Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
A report by an SCN working group and the Advisory Group on Nutrition to the SCN
Part 1. Deciding on methods
1. Introduction and rationale
Nutritional surveillance means "watch[ing] over nutrition. . . to make decisions to improve nutrition in populations" [1]. The concept stems in part from public health, where information is needed to trigger action-itself usually predefined. For example, reports of cholera cases at a certain incidence may cause an immunization programme to be launched.
A "surveillance cycle" is central to this concept, where problem detection leads to decisions for intervention; interventions are carried out, and information on their conduct and population monitoring cycle back to further decisions. This is illustrated in figure 1[2].
Following a proposal from the World Food Conference in Rome in 1974 (Resolution V, pare. 13 [3]) and an FAD/WHO/UNICEF expert committee which defined initial approaches [4], an inter-agency review of experience in 1981-co-sponsored by the ACC/ SCN-made a crucial distinction between different objectives of nutritional surveillance systems in terms of decisions, and the supporting information systems [1]. The importance of setting priorities for these types of decisions before considering data needs was stressed as paramount. Thus, at that time, three types of surveillance were defined, and details of approaches based on experience were described by Mason et al. [5]. It is still considered crucial to distinguish different possible types of surveillance systems in order to set priorities. Not all information is useful for all purposes.
The focus must be on the decisions needed, and these are very different if they relate to long-term policy (moreover, varying by sector), to current programme management, or to preventing acute food crises. Those making the decisions in these separate areas will differ, and are often located in different institutional settings. The information demanded varies also. Long-term trend monitoring may require relatively infrequent data outputs, with an emphasis on data which are demographically representative, and on causal analysis. More frequent information, including that on service delivery, is needed for programme management. Suitable predictive indicators are vital for timely warning.
Three types of nutritional surveillance were therefore distinguished, as follows:
(a) For policy and planning decisions, in the long term. Here information should be linked to policy-making, and intra- and inter-sectoral linkages may need to be developed.
(b) For programme management, generally for programmes with nutritional objectives. In this case, information should be linked to management decisions, and data on programme delivery and targeting are important. Much of the data will come directly from programme sources.
(c) For timely warning and intervention. Here the information needs relate to timing of interventions, the nature of which depend on the stage of problem development. In practice-associated often with drought-the purpose is triggering decentralized (usually, or preferably) decisions on short-term measures to prevent, or, if that fails, alleviate, acute food shortages.
Explicit in this is that the starting point is with policy and interventions-and not with data per se. Indeed, experience is consistent that beginning planning with data considerations can be counter-productive. This means that indicators must not be considered in vacuo. On the contrary, possible actions should guide the development of information systems; and information systems should, as far as possible, have response mechanisms built in. In practice, data have sometimes been used for advocacy, and for fostering changes in attitude, hence policy.* How successful this has been is still not clear; no doubt decisions relating to nutrition have been and are being made in this way, although the process and specifically how and what information contributes to this is not always readily known.
FIG. 1. Iterative procedures for generating action
The Inter-agency Food and Nutritional Surveillance Programme (IFNSP) originated substantially from concerns for the effects of economic recession and structural adjustment on nutrition. which were the subjects of the SCN's symposium at its 12th session, in Tokyo in April 1986. This was a somewhat new concern for nutritional surveillance, with overtones of international advocacy and awareness-raising, especially for international financial institutions (mainly the International Monetary Fund and the World Bank). The potential responses to deteriorating nutrition in the face of structural adjustment were usefully distinguished as (i) modifying medium-term (e.g. 1-5 years) adjustment policies and/or (ii) compensating for immediate (e.g. Iess than I year) effects by special programmes generally aimed at vulnerable groups. Responses of type i came within existing experience of surveillance for policy-making.
However, the question of initiating and targeting rapid compensatory measures to protect the vulnerable during the adjustment process is relatively new. Information requirements for such a purpose relate to both programme management and timely warning. This required a review of recent experiences and some innovative thinking, and is gone into in detail in part II of these recommendations. Such a hybrid system can be regarded as a particular application of timely warning with intervention capability for initiating compensatory action, although having features of programme management surveillance once in operation. Given the likely magnitude of resources needed for compensatory action, a pre-established commitment to intervene as required with the necessary funds seems essential.
Surveillance systems for programme management and for timely warning have to date mainly been set either in relatively stable situations (e.g. to support supplementary feeding programmes) or in acute environmental crises (e.g. timely warning systems related to public works and food distribution). In the case of economic adjustment, support both to public expenditure on existing priorities (e.g. health services for the poor) and new support measures (e.g. targeted food distribution and subsidies) may need to be advocated and implemented. Building in such support to adjustment programmes is however something of a recent idea-PAMSCAD (Program of Action to Mitigate the Social Costs of Adjustment) in Ghana is one of the first experiences. Hence nutritional surveillance in this context is also somewhat pioneering.
Ways in which experience in nutritional surveillance methods could be adapted to the concerns of compensating for structural adjustment effects thus presented a need for some original thinking.
