Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Ethiopia is one of the most disaster-prone countries in the world. Famines and food shortages brought on by drought have been a major problem through the years, as to a lesser extent have been ones triggered by flood, pests, and livestock diseases. The recorded history of famine and fond shortages in the country goes back hundreds of years, with considerable loss of human lives and destruction of property. As they were at the full mercy of nature, the victims of the earlier disasters were forced to face their full impact
In addition to climate, the roots of Ethiopia's vulnerability to disaster are in its subsistence economy. About 80% of the population are peasant farmers, with another 10% pastoral nomads. Despite the destruction and loss of lives that past disasters caused, it was only in the early 1970s, in the wake of a big famine in northern Ethiopia, that the need for an organized state response to address the immediate needs of the victims was recognized. The government at the time established a Relief and Rehabilitation Commission (RRC) in 1974 to mobilize resources and coordinate responses. The creation of the RRC brought about an awareness of the need for disaster preparedness, encompassing, among other things, the ability to provide advance warnings and to develop effective response mechanisms. Crucial among these was the need for an effective early warning system.
In the approximately 20 years of the RRC's existence, drought and food shortages have occurred almost annually. In the worst years, as in the great famine of 1984-1985, close to 10 million people starved, and in the better years up to 3 million people required support. The Ethiopian early warning system (EWS) was established in 1976, the first of its kind in Africa, with the primary objectives of monitoring indicators of food availability regularly and identifying potential areas of shortages for timely intervention .
Through the years the EWS developed into a fairly reliable system. It identified key indicators of food shortages and established adequate field infrastructure to monitor the performance of the indicators regularly. Its predictions are primarily meant to allow timely pro-positioning of food and other relief items, and, so far, the lead time given has been adequate to take such measures. To a certain extent, trends developed from the data have also been helpful in identifying vulnerable areas for longer-term rehabilitation and disaster-prevention measures.
Methods of predicting food shortage
In a country such as Ethiopia, famine and food shortages have been a part of rural life for years. In response, the people have developed complex coping strategies of their own. Thus predicting famine and food shortages is difficult. Many times, despite severe droughts and crop failures, people have survived with little or no external support.
The complex mechanisms the people have developed to cope with famine mean that neither a few indicators nor conventional methods of assessing food availability suffice to detect impending disasters with any acceptable degree of accuracy. Reliable prediction of food shortages requires that different interlinked indicators be brought together to establish beyond any doubt that a disaster threat exists.
For subsistence crop growers, the process of identifying shortages starts with monitoring crop performances. Since crops are the major source of food, their failure severely constrains a peasant household's capacity to feed itself. On the other hand, unless a household is subjected to successive years of crop failure, which may force it to exhaust all alternative sources of food, it would be rare for a single crop failure to result in a serious food shortage requiring relief intervention. One coping mechanism peasants have developed is to keep a certain number of livestock from which they get milk and which, in bad years, they sell to buy grain. In certain areas, also, the sale of labour and other commodities provides alternative sources of income.
In light of this, a second indicator in areas where crops fail is market price and supply. Prices and terms of trade can reveal people's capacity to rely on their second alternative, to purchase food. Their own food stocks left over from the previous harvest are also an important source but are difficult to measure.
In areas that experience crop failure, it is the depletion of food stocks, whether from people's own production or in the market, together with impaired access to other food sources, that affects nutrition status. As less and less food is available, the level of malnutrition increases. This can be detected over time by establishing a trend of changes in the nutrition status of the population under stress. The trend can provide early signals regarding the severity of a developing food shortage as well as the appropriate timing of an intervention if one is required. A key point is that nutrition indicators should be employed at a late stage of food shortage but in a manner that will ensure adequate lead time to make preparations. It should follow warnings based on information regarding crop performance and prices.
It is more difficult to detect food shortages affecting nomads than those among crop growers. Nomads have much more complex strategies for combating famine and food shortages. Among these is their mobility itself, which is crucial for their survival and which sometimes involves crossing international boundaries. This mobility ensures the full use of alternative grazing and watering points, by which they survive drought years. Although one can make early judgements regarding stress based on meteorological and related data, nomadism is a major constraining factor for making household-level assessments.
As an early signal of impending disasters in nomadic areas, drought monitoring is the only reliable indicator. The market is also a strong early indicator, since terms of trade can reflect distress sale of livestock and acute shortage of grain among other factors. Unlike the case in cropping areas, nutrition indicators may not provide the early signals required for timely intervention. Unfavourable terms of trade, together with the threat of death of their livestock, can force nomads to consume more meat, and this can provide a distorted picture of the nutrition status of the population at a time of great stress. By the time an increase in malnutrition is observed, it could be too late to save lives. Ethiopian nomads also have feeding habits that favour younger children, and studies conducted in disaster-affected areas have identified better nutrition status among children under the age of five years than among older children.
