The purpose in this report is to assess trends rather than levels of nutrition, and of selected likely determinants. Indicators were chosen with this purpose in mind. The major effort went into obtaining data on trends in prevalences of children underweight, as the main indicator of nutritional status widely available (see discussion in “First Report on the World Nutrition Situation”, FRWNS).
The indicators are presented in the country case-studies on two pages. The first page gives the same indicators for all countries, where available. The second page gives, at the top, sources of cereals (production, import, food aid), use of IMF credit, and the Food Price Index. On the lower part of the page, additional data may be given on child nutritional status. This is particularly relevant when only sub-national data are available, when the lower panel on the first graphics page is blank.
A brief description of the selected indicators and rationale for their choice follows.
Choice of indicators
The new data in this report concerns trends in prevalences of underweight children, or other measures of child nutritional status. All other indicators were directly compiled from data already available through the UN agencies. Three types of common indicators were sought: economic, food, and child nutritional status. Other health indicators were also investigated, as discussed below.
Economic indicators
Rapidly changing economic conditions in the 1980's have had profound effects on human welfare. Recession and structural adjustment, closely related to the burden of external debt in many countries, have interrupted progress or exacerbated underlying problems. One purpose of this update report was to further document the effects on nutrition. Two indicators were therefore selected, from World Bank publications (World Debt Tables, World Bank), to monitor debt. First, the debt service ratio (called “debt ratio” in the panels) is the debt repayment (interest and capital) divided by the value of the exports of goods and services: that is, it is a measure of the proportion of national resources used to pay for external debt. The higher the indicator, the greater the burden of debt. Secondly, the total debt is displayed, defined as debt outstanding and disbursed (called “debt outstanding” in the panels). This is cumulative through time. Both debt indicators are annual.
The estimate of gross domestic product (“GNP” in the panels) is included in a common indicator of the national economy. The value given is based on GNP estimates at constant 1980 prices in local currency1. This gives an indicator of economic performance, and average income available in the country.
1 This indicator was unavailable for Bolivia, Chad, and Cuba, and an alternative version of GNP or GDP was used.Exchange rates are given in the first pages of the panels, as an indicator in its own right, but also to indicate the points in time at which major economic adjustments were made -devaluation being a common feature of adjustment programmes. On the second page of the panels, use of IMF credit is shown: this gives some indication of when IMF- supported programmes were started and the scale of activity involved.
Many other economic indicators are available and relevant, and these can be readily looked up in such publications at IMF's “International Financial Statistics” (monthly), the World Bank's “World Debt Tables” and “World Development Reports” (both annual).
Food indicators
Food consumption by the individual is, with health, a direct determinant of nutritional status. Data on food consumption, even at household level, are rare cross-sectionally, let alone showing trends over time. As for income or expenditure on the economic side, indirect measures were sought. The indicators used estimate different aspects of food availability.
The Food Production Index, which shows relative levels of per capita food production annually, is included for two main reasons. Food production is not itself usually the major determinant of food consumption at national level (although it probably is for farmers producing food, whether for sale or home consumption). Nonetheless, food produced domestically is almost always the main contributor to food availability, as shown in the top left-hand chart in the second page of graphics. Thus, when interpreting changing food availability, changes in domestic food production are important; for farmers, including through income effects. Second, the food production index is for many countries a readily available indicator of the state of the rural economy. Where the food production index falls, a common explanation is drought, and this may be seen from the “unfavorable crop conditions/food shortages” chart within the food indicators. Botswana provides an example: in Botswana 1981 and 1982 were the last years, until 1988, to have adequate rainfall; the fall in the food production index in 1983 - 1987 is some indication of the effects of the prolonged drought. Similarly, the drought in Ghana in 1983 contributed to the crises in 1983/4, but here severe economic problems interacted with drought. Estimates of the food production index were supplied by FAO. A more detailed discussion of the definition is given in the FRWNS (p.44), and in the Supplement on Methods and Statistics (ACC/SCN, 1987, in press).
As cereals are the staple food in most countries, accounting often for 50-70% of kcals consumed, the indicator of cereal availability gives an idea of staple food supply at national level. Major changes in food supply have an effect, through prices and sometimes (though less often) through physical availability of food in the market, on food consumed and hence nutrition. This is particularly important in Africa, and in country after country we see that cereal availability declined sharply in 1983/84. Estimates of cereal availability were calculated from the data on domestic production, imports and food aid, (shown in the second page of the panels) provided by FAO.
