New method for assessing acute malnutrition in nomadic pastoralist populations
By Anne-Marie Mayer, Mark Myatt, Myriam Ait Aissa and Nuria Salse
Anne-Marie Mayer is a technical consultant for this project and carried out the first field test in Mali with ACH. She has a PhD in International Nutrition from Cornell University and has worked in Ethiopia with Save the Children UK on pastoralist nutrition surveys.
Mark Myatt was the statistical consultant for the project. He is a consultant epidemiologist and senior research fellow at the Division of Opthalmology, Institute of Opthalmology, University College London. His areas of expertise include infectious disease, nutrition and survey design.
Myriam Ait Aissa is the research manager at Action Contre la Faim- International Network (ACF-IN) since 2007. Previously, she worked as a research manager on food security issues in North Africa, South America, Canada and in research management.
Nuria Salse is the Nutrition and Health Advisor for Accion Contra el Hambre (ACH) (Action Against Hunger, Spain). Previously she spent several years working on nutrition and medical programmes in Angola, Guinea Conakry, Niger and Argentina.
The authors would like to acknowledge ACF-International Network for financial and managerial support, ACF-Spain for hosting the project in Mali and the peer review group for reviewing the method and providing information on pastoralist communities: Andrew Hall (Westminster University, UK), Phil McKinney (Consultant), William Kalsbeek (University of North Carolina, USA) , Jon Pedersen (Institute for Anvendte Internasjonale, Norway), Fiona Underwood (Reading University, UK), Tina Lloren (Save the Children USA), Francesco Checchi (London School of Hygiene and Tropical Medicine, UK), Megan Deitchler (FANTA), Andre Briend (WHO), Chris Leather (Oxfam GB), Claudine Meyers (Oxfam GB, Kenya), Orla O'Neill (Concern, Ethiopia), Grainne Maloney (FSAU, Somalia), Peter Hailey (Unicef, Kenya), Filippo Dibari (Valid international), Lio Fieschi (Valid International), Fabienne Nackers (MSF Belgium), David Crooks (Tear Fund). The authors also acknowledge the survey team and ACF administration in Kidal, Mali and the communities of Kidal for participating in the survey.
This article describes a new survey method for assessing acute malnutrition in nomadic pastoralist populations, including a case study from Mali.
The work presented here was commissioned by Action Contre la Faim-International Network (ACF-IN) and took place between May 2007 and June 2008. The research aimed to identify a novel method to assess the nutrition condition of pastoralist communities in countries where ACF operates: Mauritania, Burkina Faso, Mali, Niger, Chad, Sudan, Ethiopia, Somalia, Uganda, and Kenya. Pastoralist populations are vulnerable to shocks that result in nutrition risks, e.g. drought, animal disease, market disruption and closure of borders. However, the lack of a suitable sampling method for pastoralist surveys has contributed to the omission of pastoralist populations in emergency response and development programmes.
The two main challenges in surveying pastoralist populations are:
- Case definition for wasting (the usual weight -for-height (WH) based case definition has returned higher prevalence's of malnutrition than Mid Upper Arm Circumference (MUAC) in pastoralist children over 24 months of age)1.
- Selecting a representative sample in an area with a mobile, low density population for whom there are few reliable data on population size at the community level.
Existing survey methods and their limitations in pastoralist settings
Most surveys use a two-stage cluster sampling method, e.g. 30 by 30 cluster surveys and SMART surveys2. Both methods use primary sampling units (PSUs) selected using Probability Proportion to Size (PPS). This weights the sample according to community size, favouring large communities but does not ensure a geographically representative sample. In pastoralist areas, community-level population data are not available and hence the PPS method does not work. Furthermore, mobile communities (troupes) change size and composition throughout the year and may get smaller in crisis conditions as troupes disperse in search of scant grazing resources. Any official information is likely, therefore, to be out-of-date at the time of the survey.
Camels at a water point in Kidal.
Trials of the use of CSAS and 'snowball' sampling in retrospective mortality surveys are currently underway. The CSAS method has been adapted for nutrition surveys in pastoralist areas, for example in Mali in the same location as this field test6,7. Although addressing some of the problems of PPS methods, the main problem with CSAS is that it does not help the survey teams locate the mobile population and hence a lot of time is spent locating troupes.
Development of the new method
The new method was developed through a peer review process and designed to meet the criteria listed in Box 1. The peer review group were experts in the field of pastoralist society and economy. Existing methods, advice from peer reviewers, nutrition assessment guidelines and the experience of the consultants were used to design the method. Computer-based simulation was used to test the design statistically. The method was then field tested in Kidal, Mali with ACH.
Box 1: What are the requirements for a new method?
The follow requirements were considered essential to define a new method:
- It should be a general method that can be adapted to the field situation using information gathered locally.
- The sampling method should not require population data at the start and should not require knowledge of the location of populations ahead of the survey.
- The method should be straightforward and efficient to conduct in the field.
- It should be representative of the whole population - even remote communities.
- An unbiased estimator should be used.
- The precision should be predictable across different sample sizes and be similar to that obtained by conventional cluster sampled surveys in sedentary rural populations.
- The case definition used should be appropriate for pastoralists and a good predictor of nutritionassociated mortality.
