikely that the national epidemiologist will need to know the exact clinical signs observed in a specific village, or the name of the diseased animals' owner. In a similar vein, when international disease reporting is undertaken to regional or global bodies, this kind of detail is also unnecessary. At higher levels of a system, basic information needs will be limited to such variables as the locality where the observation was made, the numbers and species of animals affected, the diagnosis made, and whether and how it was confirmed. These considerations must be borne in mind where large systems are planned.
What is important at national level, is that animal disease data not be “buried” and forgotten in common national records, but be passed to national epidemiologists for early analysis and further action, even if action has already been taken in the field. Such data must be readily accessible to veterinary management. Data should always be looked at when “fresh”!
As mentioned above, computers have become smaller, cheaper, user-friendly, more readily available and more robust. They can be taken into all sorts of environments and programmes are available that will perform a great variety of functions. But just as horses are more suited to certain terrain than the motor car, so we must remember that computers, for all of their hi-tech functionality, simply can't do everything. Computerisation, put bluntly, is not a panacea.
What it will do, what it won't do
Computers will not replace good personnel. They do not reach far-off stock owners and collect data. Trite though it may seem, having a computer doth not a system make. The system design must, first and foremost, incorporate people and their abilities, and make provision for extensive training in the use, completion and submission of data questionnaires. In the ordinary course of events, input questionnaires should never be completed by farmers and by those unschooled in their use; only by properly trained personnel. The first principle of data recording is, and always will be, “garbage in - garbage out”.
Computers will not check data for you. They will not detect fabricated data, problems with data logic, nor will they improve poor handwriting on data forms. Such work must be done at field level by the most senior staff member available before forms are dispatched to the computer centre for typing-in.
Computers cannot replace the epidemiologist. Computers are simply data storage and retrieval devices. They cannot make judgements on data quality, notice disease trends, contact field veterinarians to ascertain their feelings on the current disease situation, or recommend a course of action. While the foregoing may seem glaringly obvious, it is all too often forgotten by those who want to create the “perfect system” without realising the crippling limitations of computer technology.
Computers are not a short-cut to easy data storage for all types of variables, either. They can be used for storing relatively simple numeric and non-numeric variables, and should only be used for this purpose. The examples given above of computerising livestock movements and storing trade routes are very fitting in the veterinary context. Another “variable” that must not be overlooked is a very simple one - the “gut feeling” of the field veterinarian. The field vet is in touch with farmers and field staff, he knows the conditions of the pasture, the livestock and the local market; he understands the way diseases behave in his area where he may have worked for some years. To computerise this? Impossible - which is why regular personal communication with headquarters is always important.
Computers also do not write interesting reports and do imaginative analyses, nor do they give constant encouragement to field staff. They may store data, but making sense of what is stored is, in the end, a human function. To calculate is computing, to interpret is human.
The centrepiece of any information system is not the computer, or the programmes installed on it: it is the people who run the system, and most particularly the epidemiologist who stands at the centre and directs operations. Without a dedicated, enthusiastic and “wide-awake” epidemiologist, the system will crash as surely as it will when the computer suffers a power failure.
National Computer systems
There is often a tendency to try to over-computerise, which can be fatal. In other words, some would have a computer in every district veterinary office - after all, what could be easier than local data input and simple electronic transfer to the central database. This concept has several major shortcomings:
Who will be responsible for data input? And who will check it before it is transmitted upwards to the central computer? A busy veterinarian will certainly not type in large numbers of data forms reaching his office, and it will be left to a clerk. What of the work load already carried by the clerk? Or will one need to employ an army of data input clerks to cope with each district office? Data input is a very monotonous but highly demanding job that requires high standards of accuracy. Ordinary office clerks are usually unable to cope with such work, and when it is imposed on them, the results are often disastrous - slapdash and shoddy, with a wide variation in data quality between various staff members. In addition, if good data entry clerks are employed, their capabilities are usually such that they could easily cope with work from a number of districts simultaneously.