More than accuracy, timeliness is a crucial factor in early warning activities. Hence, most of the data collected are often simple and qualitative. Where they are quantified, as in nutrition assessments, only the most sensitive indicator is monitored, in our case the weight-for-height of children under five. Different early warning indicators are used to serve different purposes at different times. An important requirement is that they be interlinked. As an example of the linkage, some of the major indicators can be put into two groups: one comprising the earliest of the early indicators-crop outputs, livestock conditions, and the like-which can be called food-supply indicators, and a second group, consisting of price, nutrition status, and so on, which are impact indicators. Further subdivision is also possible. Since nutrition status is the last indicator in the chain, it often is taken as a means to fine-tune the warnings the other indicators provide .
The role of nutrition surveillance in early warning
The usefulness of nutrition information as a tool for early warning is often controversial. Many challenge the idea that such information can detect impending disasters with adequate lead time. One of their arguments is that by the time the information gives a distress signal, the victims have already suffered enough. The other major concern is its cost: surveys are expensive to undertake at the level an early warning activity requires.
Despite such challenges and reservations, nutrition surveillance is an important component of the Ethiopian EWS and has been reasonably successful. As its aim is to detect transitory food insecurity, the data collected signal the prevalence of acute or short-term malnutrition among children under five (or those between 70 and 100 cm in height). This is done through a sample survey, which involves measuring the weight and length of the children in the sample.
As the purpose of the surveillance is to fine-tune the earlier forecasts based on food-supply indicators such as crop and livestock information, it is conducted only in areas where the indicators have already given the initial signal. This obviously saves a lot in terms of resources and time. Once the areas for such a surveillance are targeted, the activities involve longitudinal monitoring of levels and changes in nutrition status over time, so that a significantly declining trend can be observed as early as possible to plan interventions.
The method is a cluster survey in which the same clusters are involved every time the survey is repeated and the same children are weighed and measured every time. The primary intention here is to detect changes in the mean percentage weight-for-length (WFL%) of a population group over a given period of time. The frequency of the survey recommended in Ethiopia is monthly or once every two months. So far, most of the surveys are undertaken every two months.
In the standard set for Ethiopia, interpreting the nutrition-status result for a given community is based on the criteria shown in table 1. When the two criteria-mean WFL% and the percentage of subjects less than 80% WFL-classify a population differently, the mean WFL% is preferred since it is more precise. As far as early warning is concerned, therefore, a trend in the mean WFL% approaching 90% should trigger an early intervention. Obviously, a mean WFL% of 95%, together with 5% or less of the population under 80% WFL, should not be a matter of concern, whereas anything under 95% and declining should stimulate the planning of interventions, especially when other food-supply indicators have already given advance warnings.
In Ethiopia a full relief ration distribution is recommended when the mean WFL% is still above 90%. Intervention is required when a declining mean WFL% is approaching 90%, and only in areas where there is such evidence.
An issue of great concern in this regard is the extent to which a survey every two months is sufficient for detecting changes early. Experience shows that it may not be, and, resources permitting, a monthly survey is preferred. In our past experience, there have been occasions when results from bimonthly and quarterly surveys have indicated rapid deterioration in food situations between the survey periods, in which case the purpose of undertaking the surveillance would not be served.
Another potential use of the data, which is not yet widely practiced in Ethiopia, is to help estimate the number of people to be assisted. In countries like Ethiopia, where, as noted earlier, the coping mechanisms of the people are complex and resources are scarce, one of the difficult tasks in relief planning is to come up with a reliable estimate of the number of people to be targeted for assistance. Administratively, the number is often grossly overestimated. Nutrition information is needed to provide a more accurate estimate. The argument is that, as 90% WFL is an appropriate cut-off point, the percentage under 90% reflects the size of the population in need of assistance and that this figure should be used to determine the number of beneficiaries. Once an initial estimate is made, it can be updated and refined as additional surveys are conducted.
TABLE 1. Criteria for establishing the nutrition status of a population
|% of population
< 80% WFL
|> 95||< 5||good|
|< 85||> 20||serious|
Nutrition surveillance in relief operations
The usefulness of nutrition information for relief operations is less controversial. Many agencies employ it effectively to screen relief beneficiaries and to determine the type of assistance to provide as well as when to start and stop food distribution. It also is useful for assessing the impact of relief programmes.
As noted earlier, the targeting of areas for relief depends on a mean WFL% approaching 90%. This indicator helps to identify specific areas for intervention as well as to set priorities between areas. The fact that the percentage under 90% helps to estimate the size of the population to be targeted adds to its usefulness.
A number of operational matters still have to be decided under circumstances for which nutrition information is very important. Needless to say, when warnings are given well in advance and interventions are properly planned, measures beyond general ration distribution may not be required. Ethiopia provides 17 kg of rations per person per month for all ages-15 kg of cereals, 1.5 kg of pulses, and 0.5 kg of oil. Serious concern arises when the surveillance shows that the mean WFL% has fallen under 90% but remains above 85%. In such cases, supplementary food should be provided to vulnerable groups either in the form of dry take-home rations, which for Ethiopia is 100 gm per person per day, or wet feeding from 135 gm per person per day upward depending on the severity of the problem. An increase in the cereal component of ration distribution from 15 kg to 18 kg is recommended if the WFL% is under 85%.