It is worth emphasizing that the style of nutritional surveillance has been, and should continue to be, a low-profile, institution-building approach, supporting national or local efforts within the decision-making structure. The IFNSP started off with this idea clearly in mind, with an emphasis on facilitating developments in-country, with training and add-on funding as important aspects. Given the very limited IFNSP resources per country, as well as the usually far greater concurrent relevant activities (national statistical efforts, other surveillance support, the SDA programme in Africa, etc.), it makes sense for the programme to see how it can best complement these.
It must be appreciated that the IFNSP is intended to assist the development of national priorities and not the imposition of outside priorities, including priorities as to the type of data output to be supported. It may usefully facilitate and encourage exercises to clarify national priorities, and this needs to be sensitively approached to avoid inappropriate promotion of external views.
In this context, it cannot be stressed too strongly that the IFNSP is but one element in efforts, national and international, to foster nutritional surveillance. Equally, nutritional surveillance is only one activity within the spectrum of activities required to prevent hunger and malnutrition.
1.1. Established methods of nutritional surveillance
Nutritional surveillance systems, generally recognizable within the typology given earlier, exist in many countries, tailored to local circumstances. Some twenty or so surveillance programmes were recognized in the early 1980s, and since then there has been considerable development in line with the conventional framework.
The latter framework is described in a number of publications. The most detailed is Mason et al. [5], which expands on the previous UN position-ACC/ SCN and WHO [1]-discussed at the Cali conference in 1981. This framework provided the basis for the original proposal for an inter-agency programme, approved by the ACC/SCN at its 13th session (1987). And this in turn led to the proposal put to the UNICEF Executive Board in 1987, and a US$10-million noted project being subsequently approved [7].
It is recommended that the established procedures for nutritional surveillance be used to guide the implementation of the IFNS Programme.
One particular procedure, referred to as "initial assessment" (described in Mason et al. [5, chap. 2]), may be adapted, as appropriate, in designing nutritional surveillance systems. The primary purpose of this procedure is to help identify the priority actions to be supported by nutritional surveillance, and hence to give guidance on methods to be applied.
Specifically then, the first step is to determine the principal requirement for nutritional surveillance in the particular local circumstances. This means decisions are required on priorities for nutritional surveillance development, which can be for
(a) long-term policy and planning, or
(b) programme management, where programmes exist or are being established with nutritional objectives, or
(c) timely warning and intervention, within which we now differentiate between systems responding to (i) natural phenomena, e.g. drought, floods, and (ii) economic stress related to structural adjustment.
Deciding between these alternatives then directs attention to the appropriate type of surveillance. If timely warning in relation to economic stress (type c (ii) ) is considered the most relevant, and if resources to compensate can be built into the adjustment programme, then response mechanisms and data needs can begin to be considered. (Methods considered appropriate to c (ii) above, are proposed in part 11.)
2. Common indicators
The question of whether the IFNSP should select and promote the use of a basic set of common indicators was one of the earliest issues to arise, and the matter has been extensively discussed in the interim. (A meeting was held at UNICEF, New York, in October 1987 to consider the issue. A second meeting was held in June 1988 under the auspices of the ACC/SCN; part 11 of this paper is based on that meeting.) From what has been said above, it will be evident that for national or sub-national surveillance systems, the first concern is to clarify potential actions/interventions, and then the information needed to identify or trigger a predefined response. Hence, for the exclusive purpose of supporting national development of nutritional surveillance, it could be counter-productive to recommend a set of indicators in advance-and certainly specifying indicators without reference to their use should not be considered.
Nonetheless, three nutritional status indicators have been considered at the inter-agency level as potentially useful generic indicators in those countries where capacity exists for effective collection. These are:
(a) low birth weight-as the measure of nutritional status of the child at birth; as a predictor of the child's likelihood to grow and develop; and as an indirect measure of the health and nutritional status of the mother of the child and, in aggregate terms, of pregnant women and mothers as a vulnerable group;
(b) weight for age of children under five-for monitoring growth and development of this vulnerable group, disaggregated by more specific age groups under three and, where possible (as in household surveys), supplemented by weight for height and height for age, which are more specific measures;
(c) height for age of primary-school entrants-a measure taken at the end of the early period of growth and development which then becomes a cumulative indicator of the overall nutritional and health status of the child at the beginning of its school going years and an overall indicator of social and economic development.
Some or all of these indicators have featured in early recommendations on surveillance design [4; 8] and will, in all probability, be selected in any substantial intersectoral information system, particularly in the area of health impact assessment. (Further work is planned during 1989 to address indicator issues in connection with their use for monitoring food access, health/disease trends, and anthropometry. The reports of these meetings will provide a technical overview of these areas which will complement the present document.) However, it must be appreciated that the listed indicators are outcome measures, reflecting the impact of past events, and have limited leading or predictive capacity.
2.1. Common or core indicators versus comparable indicators
The drawback of insisting on certain specific outcome (or status) indicators-to be common to all IFNSP country programmes-is primarily that this may conflict with current in-country practice and be regarded as unwarranted and potentially disruptive. Furthermore, as discussed above, different indicators may well be more appropriate for different situations. The potential advantages on the other hand of recommending common indicators between countries (or areas) are:
-they facilitate cross-country comparisons, for international advocacy and possibly decision-making;
-inferences about the effects of adjustment etc. on nutrition are substantially strengthened when similar patterns are seen in different countries or areas;
-experience in data collection, analysis, and inter pretation is more easily shared and extended when the same variables are used;
-in the long term, this would facilitate progress to wards common standards.