Who will maintain the system? A huge network of computers strung out across the countryside means regular breakdowns, staff having problems with software, power failures, etc. This implies a number of technicians to keep order, and ensure a smooth flow of information.
More importantly, who will pay? The more computers, the more connections, the greater the cost - the cost of maintaining the large number of machines, regular upgrades, and paying for the connectivity which will, obviously, be supplied by the national 'phone company.
Large numbers of staff with sophisticated, sprawling computer networks certainly have their place in developed countries, but in developing countries the secret is rather to start too small than too big. If one takes a convenient number of 20 to 25 veterinary districts to one computer, then it is quite practical to say that in large countries, a small computer unit at regional/provincial level could handle the inputs from those administrative divisions and than pass the data on to the central unit. In very small countries, a single central data input and processing unit would be sufficient. Admittedly, one would have to trust to national postal system to get data questionnaires to the input unit, but that is often preferable to trusting a strung-out computer network with all of its associated complications. The easiest way to handle this would be for each station to post all of its inputs on a regular basis (fortnightly; monthly) to its computer unit, and to keep duplicates (carbon copies are cheapest) of all questionnaires in reserve.
Managing the flow
When designing an information system, it is important to construct a flow diagram of how data flow in the system is envisaged, and what the various control points will be. This will of great importance in managing the system once it is running.
In general, the flow of data in the case of visual surveillance will be:
Farmer→ Animal Health Worker → Veterinarian → Epidemiology Unit → Analysis & Feedback → Management Decisions
In the case of sero-surveillance, it will be slightly different:
Farmer/animals → Samples & Info → Laboratory → Results → Epidemiology Unit → Analysis & Feedback → Management Decisions
Having determined the route of flow, the next important step is to determine exactly what will happen at each step. Sending data straight from the field to the Epidemiology Unit, and then typing it directly into a computerised database would be foolhardy. Checking and validation mechanisms must be built in along the path to the database.
Taking the case of visual surveillance as an example, a more detailed flow diagram would look like this:
There are a number of “lines of defence” in data protection. The first is the person actually collecting the data. Such people should have clear handwriting, an ability to work with people and extract information from them, and a reputation for honesty.
They should be clearly briefed on what they are doing and why they are doing it. The importance of their work must be emphasised. Obviously, if the person collecting the information is a veterinarian, the foregoing points should not be a problem, however, in many cases the data collection staff will be community animal health workers or veterinary assistants, or similar groups.
The next line of defence is the veterinary district office. Here the form must be carefully checked by the district veterinarian. He is the one most familiar with the district situation, and is closest to the ground. This is the last opportunity to, if necessary, go back to the field and re-check the information before it is transferred to the epidemiology unit. It is also the first opportunity for action. If any suspicion of a transboundary disease is detected and reported, the district veterinarian must act immediately. There is no point in waiting for the data to be analysed by the national epidemiologist before doing anything about a disease outbreak.
Items to be checked by the district veterinarian would include:
Legibility of writing on the questionnaire form. Ensuring readability at this level saves expensive phone calls from data input clerks later on.
Correctness of spelling, particularly of place names. If geographic co-ordinates of places are included, these must also be checked.
Internal logic. If a diagnosis of disease is made, but, for example, zero animals are reported as affected, the field worker has made an error which needs correction.
Information must be realistic. If 100 animals are reported sick in an area where herd sizes average only 40, the farmer has probably exaggerated his problem to the field worker - or the field worker has not visited the farmer and has deliberately fabricated the data.
Data encoding. If on-form encoding is used (see later), the codes used must be checked for correctness.
Only once the district veterinarian is completely satisfied that everything possible has been done to check and verify the data, should he submit the forms for computerisation.
The next line of defence is the data input clerk. They will obviously complain about poor handwriting, but experienced clerks will often detect other anomalies, such as incorrect place names, the diagnosis of a disease in an area where that particular disease does not normally occur, and so on.