The last but not least major use of nutrition surveillance in emergency operations is in assessing the impact of various intervention programmes. In the same way that a declining trend in nutrition status reflects exhaustion of household food stocks and the emergence of food shortages, an upward trend after successful interventions reflects effective responses. During the emergency phase when relief food is being distributed, regular assessment of the nutrition status of the beneficiary population is recommended to determine the impact of the programme.
It is always important that relief should not create dependency. It should be given only when there is an absolute need for it. Provided that normalcy in food supply is being restored, reduced ration distribution is recommended as the level of the nutrition status improves and approaches the 90% mark.
Selected nutrition surveillance results
Nutrition surveillance has become an important component of the Ethiopian EWS over the last few years. Many non-governmental organizations (NGOs) that provide relief and rehabilitation have introduced it into their regular activities, often to assess the impact of their programmes. The RRC has issued a standard guideline for the collection of information by concerned agencies so that it can be compared and pooled for common use .
All agencies have adopted the guideline. Of the many NGOs working in Ethiopia, the British Save the Children Fund is by far the most involved in this activity. It undertook the programme in collaboration with and to support the government's early warning activities. The following selected results of its surveys are cited to show how the findings have been used to justify or reject relief interventions [4,5]
The district of Menzna Gishe, in central Ethiopia, is prone to food shortages. It produces twice a year- the first harvest in May and the second in December. It also depends to a great extent on sheep herding. In 1992 the May harvest was poor and was followed by major destruction of the December crops in three subdistricts. The May 1993 crops also failed. An assessment made in November 1992 about the 1993 food prospects of the area concluded that the livestock holdings in the area were depleted and that the crop losses would cause serious food shortages. It was therefore recommended that relief assistance should be planned for about 100,000 people. The failure of the May 1993 crops further strengthened this forecast. Relief distribution started in March, but the agencies involved had significant reservations regarding the accuracy of the forecast and, as a result, did not deliver enough food. Table 2 shows the results of the nutrition surveillance conducted in the area. Not only did the trend indicate a decline, but the decline from April to June was statistically significant. This was also the time when distributions were reduced.
TABLE 2. Results of nutrition surveillance in Menzna Gishe
The Merhabete district borders Menzna Gishe. It only produces once a year, in December, and is almost entirely dependent on crops. It is also vulnerable to food shortages. Assessment of the 1991 crops indicated that the district would be self-sufficient in 1992, and grain prices continued to support this conclusion. Despite the favourable forecast, relief assistance was continuously requested. The nutrition assessments conducted in the area revealed the following mean WFL%s: May 1992, 94.7%; July 1992, 95.1%; September 1992, 94.8%. These results indicated a fairly stable nutrition status, and therefore the request for assistance was not accepted.
East Harerghe is a region in the eastern part of Ethiopia that often faces food shortages. The food forecast for 1993, made at the time of harvest in November 1992, indicated substantial shortages in several parts of the region. The forecast was not, however, conclusive for four of the districts-Harer Zuria, Kombolcha, Wobera, and Melkabelo. It was concluded that the situation should be closely monitored in the first two districts, while food shortage was indicated as unlikely in the other two.
Contrary to the forecasts, requests for relief assistance came from these districts early in 1993 but could not be supported until they were justified by nutrition assessments. The assessments indicated the following mean WFL%s: in Harer Zuria and Kombolcha, 94.6% in January 1993 and 93.1% in March; in Wobera and Melkabelo, 95.9% in January and 93.5% in March. The decline in both groups was statistically significant. This, together with the request for assistance, led to the earlier forecasts being revised, and relief assistance was recommended to prevent further deterioration.
Summary and conclusion
The Ethiopian experience demonstrates that nutrition surveillance can serve a useful purpose both in predicting food shortages and in planning and executing relief interventions. The usefulness of the information for directing relief operations, as in the case of targeting areas, screening beneficiaries, and assessing impacts of interventions, is better established than is the case for its reliability in predicting food shortages. The argument against the latter is not so much on technical grounds as on its cost implication.
It is true that nutrition surveillance, particularly at monthly and bimonthly intervals, is costly. However, the cost is often viewed only in terms of the amount of money required to conduct a survey, without taking into account what the surveillance may save through better targeting of areas and beneficiaries and, most important, the human lives it saves through timely interventions. It is true that nutrition information cannot provide as much lead time as such indicators as crop forecasts. Nevertheless, it can refine the forecasts that the early indicators provide, and a reasonable lead time is still possible. The fact that, for early warning purposes, it needs to be conducted only in areas where other indicators have predicted food shortages also allows the coverage of the survey to be reduced, and with this its cost.
As regards its usefulness for directing emergency operations, there seems to be no better alternative. In Ethiopia, decisions on whether or not to start relief operations have frequently been influenced by the results of nutrition surveillance. On several occasions requests for relief food have been rejected or accepted on the basis of evidence such data provided. In disaster management, it is the decision cycle that should reflect the data cycle and not vice versa.
