Integrated Nutrition and Food Security Surveillance in Malawi
By Elena Rivero, Núria Salse and Eric Zapatero
Elena Rivero is currently working for Action Against Hunger as Surveillance Advisor in the Malawi Integrated Nutrition and Food Security Surveillance System (MINFSS). Since 2000, she has been involved in humanitarian aid and nutritional development programmes in several countries, including Algeria, Colombia, Central America, Sudan (Darfur) and Lebanon.
Núria Salse is the Health and Nutrition Officer in Acción Contra el Hambre. Previously she spent several years working on nutrition and medical programmes in Angola, Guinea Conakry, Niger and Argentina.
Eric Zapatero is the Food Security Officer in Acción Contra el Hambre. Previously he spent several years working on food security in Burundi, Tanzania, Ethiopia and the Democratic Republic of the Congo.
The authors would like to acknowledge Neil Fisher who designed, implemented and has provided continued support to developing the systems methodology. They would also like to acknowledge Phil Mckinny, Silke Pitsh, Brad Palmer, George Mvula , Maja Munk, Mercy Chikoko, María Bernandez, Jacob Asens, Rebecca Brown, Marta Valdés and Peter Hailey who contributed to launching and improving the surveillance system. They would especially like to acknowledge the work of the Ministry of Health and the Ministry of Agriculture of Malawi in implementing the methodology.
This article describes a surveillance system developed and implemented by AAH and partners in Malawi, that monitors nutrition and food security (through a food stress index) in a sample of children under 5 years at sentinel sites. The development of user friendly software programmes for data analysis is facilitating the integration of the system into government structures.
The Ministry of Health and Ministry of Agriculture of Malawi, in partnership with Action Against Hunger (ACF), have been running the Malawi Integrated Nutrition and Food Security Surveillance (MINFSS) system since May 2003. The system was initially piloted in six districts, but currently covers the whole country. The MINFSS system aims to provide information on nutrition trends of children under five years, as well as the household food security situation in Malawi. Any significant decline in the nutrition and food security situation detected by the system should lead to a detailed investigation, using standard nutrition survey and food security assessment methodologies. These would estimate the actual prevalence of acute malnutrition or the level of food security.
The MINFSS system monitors nutritional status of a sample of 9,100 children (350 per district) from five growth monitoring clinics (GMC) in each of the 26 districts in Malawi. These children are randomly selected from a population of children attending the GMCs and therefore include healthy, malnourished, and sick children. Out of the 9,100 children, 1,300 children (50 per district) are selected to gather household food security information. These same children are followed over the period of a year.
NUTRITION Information Collection and Analysis
At each of the GMC sessions, the following information is collected from the 70 selected children:
- Child Identification
- Child age in months
- Height (using a height board to the nearest mm)
- Weight (using a Salter scale to the nearest 0.1 kg)
- Mid-upper-arm circumference (MUAC) (using MUAC tape to the nearest mm)
- The occurrence of acute diarrhoea in the past 2 weeks
- Presence or absence of bilateral oedema
Five health centres are selected within each district to ensure that all livelihood zones are covered. The centres are called sentinel sites. In each sentinel site, a number of GMC sessions are conducted with caretakers of children under 5 years of age who regularly attend these clinics. In order to ensure all children attending a GMC have an equal chance of being selected, a random sampling is carried out at all the GMC sessions in a particular month.
The MINFSS system aims to create an information channel from rural Malawi Health Centres and Agriculture Extension Development Officers (AEDOs) through to their respective ministries in Lilongwe where data are entered and analysed. Results are reported to national and international decision/policy makers. Data are sent to Lilongwe at the end of each month and the 10th of the following month is the cut-off point for these data being received at Ministry offices. These data are entered into the Nutrition and Food Security databases.
The anthropometric data are analysed with ANALYNUT (see box) to produce anthropometric indices, using the National Centre for Heath Surveillance (NCHS) tables of reference. The analysis uses repeat measurements by pairing the same child from the previous month in order to make comparisons over time.