With regard to international advocacy, successful country nutritional surveillance programmes can usefully contribute without necessarily maintaining fully comparable indicators. Furthermore, movements in alternative indicators (e.g. of child nutrition) would be often highly correlated, giving much the same story.
The broad application of a few types of indicator (e.g. a direct measure of nutritional status; an indirect economic indicator of food access; use of health services) could be sufficient to provide the advantages of commonality while, at the same time, allowing for in-country flexibility and specificity. The actual indicators chosen need not always be identical in all areas within, or across, countries. The food price index as a ratio of the consumer price index, for example, may be a widely used food-access indicator, but in rural areas the retail price of staples may be the only data available. Possibly, too, this indicator could also be linked, where applicable, to the minimum wage. Similarly, a child-nutrition indicator might usually be, for example, the prevalence of 1-5-year-old children more than 2 SD below international standards, but in certain areas the prevalence of, say, 1-3-year-olds under 80% of local standards could be used. This variation would not seriously invalidate the advantages of commonality or comparability.
This implies that in practice certain data from country systems may be regularly tapped for international reporting. The most commonly available indicators are likely to be those of child growth. Experience with, for example, the SCN Secretariat's Update report on the world nutrition situation [9] has shown that precise comparability across countries has not proved currently either feasible or essential. Much more important is to have frequent, reliable, and up-to-date indicators-in the literal sense of that word- of the nutrition situation; but the indicators need not be exactly the same from each country. Irrespective of which indicators are chosen, the use of these same indicators in the same situation over time is crucial, in order that a consistent time series be available to monitor trend development. In the context of nutritional surveillance, trend assessment is arguably more important than determining point-in-time level estimates.
In sum, specifying any particular indicator in advance-the idea of "core indicators" for the IFNSP-is logical on only two grounds:
-to facilitate inter-country comparisons of trends, for the purpose of international advocacy;
-to facilitate sharing of experience in issues of data collection, interpretation, and use; this is particularly useful for biological-outcome indicators (e.g. the prevalence of underweight children) where the cross-situational biology is comparable.
Since the primary objective of the IFNSP is for in country use of information, this means that strict comparability of indicators (across countries) is a desirable spin-off but not the main concern of recommendations on methodology.
The use of closely comparable indicators is judged to be as useful as strictly common indicators.
2.2. Summary
1. The first step is to decide on the type of nutritional surveillance system to be deployed. This will be a function of the kinds of priorities and decision to be supported. Historically three types have been distinguished: (i) for longterm planning, (ii) for programme management, and (iii) for timely warning and intervention (against acute food shortages). Conventional approaches-which continue to be recommended-to the application of surveillance in these settings are now well established. If, however, the planning context is that of structural adjustment, then a new approach is advocated, which forms the basis of the recommendations in this paper.
2. Indeed, the requirement for a new approach arises because of the comparatively recent, but very widespread, implementation of structural adjustment programmes and their now well-documented impact on vulnerable sectors in society. There is, therefore, a particular need for developing a tailored information support system to provide rapid reporting of key indices, which would enable appropriate compensatory measures to be applied.
3. Such a system may be considered as a variation on the conventional nutritional surveillance for timely warning and intervention, suitably modified in terms primarily of indicator choice, data capture, and modes of analysis.
4. Finally, in connection with the IFNSP the conceptual and practical issues concerning the relative merits of common versus comparable indicators were considered, with a recommendation that the latter were judged to be as useful as the fomer.
In part II, attention is focused specifically on the nature and requirements of a nutritional surveillance system for timely warning in the structural adjustment context.
Part II. Surveillance methods related to compensatory action during structural adjustment programmes
1. Introduction
In the period since 1980, many economies have suffered severe economic recession, with serious consequences for nutrition that are now becoming clear. In the last five years or so an increasing number of these countries have begun structural adjustment programmes. UNICEF's "Adjustment with a Human Face" [10] described ten country case studies, in most of which nutrition was reckoned to have deteriorated. During 1988, ACC/SCN accumulated data covering the 1980s for over 30 countries, which document nutritional effects of economic stress, structural adjustment, and drought [9]. A classic example, from Ghana, is shown in figure 2.
By mid-1988, at least 20-25 countries in sub-saharan Africa had begun structural adjustment programmes, most in collaboration with the IMF and World Bank. This represents a majority of countries and people in that region. In Latin America, the debt crisis had led in many countries to severe retrenchment in public-sector spending, massive inflation, and devaluation: all characteristics of structural adjustment, whether or not formally identified as such. In some of the more populous countries of Asia, including China, reverberations of economic change may be longer lasting, bringing change but undoubtedly causing stress.
All this emphasizes that the economic changes happening now, in the late 1980s, are of great significance for human well-being, and for nutrition as a central part of this. The priority in the long run is for reestablishing a trend towards sustained improvement, but in the short run some segments of society evidently suffer. If we are concerned to prevent malnutrition, often we will have to address the immediate risk associated with structural adjustment.