In general it can be argued that, when food and similar inputs are required, an effective emergency operation is inconceivable without information generated through nutrition surveillance. It is required both at the planning and execution stages. The only major concern in making effective use of such an approach is its cost implication. The challenge is to design it in such a way that an acceptable frequency of data collection can be established at a justifiable cost.
1. Gizaw B. Running a national early warning system: the Ethiopian experience. Addis Ababa: Relief and Rehabilitation Commission, 1991.
2. RRC. Guidelines on nutritional status data and food relief. Addis Ababa: Relief and Rehabilitation Commission, 1990.
3. RRC. Nutritional guidelines for rations. Addis Ababa: Relief and Rehabilitation Commission, 1989.
4. Holt J. Lawrence M. An end to isolation: the report of the Ogaden needs assessment study. Addis Ababa: Save the Children Fund (UK)/RRC, 1991.
5. Holt J. Lawrence M. The prize of peace: a survey of rural Somaliland. Addis Ababa: Save the Children Fund (UK), 1992.
Martin W. Bloem, Abdul Hye, Jonathan Gorstein, Marijke Wijnroks, Gillian Hall, Helen Matzger, and Alfred Sommer
As a response to the inability of both governmental and non-governmental organizations to provide vital information during the floods of 1987 and 1988 in Bangladesh, a nutrition surveillance system (the NSP) was established in April 7990. This is a collaborative effort that involves international and indigenous non-governmental organizations and the government of Bangladesh, and is coordinated by Helen Keller International and funded by the US Agency for International Development. During the past three years the NSP has demonstrated an ability to provide regular and dependable information on the prevalence of undernutrition and morbidity in children under five years of age, household socioeconomic characteristics, food prices, and the extent of distress at household and community levels from data collected every two months by NGOs and the government in selected rural districts and urban slums in all regions of the country.
The system was established as a bottom-up surveillance system based on NGO-specific teams with a continuous central quality control system to ensure the collection of reliable data. The NSP has proved to be an excellent tool for policy makers from several sectors, involving health, agriculture, and food aid The NGOs use it for the continuous monitoring of their development programmes and to identify mechanisms through which services can be delivered most effectively. Regional-level analyses evaluated the impact of the universal vitamin A capsule distribution programme. After the cyclone of 1991, the NSP demonstrated its flexibility by expanding rapidly to cover affected districts and provide pertinent information to those involved in relief efforts. Most recently, the NSP has provided information on the role of food prices on nutrition status and assisted the Ministry of Food in its decision-making. The model is worthy of consideration for replication in other countries in the world.
In 1987 Bangladesh experienced one of the worst and longest floods in its history. The flood covered 36% of the country and caused widespread damage that was only partially repaired by mid-1988. In 1988 an even more severe flood affected 61 of 64 districts, covered 84% of the national territory, and directly affected 45 million people.
Bangladesh is one of the poorest countries in the world, and as a consequence morbidity and mortality are extremely high, especially among young children and mothers [1-3]. Frequent floods, droughts, and cyclones invariably result in a deterioration of the already poor health of its people through exposure to contaminated water, crowded and poor sanitary conditions, and, above all, decreased access to food due to crop and employment losses .
The nutrition status of young children is a very sensitive indicator of changes in food supply and health conditions. It also is a reliable predictor of child mortality . Monitoring children's nutrition status in sentinel sites is thus a potentially valuable tool to anticipate, assess, plan, and coordinate the response to continuing and unforeseen crises related to floods and other natural disasters [6-8].
After the floods of 1988 there was lack of appropriate, timely information on health and nutrition status for effective allocation of relief programmes. To meet this need, UNICEF set up a temporary post-flood monitoring system of child nutrition status with the assistance of the United States Agency for International Development (USAID), the US Centers for Disease Control, and the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B). They based their analysis on data collected by non-governmental organizations (NGOs) with nutrition-monitoring activities. The analyses were used to attract disaster assistance and to target limited food-aid resources to areas of greatest need. The data-collection procedures hastily set up during the 1988 emergency were unstandardized and lacked quality control. Although the data were adequate for the time, it was recognized that a more refined system was needed for the future.
The nutrition surveillance project (NSP) began in October 1989 with the intention of establishing a permanent system for monitoring nutrition and health status. The project was initiated by Helen Keller International (HKI) in collaboration with a number of NGOs, the Bangladesh Institute of Public Health and Nutrition, and UNICEF; it receives financial and technical support from USAID and is coordinated by HKI. The overall goal of the project is to create an interactive mechanism for planning, monitoring, and evaluating multisectoral development and relief activities to increase their effectiveness.
Data are collected every two months by ten NGOs and the Bangladesh government from 26 sentinel points, corresponding to 20 subdistricts and 4 urban slums. In each round of data collection, 5,000-6,000 households are randomly selected, and anthropometric measurements are taken from 7,000-9,000 children between the ages of 6 and 59 months. Since the first round, in April 1990, 24 rounds of data have been collected and over 30 reports circulated for wide dissemination.