The most important “data sifter” is, of course, the epidemiologist. His job will involve preliminary analyses to see what is in the database, follow-up communications to field veterinarians, and the detection of fabricated data. Additionally, he will be on the lookout for trends in disease spread, and for unusual occurrences that will alert him to possible problems brewing in the field.
The epidemiologist will be watching performance indicators, not only as far as field staff are concerned, but also checking the work of input clerks to determine typing errors. Errors can be quantified, for example as x errors per 100 questionnaires. Very often, the easiest way to look for input errors is to get input clerks to check each others' work. Although this will reduce the epidemiologist's work load, it will not free him from the need to carry out spot checks.
Ideally, there are a number of data sources that should be tapped. Although the means of collection of information, as well as storage format, will obviously differ, the streams of data will all flow into the same management system, to be analysed under the auspices of the one person, and to inform the decisions of the same veterinary service. Data may come from the field (veterinarians, lay staff), from abattoirs, and via laboratories (serosurveys, confirmatory sampling). This makes the flow more complex, and data checkpoints will increase in number, but the principles described above will remain the same.
Any information system must have a data backup capacity. Backups are kept in a variety of ways:
duplicate copies of completed questionnaires at field offices.
original copies of completed questionnaires, sorted according to district and month at the data input centre.
electronic backups of data that has been stored on computer. Data may be backed up on tapes, diskettes, or compact disk, but it must be backed up. Some computer systems make provision for automatic daily backups, and these can be timed to take place after hours. In the absence of an automatic backup facility, data should be backed up at least once a week, so that if a computer crash occurs, the amount of information that has to be re-entered is minimised.
When making backups, especially onto tapes or CDs, the “grandfather-father-son” principle should be followed. If, for instance, the first backup is made on CD no. 1, the next on CD no.2 and the third on CD no. 3, then to make the fourth backup, the user will revert to CD no. 1. This ensures that the most recent backup is still available and unscathed should a crash actually occur during the backup process.
Running a system without a computer
It may seem unusual to include this topic in a manual written in the computer age, but the fact is that information systems can exist without computers, and in some cases, they simply have to. In far-flung areas where electricity supply is erratic or non-existent, in very poor countries or regions, or in small projects run by NGOs, computers may be inappropriate, impractical or simply too expensive.
In such cases, data are stored on large tables called tabulation sheets. Each column in the sheet will contain one of the variables to be recorded, while each row of the table will represent the information contained on a single questionnaire. It is self-evident that questionnaires will have to be short and simple, and data volumes low! Data entry staff will transfer data from the completed questionnaires to the tabulation sheet by hand.
An example of a tabulation sheet is given below:
|Month||Place||Disease||Species||# Sick||# Dead||Signs||Officer|
On a monthly basis, data can be analysed using a tally sheet, for example:
|Diseases: (sick + dead)||Rinderpest||CBPP||LSD||Rabies|
|3+6 = 9|
1+10 = 11
|5+2 = 7||0||1|
Such analyses can be done (as in the above example) per disease, or per village, per species - whatever the need might dictate, but they must be done by hand, and the database must be small enough to make allowance for such work.
More will be said about computerising data later.
Questionnaires can be used in a number of ways. Firstly, they can be used as a rough guide to a discussion, and as concrete information arises from the discussion, it can be entered onto the questionnaire. Alternatively, notes can be kept of interviews (whether person-to-person or group interviews) and pertinent information can be transferred to the questionnaire at a later stage. Or, and this is most usual, the questionnaire is used directly during a formal interview, and its structure strictly followed, point for point.
No matter which way a questionnaire is used, its design is of great importance to the success of information gathering, and a number of important principles apply:
This refers to the actual variables to be recorded. They will have been decided during the design stage of the information system, and must be kept to a minimum.