Buffering of the poorest in society against the shocks of adjustment is beginning to be accepted as a necessary adjunct to structural adjustment programmes. This requires that contingency resources be made available for this purpose. Information will be needed for effective use of such resources.
This thinking in turn argues for a potentially useful role for nutritional surveillance in protecting nutrition during the adjustment process. We need to broaden the argument somewhat, however, because malnutrition is only one of the transitory consequences of adjustment that governments and financial donors may be willing to commit resources to alleviating. In this sense we should draw attention to monitoring nutrition as, first, an established summary indicator of well-being in society and, second, perhaps the most readily available of such indicators on a frequent and disaggregatable basis. Hence, nutritional surveillance in the context of structural adjustment may be useful as a core method for discovering and preventing adverse social effects. And it may be uniquely useful to complement longer-term monitoring programmes. There are few (if any) other indicators available that can provide the insights, in a timely manner, that nutrition indicators can. But even these will be of no use if the resources and mechanism to respond are not in place.
1.1. Potential for nutritional surveillance in view of recent developments
The opportunity for nutritional surveillance to contribute in the context of structural adjustment depends on three features of the situation in many developing countries today.
First, observed trends in nutrition do indicate generally that development forces (education, health services, income growth) push in the direction of improvement. This was seen globally, and in most countries except in Africa, in the two decades up to around 1980. Recent economic recession has acted widely to reverse this underlying trend. In the process, severe crises (economic and natural) have caused peaks of malnutrition-figure 2 providing a good example. Thus there is a need for influencing policies to get back on track towards improvement. More immediately critical is to prevent the effects of severe within-year and year-to-year deterioration. Structural change is perhaps the major factor affecting nutrition in many countries.
Second, the shift in major external and government funding towards adjustment programmes, coupled with concern for the possible short-run effects on the poor, means that within adjustment programmes new resources are becoming available for buffering. Effective and efficient use of these resources requires timely, appropriate, and well-targeted intervention. The existence of these resources means that a surveillance programme could be particularly effective.
Third, improvements in statistical information systems provide a framework in which the particular contribution of nutritional surveillance can be most useful. In the context of adjustment programmes, there is both increasing support for monitoring social trends in general, and a major external investment in statistics for assessing the social dimensions of adjustment (SDA) in Africa. The methods generally in use require representative household surveys and a substantial number of variables, and even in the best cases take a year or more to begin to establish trends. This is inherent in the nature of the data collection, processing, and analysis methods needed to provide the required information for medium-term (year-to-year) policy analysis. Such an approach is not usually suited nor intended for triggering within-year action. Some methodologies for nutritional surveillance-as will be discussed below-are appropriate for this purpose and furnish a complementary system to that of the large-scale survey. Moreover, both will benefit. Surveillance will be improved by the validation and greater detail provided by surveybased statistical systems; survey-based systems will benefit because surveillance can strengthen one of the major weaknesses they have-the relatively lengthy turn-round time from data collection to useful information.
These considerations lead towards recommending that under some circumstances-see part I-an adaptation of nutritional surveillance (specifically of timely warning and intervention) should be considered. The line of thinking that specifies the proposed methodology is developed below, leading to proposals for implementation.
2. General principles underlying the proposed approach
2.1. Operational characteristics
Timeliness and leading indicators
To trigger action to deal with human effects of economic shocks, indicative data of undesired (or unexpected) trends concerning the condition of vulnerable groups are required; some predictive information would be even better. This argues for relatively small amounts of data, perhaps obtained from sentinel sites, quickly summarized and reported. Monthly data, reported at least quarterly or preferably monthly, with minimum lag between data collection and information production, is desirable (see Mason et al. [5, pp. 180-85] for discussion of this point). A strictly limited number of different indicators should be used. Shifts in the pattern of these indicators, deseasonalized as appropriate, should be sought. For example, when two indicators, based on food prices and child anthropometry, are considered, both the predictive value of prices for anthropometry and their relationship at any one time (which would probably differ depending on whether the group to be monitored was net food producing or consuming) could be used. This concept is analogous to employing leading economic indicators which are used in short-run forecasting in the financial and business world.
Vulnerable-group monitoring rather than national indicators
The anticipated use of the data is for triggering action for defined population groups. The resources required to get timely national data would be much greater, and anyway more in the domain of existing statistical operations or of those planned, such as the SDA programme. It is proposed to focus on groups vulnerable to deterioration related to adjustment rather than on nationally representative data. This means that indicators would not be directly comparable to such national economic indicators as GNP or exchange rate. They would, however, show where and when to investigate further.
Empirical rather than causal analysis
Considerations of feasibility-i.e. timeliness, analytical capacity, and cost-and use for triggering rapid action argue in favour of easily acquired data on a few variables and topical analysis of current trends. It is clear that this approach would not lend itself ideally to causal analysis (although secondary analysis, with additional covariate data, might sometimes be valuable).
Sentinel sites, not representative sample
The above considerations suggest that data should be collected from specific (i.e. sentinel) sites in preselected areas. These sites could be markets and health posts. The use of sentinel sites will reduce reporting lags and basic costs. (Sentinel-site surveillance is used extensively in Europe [11] and in the developing world [12; 13]. In addition to considerations of timeliness and cost, the particular advantage of sentinel surveillance over periodic surveys, in developed and developing countries, is the continuous provision of high quality, reliable data.)