Although the first six rural sentinel points wore chosen on the basis of their levels of disaster proneness, subsequent subdistricts were chosen on the basis of representation of the five regions of the country: north-west, north central, south-west, south-east, and north-east.
Two different sampling designs are used-one for urban and one for rural sites. In both designs, a total of 400-500 children are measured within each area (subdistrict) selected, making it possible to compare results among sites at the same time and to evaluate the changes in health status over time.
For urban slum areas, where the populations are considered to be homogeneous, a systematic cluster sampling technique is adopted. Data are collected only from slum areas where NGOs implement programmes. At the beginning of each round of data collection, a household is selected randomly from a list of all households maintained by the NGO, and all children between the ages of 6 and 59 months in that household are measured. The team then moves to the next household and continues systematically until the required number of children are measured.
For rural areas, a multi-stage random cluster design is used for the sample selection. In each targeted subdistrict, half of the unions (the next lowest administrative level) are selected randomly. Then, from a list of all villages in the selected unions, 25 are identified. Within each of these villages, 20 children are selected systematically after a random start. On each field visit, the team selects a household from a list maintained by the expanded programme of immunization and visits that household. As in the urban areas, all children within the first household are measured, and then the team visits the nearest household and repeats the procedure until the required number of children from the particular village are measured. Thus, for each subdistrict, 20 children are measured in each of 25 villages (clusters), for a total of approximately 500 children.
Each NGO team is responsible for collecting data in two geographical areas (either rural sentinel points or urban slum areas) where they are currently working. There, the teams measure all children between the ages of 6 and 59 months within selected households every two months. Data are collected on four aspects relevant to disaster preparedness and the prevention of nutrition-related blindness, including nutrition status, health status, socio-economic status, and distress factors.
The assessment of the nutrition status of children 659 months old is an important tool for estimating the degree of distress within a community. Anthropometry is a sensitive indicator of the availability of food and is also a predictor of childhood mortality [6, 9-11]. Three measurements are taken for each child- weight, height, and mid-upper arm circumference (MUAC)-and are recorded together with the child's age and sex. From these data several indicators are created, including weight-for-height (WFH), height-forage (HFA), and weight-for-age (WFA), expressed both as percentages of median values and as standard deviation (SD) scores from the international reference population recommended for international use by the World Health Organization.
One of the major objectives of this programme is to identify a set of indicators that best reflects changes in nutrition status at low cost and can be collected reliably. To determine weight, nude or lightly clothed children are weighed to the nearest 0.1 kg on CMS scales (CMS Weighing Equipment, Ltd., 18 Camden High St., London NW1 OJH, UK), which are regularly calibrated against standard weights. Recumbent and supine lengths are measured to the nearest 0.1 cm on a locally constructed two-track length board. The MUAC is measured to the nearest millimetre using TALC numeral insertion tapes (Teaching-Aids at Low Cost, TALC, PO Box 49, St. Labanas, Herts AL1 4AX, UK). The date of birth is seldom accurately known but is estimated to the nearest day for every child by carefully interviewing the mother. A standard technique was developed using the Bengali calendar.
Four health indicators are used in the NSP: prevalence of diarrhoea, night blindness, acute respiratory infection, and vitamin A capsule distribution coverage. Data on the prevalence of diarrhoea, history of night blindness, and receipt of vitamin A capsules in the previous two or six months are obtained by a history from the mother or adult caretaker of the child. The prevalence of diarrhoea is defined as three or more liquid or semiliquid stools in the last 24 hours. Since August 1991, data have been collected to determine the prevalence of acute respiratory tract infection. The mother or caretaker is asked whether the child has any respiratory tract symptoms at the moment of data collection. Specific questions are asked about the presence of rhinitis, sore throat, cough, difficult breathing, and chest in-drawing. Then the health worker examines the child for any external signs and takes the child's axillary temperature using a digital thermometer. After the child has been quiet and resting for some minutes, the respiratory rate is measured for 30 seconds by one health worker while a second worker checks the exact time by stopwatch. Rapid breathing is defined as over 50 breaths per minute in children 612 months old and over 40 per minute in children 12-59 months old.
Socio-economic conditions provide an important indication of the level of welfare generally. The household characteristics most frequently reported in relation to nutrition and health status are education, land size and ownership, income, and occupation. The indicators selected for the NSP are family size; number of children under five years old; occupation of the head of the household; previous week's salary of the main earner for those who are dependent on daily wages or services; years of education of adult household members; type, size, and actual value of the main living house; and amount of agricultural and homestead land owned. A landless household is defined as one with no agricultural land at all.
General indicators of distress are collected at both the village/urban slum and household levels. Data are collected on the occurrence of natural disasters, such as floods, cyclones, crop damage, and drought. In addition, the market prices of rice, wheat flour, lentils, potatoes, unrefined sugar, kerosine, and soya bean oil are recorded. Household-level data include the sale of household assets to fulfil basic needs, which is a late indicator of household distress. These household assets are categorized into four major groups: general household items, jewelry, livestock, and land. A final distress indicator collected at the household level is the frequency of food loans.