The length of the interview (which is directly determined by the number of variables to be recorded) must be short. 15 minutes is a practical guideline, shorter is better. When doing complex surveys (eg. Rapid Appraisals), interviews may become longer, but under those circumstances, anything longer than one hour is excessive.
The questionnaire layout must be clear and logical so that the interviewer follows a logical sequence down the page from start to finish. Likewise, it must provide a logical sequence for the data entry clerk to follow when transferring data from the questionnaire to the database. It must also be clear and legible. Certain parts of the questionnaire may contain comments or information not intended for the database - these should be clearly marked.
The form must be self-contained in that all necessary information is contained on it - district, date, details of interviewer, name of place, disease information etc.
Where possible, information should be coded on the form to simplify and accelerate data input. If the code for a particular district is a set of letters, the field worker should enter his district's letters on the form - or they could be pre-entered before a field visit is undertaken. Species, such as “Bovine” could be ticked off in an appropriate “check box” on the form. Nonetheless, on-form coding should be approached with care. If check boxes for every conceivable alternative of every variable are provide on the form, it will become complex and possibly difficult to read and complete - perhaps leading to the wrong boxes being checked, or even omitted. Dealing with purely “veterinary” information means that it is often best given in narrative form and then encoded at input level - for example, issues such as clinical signs or post-mortem lesions. Giving a few alternative signs on a form will lead to all syndromes reported having very similar appearances. Such a form of information collection could also have the effect of being a set of leading questions, in which farmers are lead to describe a particular condition.
Consider the possible effects of the following interview:
“Did your animals show discharges and diarrhoea before death?”
As opposed to:
“Please describe all abnormal signs you noticed before your animals died.”
The first question is clearly leading, and if animals did not show clinical signs within the narrow range given, confusion will result. Either the stock owner will assent to the few symptoms given, even if they were not seen (after all, his animals did die) or he may be reluctant to co-operate further. The interviewer will simply record what he is told, blissfully unaware of whether it is valid or not, and garbage will enter the system.
The second question allows a full description of what really happened, and the interviewer will get closer to the truth. Instead of being forced to fit all diseases seen into what will perforce be narrow sets of clinical signs, it will be possible to identify a wider range of diseases. It will thus be possible to get early warning of new - and hitherto unknown - diseases in the area.
Clinical signs and post-mortem lesions are best left as open as possible so that farmers have a free reign to describe exactly what they saw, and are not limited by a few choices on a questionnaire. In such cases, encoding can be done at the level of computerisation, where a full list of codes to cover every symptom and lesion can be kept, and used during input.
Issues such as paper size and quality, clarity of printing and size of spaces for recording answers deserve careful consideration. A smart, simple questionnaire will further ensure good quality information. Shoddy, overcrowded and complex forms will be completed in a shoddy manner.
Questionnaires should be so designed as to accommodate all diseases, not custom-made to suit just one. Expecting field workers to carry a variety of questionnaires, each for use with a different disease is wasteful of resources and will only cause confusion. (The same, incidentally, goes for the database - creating a separate information system for each disease is senseless. The entire system should be broad-based enough and robust enough to cope with any eventuality. Anything less than this reflects poor system design).
A few specimen questionnaires are given at the end of this manual as examples.
Using questionnaires for the first time means a few things. First of all, an instruction manual should be prepared, explaining the questionnaire in greater detail. The aims of the questionnaire should be summarised. What is required under each data item (in logical order) should be explained.
Once this has been done, a number of the field workers who will be using the questionnaire should be trained in its use. A briefing on the contents of the questionnaire and how best to put the questions should be given first, and then the workers should use role-playing techniques to test their abilities among themselves. A field test is then carried out, with a senior staff member (preferably the questionnaire designer) accompanying each field worker for the first interview. Each worker is then allowed to do a few further interviews on his own.
After this exercise, the response of the field workers is evaluated, and the first data are examined in order to determine shortcomings with the questionnaire and modify it if necessary.