Generally speaking, available resources would be better used to improve data quality-accuracy and precision of measurements, care of interviewing, etc.-than to increase the numbers of sites. The reduction of non-sampling error sources is better value (within limits) than increasing the sample size. Some research is needed, with existing and additional data, into how far movements in indicators from a few sentinel sites do correspond to changes in the larger population.
Supplementary information
A further consideration is that if the primary data collection system is to be as economical as possible, some resources should be saved for supplementary surveys. These would be required to follow up changes detected in indicators, determining such factors as occupation groups affected (e.g. "exit polls" from clinics) and possible causes of price changes (e.g. to get an idea of volumes of staples in the market).
2.2. Implications of the approach
Support from the IFNSP is therefore considered one means of providing a simple, timely source of information, supplementary to other survey and related data. Where household survey programmes (e.g. SDA) exist, nutritional surveillance could benefit from such data. It is not considered that household surveys based on nationally distributed probability samples are to be recommended for the provision of the data needed rapidly and at high frequency. For this purpose the preferred methodology would be to use a sample of sites-usually clinics and markets- from which a regular flow of data would be recovered and quickly analysed. Occasional surveys would be undertaken within the catchment areas of the selected sites to provide more detail regarding the socioeconomic profile of the population groups using the clinic, or further detail on trading practices affecting the market. The sites of both clinics and markets would be selected from within a limited set of geographically defined areas which are known, from available data series (censuses, budget surveys, etc.), to have the highest concentrations of vulnerable people. Other areas containing the relatively better-off groups in the population would be excluded from this surveillance programme. Those areas eventually selected as meeting the necessary socio-demographic criteria are designated the "domains of study." The domains of study might include, for example, one or two low income areas in major urban locations, a small-holder area growing the staple food crop, and one or two areas classified as marginal according to agroecological criteria.
It is recognized that the domains of study may display considerable heterogeneity and may include population groups that are not of direct concern to the programme, but further refinement of the classification process is possible at the next stage of selection- the choice of sentinel sites (clinics and markets). These should be a number of sites within the domains of study around which vulnerable groups are known to be concentrated. It is important to note too that the selection of such sites is not random, but purposive according to the catchment area they are know to be servicing.
The purpose of the approach recommended in the stated context is to measure movements of indices over time in a limited number of identified sites, not to provide nationally or regionally unbiased point-in-time estimates.
The indicators that could, as a pair, be an integral part of a nutritional surveillance programme are an indicator of food access (most probably derived from the retail food price index) and the prevalence of malnutrition among young children (most probably weight for age). (It should be understood that the indicators cited are intended as examples only and additional indicators or indeed alternatives-e.g. immunization rates for selected diseases-might be considered more appropriate in specific cases.) Ongoing research is investigating the relationship between the movements of those indices over time, and certain propositions regarding their use can be put forward.
Where the two indices show a significant relationship (with appropriate lagging) over time, it is important to maintain the paired comparison
-to trigger provision for intervention when the food price indicator rises above the seasonally corrected trend, and
-to monitor for break-points where the relationship no longer holds.
The latter will be of particular importance when the adjustment process involves manipulation of food paces.
2.3 Complementarity with other monitoring activities
Many countries currently executing a policy of economic and structural adjustment are concerned with the problem of measuring the overall impact and effects of the adjustment process and are planning study programmes to investigate these effects. Perhaps the most prominent of these is the Social Dimensions of Adjustment (SDA) programme, which is scheduled to be carried out in some 20-25 countries in Africa where the World Bank is supporting the country with a structural adjustment loan. Under the SDA programme, it is intended to implement a series of national household surveys that will investigate the effects of the adjustment process on different population groups. The parallel interests of the SDA and IFNS programmes are obvious. The surveillance programme however has a somewhat different role to play because it is very much smaller and is focusing on the rapid feedback of information into the system, at least initially in a more direct manner than the large scale survey programme. The SDA programme, although having broader interests, is expected to include nutritional monitoring in its household surveys. These data will be of considerable use to the IFNSP and can provide a complementary source of information for checking or verifying the quality and bias of the prevalence data being generated by sentinel sites. Although the nutritional surveillance systems proposed here are primarily concerned with monitoring the short-term effects of economic policy on nutrition, they should also liaise regularly with institutions that are concerned with drought monitoring. In many countries, an early warning system of drought and crop failure is in operation. The activities of such a system are complementary to a food and nutritional surveillance programme, and information generated by an early warning system can be utilized by the surveillance programme to predict areas of high prevalence of malnutrition; equally, nutritional surveillance data are germane to the corroboration and operation of early warning systems. A working relationship between a food and nutritional surveillance programme and an existing drought and early warning system could be developed to the mutual advantage of both. For example:
-the surveillance programme may reflect or confirm forecasts generated by the early warning and drought monitoring system;
-the surveillance programme would benefit from the information provided by the early warning and drought monitoring system by extending its coverage to those areas identified as possible areas of food shortage.