Field worker training
From the very onset the NSP incorporated extensive training-both training for the initial implementation of activities and refresher courses throughout the project. The initial training sessions of the data -collection teams were held in March 1990 and were organized into four one-week periods. Separate protocols were designed for the field supervisors and the assistant field officers. The field workers received detailed instructions regarding anthropometric measurements and administering the questionnaire covering socio-economic, distress, and market price information. Emphasis was placed on ensuring that the workers were aware of the objectives and importance of the surveillance system.
In addition to the initial training sessions, field manuals were prepared and printed both in English and in Bengali. Final manuals have been published for anthropometry, and manuals covering other details of the NSP have been projected. Bimonthly refresher training courses were established for field teams, who come to Dhaka to interact with HKI trainers and share experiences from the field. During these sessions, the results of data collection are presented. Problems associated with data quality are shared, and, where necessary, the sources of problems identified and resolved.
One of the most important attributes of the NSP is the importance placed on data quality. For each round of data collection, an HKI monitoring team visits the field sentinel points for each NGO. Each team is responsible for two to three NGOS, checking and calibrating equipment, and supervising the data collection and anthropometric measurements. Detailed monitoring check-lists are maintained to verify whether appropriate techniques are being employed for each component of data collection.
In addition, during each round of data collection by the NGOs, a quality control team conducts random, unannounced visits to the sentinel points. The teams select a 5%-10% subsample of the children who have been measured and repeat the measurements the next day. The data collected by the HKI anthropometrist are compared with those of the NGO field worker, and the differences in observations are registered. The average difference for weight, height, MUAC, and age are detailed for each NGO team. The mean error (for each team) and standard deviation (across all NGO field teams) are compared every month for the four measurements. With the exception of the first data-collection round, which was experimental, the accuracy of the measurements has been well within acceptable ranges of variability.
A key component of the NSP is the timely reporting and dissemination of data, which is heavily dependent on appropriate mechanisms for the rapid processing and analysis of data. This has been no small task, given that the data collection involves a number of independent NGOS, each operating in different regions of the country. To facilitate data processing, a standardized data entry/management software package was developed by HKI, and all data entry operations are undertaken by clerks hired exclusively for the NSP in the field offices of the associated NGOs.
The data entry programmes have a number of quality controls, including validity checks, duplicate detection, and verification procedures, written in FoxPro and provided to each NGO at the time of clerk training, when all programmes were introduced and standardization exercises performed. A series of five data base files is prepared for each round of data collection for each NGO. Once completed, the files are transferred to the HKI office in Dhaka, where they are converted to SPSS system files, from which all analyses are executed.
Once in the central office, the data files are all joined into a single flat file for each round of data collection and are submitted to another series of verification checks to ensure that the proper codings of subdistrict and union identification information have been recorded. If any combination of these codes is incorrect or any other data variables are in error, the data are edited and the NGO teams responsible for the problems are notified.
The prevalence of undernutrition is computed for the various areas using cut-off points for each of the four anthropometric variables. The proportion of children falling more than 2 and 3 SD below the NCHS reference median are computed for each of the three indices, WFH, HFA, and WFA, and the proportion of children with MUAC values below 125 mm is also calculated. Results are presented for each sentinel area for all children between the ages of 6 and 59 months. In addition, the percentage of children receiving vitamin A capsules and the prevalence rates for night blindness, diarrhoea, and acute respiratory infection are tabulated. For each round of data collection, a series of socioeconomic variables is summarized for both the urban slum areas and the rural sentinel sites, including education, salaries, distress sales, and values of sales, food expenses, and food loans. Finally, data on the market prices of four basic foods, rice, wheat flour, lentils, and soya bean oil, are analysed for each of the sentinel points included in the NSP.
Data for other socio-economic characteristics and distress factors are maintained in the analysis files and can be used for special presentations and in supplementary analyses of the causes of undernutrition. After the data have been analysed for the summary reports, the labelled files are stored on high-density diskettes in a series of catalogued boxes.
Every two months a report is compiled on the data collected from the previous round. From 1992, interested organizations have been sent a summary paper that highlights the key findings. Organizations and interested individuals who would like further information can contact HKI for more detailed results.
Figure 1 shows the results of the quality control measurements on weight and height over the last 22 rounds. In earlier stages of the project, certain teams were identified as having larger mean errors for some measures relative to the other teams. However, these problems were quickly identified, and measurements improved in subsequent rounds. The later rounds had a few individual NGO data quality problems that were quickly noticed and rectified. This experience underscores the importance of maintaining continuous data quality exercises over time to ensure the quality of the results.
The following results are examples of what the system has been able to accomplish since 1990.