Databases, in their most general sense, are simply a means of storing data - whether it be on a computer or on a tabulation sheet, in a card index system or in a ledger. In recent times, however, the word “database” has almost become synonymous with “computer”. Computers were, in fact originally conceived as high-speed data storage and retrieval mechanisms.
As computers are evolving at such a frightening speed, no attempt will be made to suggest specifications in this manual. Prospective users should consult with a number of vendors in order to gauge trends and make intelligent purchases.
Database software is generally able to store most types of data - whether as numeric or non-numeric variables, sort data, and retrieve specific items or subsets of data in response to user queries. Data files, or specific subsets, can also be exported into spreadsheet programmes for graphic analysis, and where data are georeferenced, can be introduced into GIS software to be visualised as maps.
As mentioned earlier, it is important to distinguish between data on questionnaires that are destined for computerisation and those that are not (eg. specific comments from field staff to their supervising veterinarian, or information cattle movements that needs to be followed up), and also to take note of the fact that certain information, while it must be computerised, will need action before it even reaches the computer (eg. a suspicion of a new disease outbreak).
Various database programmes are currently available, and while these may change in the future, they can probably (at the time of writing) be divided into three broad categories:
Software for small business (and therefore for relatively small datasets) - eg. Microsoft Access.
“Middleweight” software for medium-sized databases of all types - eg. Microsoft Visual FoxPro, Borland Visual dBase and Borland Paradox.
Software for large-scale databases (often used by banks, large companies, government ministries, international organisations) - Oracle, Sybase.
No blanket recommendation can be made here regarding the “best” software to use, and the above can be taken as examples only. Software requirements will vary from system to system, and obviously according to data volumes collected. What is important is ensure that the hardware (ie. the computer) has the capacity to run the chosen software, and the system overall can cope with the input volume and the outputs required.
What also needs mentioning is that there are some more or less custom-made software packages available for epidemiological data storage and analysis. Some of these packages are old and most have very limited capacities.
Database software is normally good for storage, quick sorting and retrieval of data, and small statistical manipulations (eg. range, mean, std deviation) are sometimes possible. In order to carry out advanced data analysis, spreadsheet packages (eg. MS Excel, Corel Quattro-Pro) and statistics packages are needed (eg. StatGraphics, Kwikstat, Epi-Info). For spatial analysis, one moves into the advanced world of Geographic Information Systems (eg. Arc View, MapInfo).
Information is stored on computer in entities known as “files”. Naming of files, and what goes into them (ie. their structure) is the prerogative of the user.
Database files are structured, with each observation (in our case, the equivalent of a questionnaire) recorded in the file as a “record”. The variables in each record (eg. locality name, animal species, number sick, etc) are known as “fields”. A group of records with a homogenous structure, and stored together, is known as a table.
There are different types of variables, and the software will handle each type differently - for example, a place name would be stored as a “character” variable, and the programme would be able, for example, to sort them alphabetically. The number of animals reported dead in an outbreak would be stored as a “numeric” variable, and calculations could be performed upon it. Variables and their types must be defined during database design, and must exactly follow the structure of the questionnaire.
As this is not a manual on database software, there will be no further discussion of software at this stage. Enough information on databases is available elsewhere.
As part of database planning, it is essential to choose a set of codes for each variable before the database is launched. For example, one disease may have many names, such as blackleg, black quarter, or quarter evil. Which of these to use in the database? It would be preferable to type in a standard code for the disease, such as BQ (or some other suitable, but recognisable, code). The data entry clerk could be supplied with a “look-up table” (which may be programmemed into the computer, or kept in a separate manual), look up one of the synonyms, and enter the correct code. Likewise, when it comes to symptoms, “drooling” and “salivation” might be encoded as SALIV.
Encoding means that information can easily be retrieved. When one wants to enquire about black quarter, it would not be necessary to ask separately about each of the three synonyms in order to get a full picture of the disease - one query using the code “BQ” would immediately render all the information available to the enquirer.