3. Data sources and methods
Given the choice of indicators as discussed above and the recommendation to use existing data recording sources, the proposed sites for data collection are selected clinics and retail food markets (urban and/or rural, as appropriate). A sample of these sites will be required for each domain of study, which will have been agreed on according to the criteria also outlined above. It is recommended that the selection of sites be based on purposive selection and could be limited to as few as two of each type for each domain; i.e. the concept should be the use of sentinel sites rather than a random sample of adequate size to provide a valid estimate of the indices for the domain as a whole. The justification for this can be summed up in terms of the analogy with leading economic indicators, namely speed of reporting and analysis requiring a limited total number of sources, as well as emphasis on data quality.
The selection of the sites should take account, amongst other factors, of the following:
-the existence of co-operative resources for recording the required data,
-fast communication potential for rapid transmission of data,
-the location of clinic sites in terms of servicing a vulnerable group,
-market sites providing adequate volumes of sales of staple food crops, and not dominated by a monopoly seller or buyer.
The frequency of recording the data would likely be weekly for market prices and continuous for clinic records. The frequency of transferring the data to a centralized time-series data base should be monthly, using monthly average statistics compiled from the source documents. The frequency of analysis of the time series might be monthly or possibly quarterly, which should be adequate for most requirements unless a critical situation is encountered.
3.1. Supporting site surveys
An integral supplementary part of such a surveillance system would be the conducting, under the supervision of a central unit, of occasional surveys of the users of the sentinel sites. These surveys would be prompted by three main objectives:
-the need to provide explanatory variables of a socio-economic nature, for example concerning the parents of children utilizing a sentinel clinic,
-the need to check for bias changing over time in the population utilizing a sentinel site,
-the need to investigate the reasons for unexpected movements in an index by conducting a diagnostic study-for example, interviews with traders to establish the reasons for unexpected sharp price movements.
The methodology for such a study would be akin to that of an exit poll as used in market research, i.e. interviewing site users (e.g. by sequential random sampling either on arrival or departure from the site). The availability of resources and flexibility in their utilization are vital means of strengthening the very low-cost sentinel site monitoring using only two (or a few) indicators.
3.2. Data analysis
Data analysis is based on the movement of the indices in each site rather than on averaging the indices across sites-hence the emphasis on quality of data for each site and the impracticality, in a modestly funded system, of achieving adequate random samples for valid aggregation of estimates to a domain (or higher) level.
Analysis of a few indices (e.g. of food access and child nutrition) should be based on a longitudinal series emphasizing the movement of the indices over time rather than their absolute level at any moment in time. The expectation is that movements in an index and changes in the lagged relationship between the indices, when reviewed on a site-specific basis, will lead to plausible inferences regarding changes under way. (This is illustrated in detail in the paper by Taba-tabai [14] and in a number of country reports, most notably that for Ghana, in the Update report [9].) Such changes need to be addressed if deleterious nutritional consequences of economic adjustment are to be alleviated or avoided. The analysis should produce seasonally adjusted changes in the indices between two reporting periods, coupled with the estimate of the slope of any existing trend. The emphasis on detecting movements that accentuate or oppose the underlying trend highlights the need for quality of source data. This reinforces the recommendation for a small number of sentinel sites that can be serviced and assured in the maintenance of data quality.
The trigger points for management review could utilize:
-the detection of a seasonally corrected trend in one or both indices,
-the significant departure of the latest time point from the seasonally corrected trend if maintained over a predetermined number of months or if supported by a similar deviation in the second site of the pair within the domain,
-the strengthening or emergence of a lagged relationship between the indices,
-the de-linking of the two indices from a previously existing relationship.
As already noted, the choice of indicators will depend on local circumstances; however, in order to exemplify a number of issues on selection and analysis, it will be assumed that these will be a food price indicator and an indicator of the prevalence of child malnutrition, some details of which are given here.
Food price-based indicators
Vulnerable groups tend to spend a large proportion of their income on food. Food price changes could then have major effects on real incomes and hence access to food. Theoretical expectations as well as a fair amount of empirical evidence suggest that sharp increases in the price of food usually lead to lower food consumption by the poor and hence to a worsening of their nutritional status after a certain period. An appropriate food price indicator may thus be a useful predictive signal of the likely change in nutritional status. Evidence from some sub-Saharan African countries suggests that, in a number of cases, a rise in the national prevalence of malnutrition is preceded by a rise in the real food price (food price index divided by the general price index) with a lag of one to three months [14].
Food-related indices must be defined in such a way as to be appropriate for the particular vulnerable groups in question. For the urban poor, the real food price would be the most appropriate and would be calculated as the ratio of the food price index to the general consumer price index. These two pieces of information are generally available in most countries for the capital city or urban areas as a whole, with a lag of one or two months. In the rural context, where such indices are not widely available, a suitable substitute would be the retail price of one or two major staples in a moderately sized local market. This information must normally be collected on a weekly basis and used to derive the monthly average price. For want of a more appropriate deflator, the CPI may be used for this purpose.
The relative ease with which the above-mentioned indicators may be had is a major reason for using them as appropriate common indicators. Where feasible, however, additional food price-based indicators may be used as well. One possible improvement is when a measure of real income may be constructed on the basis of readily available statistics. For poor wage earning classes, the minimum wage, where in force, would be highly useful. This indicator should be deflated by the food price index for urban wage earners and by the price of major staples for agricultural labourers. All these indicators should be used to identify situations where significant positive deviations (actual value being above the trend level by a predetermined amount) might justify the activation of the response mechanism.