FIG. 1. Mean errors in weight and height data (differences between NGO data and HKI quality control measurements), June 1990-December 1993
Figure 2 shows the malnutrition rates measured by WFA less than -2 SD, in comparison with changes in rice prices over the past four years. The seasonal pattern in rice prices reflects the drop after the aman harvest in December and the bore harvest in June. Furthermore, the trend over the four years shows a steady increase of the price after the June results until June 1992, when the bore harvest was very good and as a result there was a free fall of the rice prices until June 1993. The malnutrition rates show a similar pattern. The seasonal pattern means that the rates of malnutrition are highest in October, just before the aman crop, and lowest in February, just after the harvest. The trend over time follows the trend of the rice prices, but a little bit more slowly. The malnutrition rates increased in the first three years but dropped to a level below that at the start of the surveillance.
The nutrition status of a population can be described by a normal distribution curve, which is mathematically expressed in terms of mean and SD. This method has the advantage of describing the nutrition status of the entire population directly without resorting to a subset of individuals below a set cut-off . We found that the change in the food prices not only changed the percentage of malnourished children but was responsible for a shift to the right of the nutrition status of the whale population. The difference in the mean of WFA between June 1992 and June 1993 is 0.25 SD, which is the reflection of a drop in the price of rice from 12 to 8 take.
Figure 3 shows the WFA means for the various rural subdistricts between June 1990 and August 1994. Regional and programmatic variations within Bangladesh became apparent. The ICDDR,B has had a community health programme in the Matlab intervention area for the last 20 years. The difference of 0.10 SD between the Matlab area and Daudkhandi, its control area, which are adjacent to each other, can be considered as the impact of a health intervention on nutrition in Bangladesh.
Figure 4 shows the mean of WFA by month for the five socio-economic groups in the rural areas. Rural Bangladesh can be divided into five groups based on land ownership: landless (41.5%), marginal (00.5 acres, 15%), small (0.51-2.50, 29%), medium (2.51-5.0, 9%), and large (<5.0, 5.5%). The structure of farms in Bangladesh is such that farmers (56.4% of the total rural population) are both producers and consumers of rice. The International Food Policy Research Institute has shown that only the medium and large farmers (14.4% of the total rural population) produce surplus rice in both good and bad years [11,12]. About 90% of the marketable surplus is produced by these groups, who can be called net producers in the neoclassical framework of price analysis [11,12]. This means that 85.5% of the rural population are net purchasers of food grains. Although the nutrition status of the higher socio-economic groups is better (0.5 SD), their children are nevertheless faltering between the ages of 6 and 12 months, with more than 1 SD. Although the lack of good-quality weaning food, the high rate of infectious diseases, and extremely poor nutrition status of the mothers are important contributory factors, poor caring practices must play a crucial role in the aetiology of this phenomenon.
FIG. 2. Prevalence of underweight (weight-for-age <-2 Z-scores) in children 6-59 months old, and the price of rice-Bangladesh, June 1990-August 1994
FIG. 3. Underweight (mean weight-for-age Z-scores) in children 6-59 months old, by subdistrict, June 1990-August 1994
The success of a surveillance project depends on the goal of the surveillance system . In Bangladesh general development is hampered by poor socio-economic conditions and recurrent disasters. Although the surveillance system was set up to monitor and evaluate multisectoral development, it had to have an early warning component to respond to both natural disasters and economic stress related to structural adjustment. Since nutrition is an outcome of several components (e.g., food security, health, caring practices), the project was designed to collect information to quantify the relative importance of the various components during certain seasons or years.
Many obstacles stood in the way of implementing a successful nutrition surveillance system. First, the technicians and information experts involved in surveillance are not the decision makers. Furthermore, it was difficult to ensure that the surveillance data would be accessible to policy makers and the information would be useful in improving the delivery of health services because of a number of factors, such as incomprehensible output, lack of timely information, and misunderstanding of information needs at different levels.
FIG. 4. Underweight (mean weight-for-age Z-scores) by age and household land-ownership, June 1990-August 1994
The NSP was set up as a decentralized system, taking into account the pitfalls of previous nutrition systems. One important characteristic of the project is the involvement of NGOs in the surveillance activities, which are instrumental in assisting the government to provide preventive and therapeutic services to communities. Most of the NGOs have an integrated, community-based approach toward development activities. The data generated by the NSP provide them with rapid information concerning health and nutrition status in the communities where they are active, thereby greatly improving their familiarity with general conditions and providing a sensitive indication of changes occurring over time so that responses can be taken at the local level. Furthermore, the project greatly increases the general awareness among the NGOs and donors of both the magnitude and distribution of various indicators, which leads to rapid decisions and actions. Recognizing that a disadvantage of a bottom-up approach is the difficulty of guaranteeing the quality of the data collected, the NSP has concentrated on technical improvements in data collection, processing, and analysis.
Data are collected and reports produced at regular two-month intervals and more frequently during disaster periods, providing a continuous assessment of whether conditions are stable, deteriorating, or improving. Information is also produced on a regular basis for decision-making at the policy or programme level. The NSP is designed so that the information that is produced is relevant to Bangladesh and is meaningful to officials and programme staff at all levels as well as to the lay people. It is well known that assessing the nutrition status of children 6-59 months old is the best means of rapidly estimating the degree of distress in a community. The specific measures selected for use in the NSP are relatively easy to obtain, can be highly standardized, and are sensitive indicators of food availability and predictors of childhood mortality.