The indicator of cereal availability should be seen with the top left-hand chart on the second graphics page, on sources of the cereals available. This breakdown of the cereal availability data into the components of production, imports, and food aid provides some important details on where staple foods are coming from, and illustrates responses to food shortage. It also gives a sense of the scale of the relative contributions to staple food availability. The lags in replacing food production losses are evident1.
1 Drought effects (in Africa) usually correspond with low points in production - of the many examples, Lesotho and Mali provide illustrations. Generally, it is not until the year following the drought that imports have begun to climb to cover the deficit. Often it is two years after the drought that food aid responds fully, and by this time production may have been re-established.Reports of (a) “unfavorable crop conditions” and (b) “food shortages”, as available at the time, are taken directly from reports of FAO's Global Information and Early Warning System. These are included only for Africa. In other regions, this panel is replaced by the food price index, discussed below. They are displayed as “Yes/No” variables, depending on whether a country was scored at that time as affected (a) by drought (occasionally by other factors such as flood or pests destroying crops), and (b) reports of food shortages. These indicators are included to document the course of drought, and as a rough indication of food problems.
Results from food balance sheet calculations are plotted, under the heading “kcals per day”, as a summary indicator of food available for human consumption. These results are related to cereal availability, since cereals provide much of the dietary calories in most countries. The calculations take into account all foods (produced, imported, etc.), other uses of food (e.g., seed, animal feed), and are standardized by population (the unit is average kcals/caput/day). Clearly they do not estimate distributional effects, and in times of difficulty the poor may suffer disproportionately as the equity of food distribution may worsen. Results from food balance sheet calculations were provided by FAO, and a fuller description of the method, and interpretation, is given in FRWNS, and its Supplement.
Purchasing power is a main determinant of household welfare. The relative price of food -food price index as a ratio to the general price index - is a readily available indicator of food purchasing power, at least for the vulnerable population that does not produce food. It is particularly relevant to the urban poor, who may be badly affected by economic adjustment. In fact, the often close correspondence (see Ghana for example) between the relative price of food and observed trends in prevalences of underweight children was a new finding during preparation of this report. For these reasons, the consumer (or general) price index (“CPI”) and the ratio of the food price index to the CPI (“FPI/CPI”) are included as indicators. These data are normally available monthly, reported to ILO, and were obtained from ILO's “Monthly Bulletin of Labour Statistics”. The food price index (“FPI”) is shown separately on the second page of graphics for African countries, on the first page for other countries.
Before discussing nutritional data, the issue of health indicators should be raised. These are included at present only for a few countries (e.g. diarrhoea incidence in Lesotho). One issue is the availability of monthly information, which is patchy; but more important in this context is that conceptually there is no summary health indicator available - except perhaps for child growth. Indeed, child nutritional status is considered a summary health indicator in WHO's monitoring of programme towards “Health for All”. For this report, including for reasons of time and resources, we decided to use disease incidence data illustratively (cf. Lesotho).
The indicator of “underweight children” (occasionally wasting or stunting) is the new feature here: all other indicators are available elsewhere, although not integrated in this way. It cannot be too strongly emphasized that this indicator is not intended as a measure of level, but of trends in the population. An estimate of the level of the prevalence of underweight children is given under the heading of each first page of country graphics. This estimate is taken directly from the calculations reported in FRWNS, in which definitions of the indicator and its calculation are given, (and in more detail in the supplement).
The indicator is defined as % of children, usually 1 through 4 years old, of less than 80% weight-for-age by NCHS/WHO standards. This definition may vary slightly between countries, but this is unlikely to affect estimated trends. Interpretation depends partly on the source of the data, as follows.
Health system-derived data
The sources of child underweight data are usually from health centres; details and methods of analysis are given in FRWNS. It should be noted that the results shown represent many millions of child weighings. Other indicators, when available, are given in the lower parts of page 2 of the country graphics.
Data are presented, both including the seasonal effect and de-seasonalized (by the method given in FRWNS, p.62) to highlight underlying trends.
The major issues in the use of these data to describe trends revolve around the risk of changing selection bias between monthly data points. There is little direct check on this, but there are a number of reasons pointing to the likelihood that these results from health-centre data do reflect real changes.
First, all data obtained from health-centres in which there was selection into the weighing programme based on wt/age (known often as “medical selection”) have been excluded. The data are thus, at least, from a sample of all children attending the health centres.