This is not a fail-safe system, and circumstances in which a significant positive deviation may not call for a response should be identified with the correct interpretation of the observed changes. In cases where an increase in food price is a deliberate policy objective, food producers' income would also rise and an increase in the real price of food would not be evidence of a likely deterioration in food intake and nutritional status. In the case of subsistence farmers, the food price-based indicator may not be relevant at all. To take another example, in the pastoral context a sharp drop in cattle prices after drought is an indication of distress, not of the cheaper availability of food for pastoralists.
Child nutrition indicators
The commonest available indicator of child growth failure is the prevalence of underweight children, derived from measurements taken in clinics or health posts for purposes of growth monitoring. This indicator is used as a measure of child nutritional status, and also, for example, as a summary indicator of child welfare in WHO's "Health for All" indicator series. Changes in child growth (hence underweight prevalence) result from changing food availability and health conditions. In this context of adjustment, there is both theoretical reason to consider that underweight prevalences will respond to changes in food availability, and evidence from clinic data that prevalences have direct linkages with economic stress, as well as drought, and show the expected seasonal changes related to harvests and lean periods.
It is felt that, of the possible anthropometric indicators, low weight for age is likely to be the common choice, partly on grounds of its wide and increasing availability. Even if weight for height and height for age are used, weight for age should, if possible, be included. On the issues of standards, cut-off points, and age ranges of children included, current local practices would have to be taken into consideration, but recommendations from WHO guidelines should be followed where possible for the sake of international comparability. The trade-off in either selecting the most vulnerable age range (e.g. 12-24 months) or tabulating by year age-bands, with the consequent need to increase sample sizes, should be borne in mind. Also, for summary indicator purposes, maintaining the common practice of pooling 1-5-year-olds may have an advantage in this context, since this summarizes different causal pathways (e. g. 1-year-olds, may be particularly affected by the mothers' conditions, the 5-year-olds by household food availability).
Low birth weight may be an alternative anthropometric indicator in some circumstances. However, probably not enough is known to recommend substituting it for indices of prevalence of underweight or stunting. Low birth weight obviously reflects maternal health and nutrition, as well as that of the child, but its sensitivity is less well established.
Movements in the prevalence of low values of anthropometric indicators are likely to be similar, within limits, for most current cut-point choices, standards, and age ranges. Equally, in the short run, weight for age and weight for height would be similar. Height for age is not suitable for measuring seasonal changes.
Prevalences of low weight for age (typically less than 80% of the median, or more than 2 SD below the median-which are nearly identical-in 1-5-year-old children) will usually be obtained through the health system. Children's ages (or birth dates) are recorded on first contact with the health system, when a growth chart is set up. A "master chart" or similar summary method is also maintained on which weight-for-age data for all children are recorded. These values may be retrieved every one to three months for community surveillance purposes and then reported to the district level. In the proposed system, these clinic summaries would be copied directly (monthly or weekly) to a central surveillance analysis unit.
3.3. Institutional issues
The food and nutritional surveillance programme should, as far as possible, use the existing institutional structure to develop all its activities.
The planning, development, and overall coordination of the programme should be effected through a small central co-ordinating unit, which might need to be established specifically for this purpose, clearly assigned both the responsibility and the resources to meet the programme's needs. It is essential to the success of the effort that the unit should be administratively located close to the policy-decision making authorities and that its activities should be built into the overall policy planning and implementation process. The unit would most probably be made up of a senior analyst with experience of time-series analysis, an assistant, and supporting clerical/ secretarial staff. Aside from basic office equipment, the unit would need to have computing facilities as well as a means of transport to liaise regularly with the reporting sites. It would also probably be necessary for a small mobile team of field enumerators and a supervisor to be attached to the unit. This would be used to undertake the periodic supporting site surveys described above.
The main functions of the central unit would be the following:
-analysing the incoming data, -reporting rapidly on the findings.
-establishing a timely and permanent link with policy makers for adjustment decisions,
-identifying the effects of adjustment processes on the food and nutritional situation in the selected areas and suggesting immediate intervention and/or policy options to overcome these effects,
-administering and supervising the surveillance programme and providing training and technical assistance to data-generating units, and
-identifying research needs and developing priorities for the programme.
The key, and somewhat specialized, data-analysis and information-dissemination role of the head of the unit would probably mean that supplementary training in the analysis and interpretation of time-series data would have to be undertaken.
3.4. Research and training
The proposed surveillance system is based on the idea that movements in price and nutritional indicators collected from a limited number of purposively selected sentinel sites, although biased, will reflect similar movements for the domains of study at large. This is because the primary interest is in the movement, not in absolute levels of prevalence, for any particular area or group. Further research is needed, however, to check and study the relationship between data generated from sentinel sites and that generated from probability surveys.
Data may already be available for some countries or regions within countries, which would allow for the comparison of time-series data of clinical observations with sample-survey data. Additionally, it is recommended that a small validative research programme be planned for one or two areas in which survey data (possibly generated by the SDA programme) would be collected in parallel with data from sentinel clinics and compared.