The weight for height z-score (WHZ)1, as an indicator of acute malnutrition, is of primary interest. The proportion of children with GAM (global acute malnutrition) and SAM (severe acute malnutrition) are calculated. Data on trends in height for age z-score (HAZ)2 are used as an indicator of chronic malnutrition or stunting. This information is disseminated once a year but can be made available upon request. Other nutrition indicators, such as the proportion of children with low mid upper arm circumference (MUAC), diarrhoea and oedema are also calculated using the ANALYNUT programme.
ANALYNUT can be used to calculate monthly trends by extracting the relevant information from the
monthly data files and calculating the mean and standard deviation for each sentinel site that are
entered onto a worksheet. As a second stage, ANALYNUT carries out an adjustment for missing sites. The
programme estimates a value for every child included at least once in the run of months for which the
analysis is set up. This is done for every missing site in a district. The programme essentially calculates
a best estimate of the value for each missing cell in a reiterative fashion using a missing data procedure.
The programme provides a best estimate for the district based on the available data of previous months.
For those districts with missing sites, the best estimates are only calculated for the mean, as confidence
interval calculations need information on the number of children.
ANALYNUT is an ideal data analysis package from the point of view of capacity building of the Ministry
of Health, since the programme is easy to use and can be run by individuals that have limited data analysis
experience. The programme is purely Excel based (i.e. data entry, cleaning, flagging of biologically
implausible indices, analysis, and output is all carried out in an excel programme). The results are
comparable with other anthropometric data analysis packages such as EPI NUT contained within the Epi-
Info programme (http://www.cdc.gov/epiinfo/) and Emergency Nutrition Assessment (ENA) Software
More information on the ANALYNUT programme is available from the authors (see contacts details at the
end of this article).
FOOD SECURITY Information Collection and Analysis
Once the sampling has been carried out, a baseline survey is conducted to gather household food security related information. The main topics covered are:
- Structure of the household
- Assets ownership
- Land, crops and cultivation practices
- Cash income and income sources
- Food consumption and preferences
- Sickness and health
- Water and sanitation
A shorter questionnaire is used on a monthly basis to monitor changes occurring in the household. It includes:
- Changes in the household (births, deaths, movements.)
- Cash income and income sources
- Food availability
- Food consumption patterns
A sample of 10 children is selected from the list of the 70 children enrolled in nutrition follow-up. The selection is random using a sampling interval procedure. This child's household is then assessed via questionnaire each month to provide information on the main food security indicators. This sampling method allows for tracking changes over time and takes account of the need to combine nutritional and food security indicators in the same locality.
The analysis of the food security data is conducted using two programmes developed by ACF for this project. These programmes were developed in Excel using Visual Basic in order to make it more accessible for users.
Surveyprogramv10 is used for:
- Simple accumulation of coded categories
- Mean, standard deviation, median and quartiles
- Accumulation of coded categories with more than one per cell
- Mean, median, etc with zero values ignored
- Accumulation of string variables
It is used for cleaning the data and to calculate the Food Stress Indexes. It can also be used to cross tabulate sets of data and calculate means and medians, etc.
Monthstrendsv5 is the latest version of a programme that is used to calculate monthly trends. It extracts the relevant information from the monthly data files, and assembles the mean for each sentinel site onto a worksheet. In a second stage, it carries out an adjustment for missing sites. Through the analysis of the data it is possible to follow several food security indicators each month in each district. It is also possible to compare the data between districts, regions and years.
Food Stress Index
The nine indicators from the monthly questionnaire which showed a discernible trend over the hunger period were combined into a 'food stress index'. This provided a useful summary of food insecurity for the 2004-05 hungry season and allowed comparison with 2003-04.