Second, the results show remarkably consistent and explicable seasonal patterns. In almost every case, a peak of underweight prevalence precedes the harvest, falling again after the harvest. The fact that the peak prevalence may coincide with periods of high disease incidence in the rainy season (e.g. diarrhoea in Lesotho) only strengthens this finding. In some cases, such as Burkina Faso, a more complex bimodal pattern is seen; the late-year peak in fact may correspond to a time when food is being conserved for later needs when labour in the fields is intensive. In others, the post-harvest improvement disappears in times of drought: contrast Botswana in 1980-81 (non-drought years) with 1982 through 1987. In still other cases, such as Rwanda with more continuous rainfall and harvest, seasonality is much reduced.
Third, disaggregation of the data to sub-national areas continues to give both meaningful patterns, and differences that correspond to known events. For example, for Madagascar (see second graphics page), the results for the area around the capital (Antananarive) are compared with an area in the south. Not only are the seasonal patterns maintained with this disaggregation, but it can be seen that deterioration started later in the south, with the drought starting in 1985 - and indeed the seasonal pattern flattened (as in Botswana) in 1986. Similar graphs at area level appear meaningful, for example in Burkina Faso and Ghana.
A fourth thread of evidence supporting the validity of these data in assessing trends came from an unexpected source: close correlations with food price data. By simple inspection, the FPI/CPI ratio in Ghana shows a clear association with underweight prevalence values. Moreover, a few months lag between FPI/CPI changes and prevalence changes appear to exist. A similar picture is clearly seen in Togo, probably in Madagascar and Burkina Faso, and on more careful analysis in Botswana. This is a simplification of a complex situation, (for instance, opposite relationships might be expected for net food producers, or between urban and rural populations) and counter examples exist. But that results from totally independent sources should support each other in this way is highly suggestive that they are meaningful of some underlying reality.
Fifth, a formal check on the influence of one possible source of confounding, that is changing population coverage of health centres, has been done. For data from Burundi, Burkina Faso, Lesotho, Madagascar, and Rwanda, analyses taking account of changing coverage failed to show any consistent influence of coverage on observed prevalence - indeed, the changes over time were if anything sharpened (see Test et al, 1987)1. Less formal observation, for example in Ghana, shows that, although clinic attendance varied substantially over 1981-87, it did not show any obvious relation with prevalences. Thus, the explanation is at least not as simple, in the cases examined, as that more and worse-off people go to health centres in times of exacerbated need.
1 Test, K., Mason, J.B., Bertolin, P., Sarnoff, R., "Trends in prevalences of malnutrition in five African countries from clinic data: 1982 to 1985"; Ecology Food & Nutrition (accepted for publication).Survey-derived data
Data from repeated surveys have been used where available to estimate trends. Here, too, problems arise and have been checked for (as far as possible) in drawing inferences on trends. The main issue again is comparability of samples over time. This comes down to the following.
First, the representativeness of each survey with respect to the same population needs to be assessed. Populations represented by the sample may differ in terms of geographic location, socio-economic status, and age/sex composition. Any of these may confound comparison. Sometimes, as in Indonesia recently, a second survey may usefully give priority to comparability of sample with a previous survey rather than change the population sampled, allowing useful inferences on trends rather than on population levels at one time.
Second, cross-sectional surveys at different seasons are hard to compare (unless there is an independent estimate of seasonal effects). This factor is often ignored if the representativeness issue is resolved, but should be kept in mind. In Madagascar, for example, as much as ten percentage points difference in prevalence is observed seasonally, compared with about a five percentage point year-to-year (de-seasonalized) difference.
Third, different standards and cut-off points may be used between surveys. This can usually be overcome by analysis, particularly if the original data are available.
In this report, a somewhat less rigorous approach than in the FRWNS has been taken. The FRWNS attempted to assess actual prevalence levels, as well as trends, whereas here only trends are investigated. Thus, for example for Bangladesh, although the source document itself trends are investigated. Thus, for example for Bangladesh, although the source document itself points out that four survey results at different times are not fully comparable, we have used the results as a probable indication of trends.
Mixed administrative and survey data
In some cases, for example in the Philippines, data points at different times are derived from both surveys and programme (health-system) sources. It was felt worthwhile to present the data together, but the inferences on trends are particularly weak in such a case.