A second area for further research is the relationship and linkages that have been shown to exist between the two main indicator categories (nutritional status and food access) on which the type of surveillance programme discussed here would be based. Current research has suggested that sufficiently strong linkages may exist, at least at the aggregated national level, to make the latter a potentially useful predictive indicator. Further work is needed however to investigate the relationship for different sub-groups of the population (the linkage is for instance likely to be substantially different for poor urban populations than for rural food producers, and in different regions).
With regard to training, mention has already been made of the likely need to provide the senior analyst with supplementary training in the analysis of time series. To assist in this, the compilation of a manual and/or case-study material showing how to analyse time series using actual nutritional-index data from one or two countries representing different infrastructural and socio-economic conditions, is strongly recommended. As the IFNS programme grows in impetus, so further data and examples could be added to the manual which could then provide the basis for periodic regional inter-country workshops.
3.5. Next steps
The state of development of these proposals is as follows:
First, there is already substantial evidence that strong relationships exist between frequently reported (e.g. monthly) indicators (e.g. prices and child underweight prevalences) such that systems using these data, from sentinel sites, for example, may be expected to generate meaningful and timely information. This needs to be verified both with further analysis of existing data and in the field. Extensions to regions other than Africa must be considered a priority. On the technical side, development will entail the follow-up of several regression-based data-analytic approaches, using (i) Almon lagged structures, (ii) transfer function models, and (iii) locally adaptive models. Analyses already undertaken suggest that these various techniques are productive, but at this point it remains unclear which approach will prove to be the most robust and feasible.
Second, the practical operation of the type of minimalist approach advocated here is based on experiences in the health and economic fields in developed countries, on the application of local area monitoring in developing countries, and on clinic-based reporting systems in several African countries. However, the installation and performance of this form of surveillance system-specifically geared to timely decision-making for compensating for economic stress-has yet to be tested. Substantial information on practical issues of data management, trigger levels, data presentation, response times, etc. is therefore unknown at this time. These need to be realistically assessed before the system can be recommended widely with a degree of confidence.
Third, and most important, the assignment of substantial resources to compensate for short-term economic problems during adjustment is only now beginning to take place. It is considered essential to link the surveillance procedures discussed here with actual programmes. This means that efforts to implement surveillance must take place at the time that structural adjustment/compensatory programmes are being planned and acted upon.
From this, the most important next step is to test this methodology in the field, as far as possible with IFNSP support. For example, a useful opportunity now arises with the IFNSP-sponsored Madagascar project.* However, at present the approach outlined in the paper seems sufficiently likely to be useful that it could be considered more widely when information systems are being designed in the context of structural adjustment. This applies especially to SDA programmes in Africa, but may also be relevant where more targeted and timely information is needed to protect nutrition in other regions.
All the same, as noted, much development can nevertheless continue on the basis of existing data.
References
1. ACC/SCN. Report of the International Workshop on Nutritional Surveillance, Cali, Colombia, 14-17 July 1981. (Doe. SCN 82/10) Geneva: ACC/SCN.
2. ACC/SCN. Proposal for expanding nutritional surveillance. Report of the ACC/SCN Working Group on Nutritional Surveillance. (Doe. SCN 87/SURVL) Geneva: ACC/SCN, 1987.
3. World Food Conference. Report of the World Food Conference, Rome, 5-6 November 1974. (Publication E/Conf. 65/20) New York: United Nations, 1974.
4. WHO. Methodology of nutritional surveillance. Report of a joint FAD/UNICEF/WHO expert committee. Technical report series, no. 593. Geneva: WHO, 1976.
5. Mason J, Habicht J-P, Tabatabai H, Valverde V. Nutritional surveillance. Geneva: WHO, 1984.
6. Kelly A. Nutritional surveillance in Europe: a critical appraisal. EURO-NUT report no. 9, Wageningen, Netherlands: Agricultural University, 1987.
7. UNICEF. Strengthening food and nutritional surveillance systems. (Doe. E/ICEF/1987/P.L.37) New York: UNICEF Executive Board, 1987.
8. Beaton GH, Bengoa JM. Nutrition in preventive medicine: the major deficiency syndromes, epidemiology, and approaches to control. WHO monograph series, no. 62. Geneva: WHO, 1976.
9. ACC/SCN. Update on the nutrition situation: recent trends in nutrition in 33 countries, Geneva: ACC/SCN, 1989.
10. Cornia GA, Jolly R, Stewart F, eds. Adjustment with a human face. Vol. 2. Ten country case studies. New York: UNICEF, 1988.
11. Eylenbosch WJ, Noah ND, eds. Surveillance in health and disease. Oxford, UK: Oxford University Press, 1988
12. WHO. Measurement in health promotion and protection. WHO regional publications, European series, no. 22. Copenhagen: WHO, 1987.
13. Kirsch T. Local area monitoring (LAM). World Health Stat Q (WHO) 1988; no.41.
14. Tabatabai H. Economic indicators of access to food and nutritional status. Paper commissioned by the ACC/ SCN and the Inter-agency Food and Nutrition Surveillance Programme. Geneva: ACC/SCN, 1988 (draft).