Data input staff soon become very familiar with the most common codes and after initial “teething” periods, will be able to enter the correct codes for most variables - districts, species, diseases, clinical signs, etc, almost without thinking.
Ease of input
Making inputs into the database must be a simple operation. The “user interface” through which the input staff have to work needs to be simple and friendly. Typing into a template and having on-line assistance available is a great help, and the order of input of variables must follow the order of the questionnaire. Having an easy-to-read manual written to help data entry clerks is ideal, and of course, training is essential. Data entry staff must also be chosen for their speed and accuracy, and their ability to cope with monotonous, repetitive work.
Querying the database
Most queries will be fairly simple, and most modern software has “query builders” built in which the user can employ in analysing data. Queries usually take the form of something like “list all the foci of pasteurellosis in bovines in district x for the month of June” or “calculate the sum of all CBPP cases diagnosed during the year.”
Exactly how to formulate queries will be described in the software manual, and on-line help and hints are usually available within programmes. It is important that epidemiologists perform a standard set of queries very regularly (say each month) for reporting purposes. Having a set of standard outputs and analyses keeps field staff up and management alike up to date with the disease situation in the country, and helps foster confidence in the system.
Although data control has been mentioned earlier, it deserves further treatment at this point as a separate entity.
Data quality control is an integral part of information management. As has been made clear elsewhere in this manual, it is a fatal mistake to assume that all data entering a system are good data.
Data move from the field to the district office to database input. The more checks are conducted before input, the better. If a problem is detected while a piece of information is still relatively near to its source, it can be followed up and corrected with relative ease, the further data move from their origin, the more difficult - and costly - corrections become.
Checking levels and what is checked are as follows:
In the field:
Careful questioning of the farmer to capture a true reflection of epidemiological information. Leading questions should be avoided. If information comes from farmer recall, it may be worthwhile to cross-check information with other family members or incontact farmers.
At the district office:
Completed questionnaires are evaluated for legibility, correctness (eg. place names, code usage) accuracy and internal logic. What is written must, in other words be clear, neat and make sense. Where a query arises, the district supervisor (preferably a veterinarian) must first contact the interviewer concerned to clarify the issue with him. If necessary, and if possible, a return should be made to the original data source (the farmer) to follow up. Not only is it easier (nearer) to do this while still at field level; it is also possible to recapture information while it is still within reasonable recall and important details are not yet forgotten.
At the epidemiology unit
The data entry clerks will detect - and complain about - poor handwriting. The epidemiologist will further do spot checks on individual questionnaires before data entry, and also cross-check data entered onto the database with the questionnaires from which the data came on a random basis.
Data input staff will need good training and careful monitoring. It essential that data typists do not sit in front of computers for extended periods, as this leads to physical tiredness, eye and mental fatigue and a lack of concentration. Where possible, data entry should be interspersed with other tasks, such as the sorting and filing of questionnaires, doing data backups, sending enquiries to the field about data quality, etc.
A very important task of the epidemiologist, mentioned only in passing thus far, is the detection of fabricated data. It is a painful truth that some field workers will not always visit each place on their visiting programmes, but may simply sit at home and complete questionnaires in their easy chairs. Discovering such data is not easy, but a few pointers might help. It is helpful to carry a preliminary analysis of data and look for tendencies. First, simply view recently entered data in its “raw” form in the table, and then carry out a few simple statistical procedures, such as calculation of range, mean, standard deviation, construct histograms, view data spatially with a GIS. Look for the following:
repetition of the same, or similar values (eg. herd sizes, or numbers of animals affected by a particular disease).
variables having a small standard deviation, with many values clustered around the mean.
a very homogenous disease/clinical signs pattern in a particular area.
many values ending in zeros or fives.
herd size distributions that are vastly different in adjacent and very similar areas.
many observations of herd inspections with little or ancillary information given on the questionnaires (eg. questions on clinical signs or pasture conditions may be unanswered).
details given for a visit to a particular place may be identical to another visit to the same place six months previously, ie. the previous visit's questionnaire was copied.
a particular disease being widespread in one worker's area while being absent from an adjacent area where prevailing conditions are very similar.