Food Stress Index mark 1(FSI1)
Nine of the variables which showed significant differences between months were combined into a food stress index. The percentage of households reporting each indicator is shown in Table 1. The mean of the nine indicators is the food stress index mark 1 (FSI1). A value of zero would imply that all households:
- had more than 50kg of cereal in store
- had root crops available
- had a cash income greater than MK1000 per month
- ate three meals per day
- ate cereals every day
- ate groundnut or other legume every day
- did not have to undertake any form of food rationing.
At the other extreme, a value approaching 100 would indicate that few households experienced such conditions.
However a number of problems became apparent with the FSI1 indicator. The FSI1 was too heavily based towards maize whereas in some areas, other staple foods have a very important role, e.g. cassava in Nkhata Bay. Following discussion with the Malawi Vulnerability Assessment Committee (MVAC), it became apparent that new indicators needed to be included in the index. For example, ganyu, the system of casual piecework which sustains the most vulnerable non-food-selfsufficient households through the hungry season. Also, the emphasis placed on meal frequencies and enforced rationing (reduced portions and 'whole days without a staple food') was felt to be too high.
Food Stress Index mark 2 (FSI2)
Given the limitations of FSI1, a new index was developed, the 'food stress index 2' (FSI2). Eight indicators were combined into the FSI2. Adjusted weightings reduced the importance of food consumption indicators relative to the availability of starch food (cereals and root crops) and cash earnings. FSI2 is better able to deal with districts where cassava is one of the staple crops. It also includes information on ganyu, which was not available from the questionnaire in use b e f ore October 2004.
The following eight indicators were proposed for the new food stress index (FSI2):
- The proportion of households that have very low supplies of starch staple food: less than 20 kg of maize, other cereal or dry cassava and no cassava or sweet potato ready for harvest (weighted 1.0).
- The proportion that have a potential shortage in the longer term: less than 50kg of maize, other cereal or dry cassava and no cassava or sweet potato ready for harvest in the next two months (weighted 1.0).
- Households with income less than MK1000 per month (weighted 1.0).
- Households having difficulty finding ganyu (weighted 1.0).
- 100 x (3 minus meal frequency) (weighted 0.33).
- Households who have not eaten groundnut or legume on the previous day (weighted 1.0).
- Households reporting that they did not have enough food at some time in the month (weighted 0.33).
- Households going whole days without a staple food (weighted 0.33).
The first two indicators are closely connected but including them both emphasises the importance of immediately available food to the index. The meal frequency indicator operates on the assumption that most households would eat three meals a day if they had adequate food. Occasionally households report 4 meals per day but this is treated as 3 meals. Indicator 6 on legume consumption is indicative of food quality as well as changes in food access. Having experimented with various weights for the different indicators it was decided to downgrade the importance of the three indicators of food consumption (meal frequency, not enough food and whole days without staple) so that the three together carry the same weight as any one of the other indicators.
Each month, a Bulletin is issued with the results of the data analysis for both the Nutrition and the Food Security information. All reports and bulletins are shared with the ministries and any other interested institution. Results are also sent to the district representatives of the ministries and presented at Nutrition and Food Security meetings.
|Table 1: Percentage of households reporting each indicator for Food Security Index mark 1 (2003/04)
|Less than 51kg in store
|100-access to cassava/sweet potato
| 100-cereals taken
|Entire days without staple
Challenges and Constraints
Possible biases of the methodology
Defaulting is an issue that arises and may be a source of bias because the same children are followed over time. If a large number of children default, the sample may cease to be random. However, for the purposes of tracking trends over time, the system needs a broadly representative sample rather than a random one. The statistical theory and methods are essentially those of time series rather than of surveys. All that is necessary is that the sample should be representative of the population.
Another potential source of bias is the nature of clinic attendees. For example, if there is a tendency for especially sick children or children of sick mothers to attend the GMC, sick children will be over-sampled. However, a counterbalancing factor may be that that more health conscious caregivers attend the GMC.