The NSP has demonstrated the capability to respond quickly to the unexpected needs of users brought about by special circumstances. HKI extended surveillance activities to coastal areas devastated by the 29 April 1991 cyclone. Special teams were rapidly recruited and trained to carry out continuous monitoring in the area, using key indicators that provided information for establishing effective allocation of relief and food aid. During this period, the NSP produced weekly reports that were distributed to NGOs, donors, the government of Bangladesh, and other groups involved in the relief efforts, so that each could be aware of changes occurring in the area.
Besides those elements of NSP projects, the strength of the system is that it is able to link macro-economic information (food prices, vitamin A capsule delivery, floods) with micro-economic information. The results discussed here are good examples of what the system has been able to accomplish in the past four years.
Although most experts agree that nutrition status is an outcome of food security, health, and caring practices, many debate the relative importance of these components. As mentioned earlier, describing a population by its normal distribution curve, mathematically characterized by the mean and standard deviation, is a good approach to describe the entire population. This way of describing the data not only points to those below the cutoff but shows that a whole population is affected. The NSP has quantified some of the additional value of these components using the means and standard deviations of certain anthropometric indicators .
These indicators help in making effective decisions. The in-depth analysis of the impact of rice prices on malnutrition helped the Ministry of Food to make decisions on several issues concerning food aid and food subsidies. A recent analysis of the effectiveness of the vitamin A capsule programme led the Ministry of Health to change the distribution schedule to make the programme more effective. Analyses of family size and gender issues were and are being used by both the government and NGOs in their preparations for the United Nations conferences on population (Cairo, 1994) and women (Beijing, 1995) [14, 15].
In the past four years the NSP has developed into a useful tool for policy makers at the grass-roots as well as the national level. In the next six years the challenge to the NSP will be to institutionalize itself in a way such that both the private sector in the form of NGOs and the government of Bangladesh can participate and benefit from it equally.
This study was carried out as a part of the Nutritional Surveillance Project for Disaster Preparedness and Prevention of Nutritional Blindness under grant No. 388-0083-G-SS-9127-00 between the US Agency for International Development (USAID), Dhaka, and Helen Keller International Bangladesh.
1. BBS. Report of the child nutrition status module: Bangladesh household expenditure survey, 1985-86. Bangladesh Bureau of Statistics, Government of the People's Republic of Bangladesh, Dhaka, Bangladesh, 1987.
2. Chowdhury AKMA. Child mortality in Bangladesh: food versus health care. Food Nutr Bull 1988;10(2):3-8.
3. Schulz LO, Csele J. Nutrition concerns in Bangladesh: the focus for improvement. Progr Food Nutr Sci 1990; 14:259-76.
4. Roy SK, Haider R. Is nutritional status deteriorating in Bangladesh? Health Pol Plan 1988;3(4):325-8.
5. Stewart NM, Fauveau V, Chakraborty J et al. Post-flood nutritional anthropometry of children in Matlab, Bangladesh. Ecol Food Nutr 1990;24:121-31.
6. Mason J. Habicht JP, Tabatabai H. Valverde V. Nutritional surveillance. Geneva: WHO, 1984.
7. SCN/AGN. Suggested approaches for nutritional surveillance for the Inter-agency Food and Nutritional Surveillance Programme. Food Nutr Bull 1989;11(2): 62-73.
8. Tucker K, Pelletier D, Rasmussen K et al. Advances in nutritional surveillance: the Cornell Nutritional Surveillance Program 1981-1987. CFNPP Monograph 982. Ithaca, NY, USA: Cornell University, 1989.
9. WHO. Methodology of nutritional surveillance: report of a joint FAO/UNICEF/WHO expert committee. WHO Technical Report Series, no. 593. Geneva: World Health Organization, 1976.
10. Yip R. Victora C. The use and interpretation of anthropometry: report of the Subcommittee for Children. Geneva: World Health Organization, 1993.
11. Ahmed AU. Food policy in Bangladesh: an analysis of economic efficiency and distributive justice. Doctoral thesis, Colorado State University, Fort Collins, Colo, USA, 1990.
12. Ahmed AU, Bernard A. Rice price fluctuations and an approach to price stabilization in Bangladesh. IFPRI Report no.72. Washington, DC: IFPRI, 1984.
13. Beaton GH, Kelly A, Kevany J. Martorell R. Mason J. Appropriate uses of anthropometric indices in children. ACC/SCN State-of-the-Art Series, Nutrition Policy Discussion Paper no. 7. Geneva: ACC/SCN, 1990.
14. Government of Bangladesh. Bangladesh country report to the International Conference on Population and Development, September 5-13, 1994, Cairo, Egypt. Dhaka, Bangladesh: Ministry of Health and Family Welfare, 1994.
15. Government of Bangladesh. Women in Bangladesh: equality development and peace. National report to the Fourth World Conference on Women, Beijing, 1994. Dhaka: Ministry of Women and Children Affairs, 1994.
Contents - Previous - Next