All of the above are reasons to be suspicious about data. Where there is a pattern of repetition in a particular worker's information, then his data are suspect. Where there are vast disparities between data of two adjacent field worker's areas, it will be difficult to say exactly which of the two is at fault. In general, however, it may be necessary to send an independent team into the field to conduct random inspections in the areas from which devious data have come as a verification study. Results can be compared afterward. Where it is obvious that a field worker has fabricated data, severe disciplinary steps must be taken.
Needless to say, field management of veterinary staff remains an important aspect of basic management, not just data management. Staff must work according to fixed programmes, and spot checks must be made by supervisors from time to time to ensure that they are actually “on programme.”
Errors need not only be the result of fabrication, though. Biases in data can arise for other reasons, and epidemiologists must be aware of these. Routine surveillance is very often not randomised, but pre-programmed, and anomalies may arise:
field workers might use motor vehicles and not always reach more remote areas. This may confine field work to areas near main roads where richer farmers often live. Such farmers usually have larger herds and flocks, and use more medicines and vaccines - with a correspondingly lower disease incidence.
field workers are usually male. For various reasons associated with personal prejudice or cultural taboos, they will not interview women, and so gain inaccurate data on animal species with which women usually work, eg. poultry and small ruminants.
Even where surveys are randomised, errors will creep in. If livestock numbers are incorrectly estimated, serum sampling tubes may be too few, resulting in unrealistically small sample sizes. Sampling animals of unknown vaccination history may result in the detection of vaccine titres during sero-surveillance. Farmers may lie about what diseases their animals have had.
Data input staff may quite literally have a bad day and miss the typing-in of a batch of data forms - or forms may get lost in the post, leaving a “hole” in the database.
Lists of what can go wrong are endless and very depressing, but can be minimised through:
thorough staff training (at field and headquarters level)
creating strong farmer awareness and gaining their co-operation
good planning of data collection, routine surveillance and special surveys
enforcing strong discipline amongst staff
having a vigilant epidemiologist
Verification has been mentioned already, but it bears repeating.
Giving feedback to the field in the form of analysed data regular reports is a good way of seeing whether trends registered in the database are a good reflection of trends on the ground. Another tactic - and this can be expensive - is to send each field veterinarian a monthly printout of all data entered from his district during the preceding month for him to inspect, correct and return to the epidemiology unit. In this way, records that were incorrectly entered can be corrected, and should the veterinarian wish to update some information on particular incident, he has the opportunity to do so. If this is too expensive, sending each veterinarian a six-monthly summary of his district's data may be a more viable option.
Visits by epidemiology unit staff to the field are an indispensable means of maintaining contact, and an opportunity for on-the-spot validation studies.
The importance of feedback has thus far been mentioned several times. It maintains the chain of communication and ensures interest on the part of field staff. The question is, how?
Regular reports giving summaries of the disease status in various parts of the country are probably ideal. Such reports must be clear and interesting, well illustrated with graphs, maps and tables. Veterinarians must be encouraged to comment on such reports, write “letters to the editor” and contribute short articles on interesting cases. A lively rapport between the epidemiology unit and the field is certain to keep the information system alive, while at the same time ensuring that all field staff are kept well informed. Part and parcel of this communication will be regular meetings between district vets and their field staff to discuss the contents of such regular reports. It is of the utmost importance that this feedback reach the people who collect the data on the ground. They must be made to feel part of a team.
It is a good idea, as mentioned in the previous section to send each district - or even, where practically possible, each staff member a short summary of their reporting every six months. A short table giving disease totals together with one or two illustrative graphs is sufficient for each person to know that his data area received and appreciated. It also gives each person a “snapshot” of the situation in his area.