Another potential bias is introduced by sampling children at the health centres and not the outreach clinics, so that more remote settlements are under-represented resulting in wealthier families being over-sampled. However this does not appear to be the case. At the start of implementation, both static and outreach clinics were used in each district for the follow up of the nutrition and food security situation. Following an evaluation of the pilot, it was concluded that there was no significant difference between the population sampled from the outreach posts and those sampled from static GMCs (for both Nutrition and food Security).
Family's tire of being asked the same food security questions each month without any kind of 'reward' In addition, it is sometimes difficult for AEDOs to complete the questionnaires, resulting in low questionnaire submissions rates.
Despite our best efforts, the dataset is at times incomplete, as completed questionnaires from specific sites in specific months are not always received in Lilongwe. There are two problems that affect data return. First, there are not enough children to be measured/monitored. Secondly, not all data are used due to poor quality. Insufficient numbers of datasets received every month determines that, currently, comparisons between regions are more useful and give a more accurate picture of the situation than district by district comparisons.
|Table 2: Components of the Food Stress Index mark 2
|Starch not available now
|Starch not available soon
|Income < MK1000
|Difficult to get ganyu
|3 minus meal frequency
|No legume previous day
|Not enough food
|Whole days without staple
The surveillance system implemented in Malawi has several important practical and statistical differences from a survey based system. The same children are measured over months as a time series whereas surveys use a point time data analysis methodology. Surveillance is effectively a moving picture while a survey is a snapshot of a movie. The sample size for surveillance is different from a survey. The pairing of the data (following the same children each subsequent month), requires a smaller sample size to produce good statistical power and precision in demonstrating significant changes over time. Currently the nutrition surveillance system at its maximum, providing a sample size of 350-paired data per district.
A priority aim of this surveillance system was to understand the linkage between nutrition and food security. Thus, a sub-sample of those households with children monitored under the nutrition surveillance component was chosen for food security assessment monitoring. It was hoped that this would clarify the impact of the food security situation at household level on the nutritional status of the children.
Measuring MUAC in a GMC session
The Surveillance System is proving an important source of information for all stakeholders and key decision makers. This was demonstrated during the food crisis of 2005 when the Integrated Nutrition and Food Security System provided information which allowed timely intervention when the situation appeared to be deteriorating.
The surveillance system is also providing useful and pertinent information to the MVAC3 in Malawi. It provides secondary data on food security and nutrition indicators which are used by the MVAC for the annual assessment report. This information is provided at harvest time when the MVAC is analysing the situation in Malawi. The system is also used by the MVAC to track the situation over time and monitor the predictions made.
Strategies are currently in place to ensure a smooth and sustainable integration of the system into governmental structures. Comprehensive handover to the Ministry of Health and Ministry of Agriculture is in progress and will hopefully lead to full Government ownership.
MINFSS applies statistical tests that are based on analysis of variance of WHZ from repeat measurements on the same children. This pairing of the same children improves the precision of the analysis in demonstrating changes over time, reduces the need for large samples and is more capable of showing a change in acute malnutrition over time. The food stress index allows comparison of food security between months (is it improving or faltering?), between years and between geographic units (regions, districts or livelihood zones). It does not predict the situation that will emerge later in the year nor does it provide an exact description of the situation in each district. However, after some years of running the system, it should be possible to predict the scale of deterioration in a 'hungry' season as there will be data over a number of years on which to draw. For the sake of continuity of the information and in order to establish valid trends over time, it is recommended that the same households are selected for one year and a new set of households selected thereafter on an annual basis.
For further information, contact: Núria Salse, email: email@example.com and Eric Zapatero, email: firstname.lastname@example.org tel: 00 34 91 391 53 00
1WHZ = (actual weight - reference weight)/reference standard deviation
2HAZ = (actual height - reference height)/reference standard deviation
3MVAC (the Malawi Vulnerability Assessment Committee) is a consortium committee of government, NGO and UN agencies that is chaired by the Ministry of Economic Planning and Development. MVAC members have contributed to livelihood zoning of Malawi. Livelihood zones feature baselines profiles that are monitored and re-assessed yearly.
Taken from Field Exchange Issue 33, June 2008