Description of new method (Pastoralist Survey Method - PSM)
An appropriate case-definition for malnutrition in pastoralists was developed based on MUAC for several reasons. Pastoralists tend to have different body shapes; compared to sedentary agrarian populations their sitting height to standing height ratio is small. The body shape of pastoralists tends to produce inflated estimates of the prevalence of wasting using a WH based casedefinition8. MUAC gives more consistent estimates of malnutrition across different body shapes9 and is a better predictor of mortality than WH10, which is only weakly associated with nutrient reserves, nutrient requirements, and mortality. Compared to WH, MUAC measurement is quicker and more straightforward and requires no heavy field equipment to carry.
The sampling design makes use of local knowledge to construct the sampling frame using qualitative methods. This information is collected primarily from central locations in order to:
- Define the organising factors (factors that define the way the population is socially and/or geographically organized) and the troupe (group of pastoralists who move through the territory (survey area) together.
- Produce a detailed organisational tree (organogram) of the troupes for each organising factor. This organogram forms the basis of the sampling frame (Figure 1).
- Estimate the average size of the troupe, i.e. the number of eligible children in the troupes.
The statistical design of the method was tested by computer-based simulation using simulated populations of pastoralists with data provided by peer reviewers. The simulation exercise produced nomograms that can be used to estimate the required sample size using the estimated size of troupe, precision required and estimated prevalence of malnutrition (see Figure 2).
The method uses an exhaustive sample of children from each PSU (troupe). This way there is no need to 'spin a bottle' to select households to be sampled; all eligible children from selected PSUs (troupes) are included in the sample. The use of MUAC facilitates the collection of an exhaustive sample. In the unlikely event that the troupes are too large to include all eligible children in the sample, a second level of sampling may be introduced. This eventuality has not so far occurred and hence a method has not been developed for it, although techniques such as map-segment-sample that have been used to improve the accuracy of cluster sampled survey and in community level assessments of trachoma prevalence are likely to prove suitable.
Computer-based simulations found the method, when used with a bootstrap estimator, to be accurate (i.e. unbiased) and capable of returning estimates with useful precision (i.e. comparable to conventional designs) with readily achievable sample sizes.
Particulars of the survey design
The quantitative survey is designed using information gathered in the foregoing steps (i.e. the organogram and the average troupe size).
One or more troupes are combined (for the purposes of sampling) to form a Primary Sampling Unit (PSU). This is done if the average troupe size is less than about 10 children. The number of PSUs to be sampled is then read from the nomogram and a systematic sample drawn from the sampling frame (constructed as described above using the organogram). Only when the PSUs (troupes) to be sampled have been chosen will it be necessary to locate them in the field. Qualitative methods are used to find the identified troupes. Information will come from the initial qualitative methods and by using observation and information from key informants as you get closer to the selected PSUs (troupes).
The nutrition assessment is carried out by measuring/assessing all eligible children in the sampled troupes. Data may be collated at the level of troupe or PSU and computerised using specially designed software. Test versions of the software are available from: http://www.brixtonhealth.com/psm.html
Data may be collected using simple tally sheets and the data-entry and data-management costs are considerably lower than required for a cluster-sampled design using WH.
Measuring MUAC in Kidal
How these design features meet the requirements
The PSM is a general method that is adapted to the field situation when the organising factor and troupes can be defined. It is adaptable to different sized populations, sub-groupings and locations. This is important as these may change seasonally or in response to shocks such as drought. The method was shown to be straightforward and efficient to conduct in the field (see below).
The results should be representative of the whole population, even remote communities, if the qualitative assessments are carried out thoroughly and no sections of society are excluded. The PSM does not require population data at the start and does not require knowledge of the location of populations ahead of the survey. Only once troupes have been identified for sampling is it necessary to find them. At this stage you should already have a good idea of their location.
The sampling method has been tested statistically using computer-based simulation and has also been field tested, as reported below. Computer-based simulations showed that the estimator was unbiased and precision predictable across a range of prevalences and sample sizes. The method may also be used to measure many other parameters, such as prevalence of diarrhoea and vaccine coverage.
Use-study in Mali
The survey took place in Kidal commune of Kidal district, eastern Mali in collaboration with ACH. The population is approximately 30,000 of whom about 6,000 reside in Kidal town at any one time. The area of the commune is approximately 100,000 square kilometres. The population is largely nomadic pastoralist. The survey was conducted during the cooler dry season between January-March 2008. This was not a food insecure season or a season with high risk of malnutrition.
A previous nutrition survey carried out in Kidal by ACF11 in December 2006 used the CSAS sampling method. This survey reported a prevalence of acute malnutrition of 5.0% (95% CI 3.8-6.2) of children 6-59 months based on the WH z-score, with 1.9% in the rural areas and 6.8% in the town. The rate of severe acute malnutrition was 0.3% (95% CI 0.1-0.7). Using MUAC, 0.8% were moderately wasted (defined as MUAC < 120 mm & > 110 mm) and no children were severely wasted (defined as MUAC < 110 mm)12. No children in the rural areas were wasted.
Application of the design to Kidal, Mali context
Information was gathered and assembled with the help of key informants.
The organising factor was identified as 'water point' because most troupes were located close to water points13. The water points were divided into small medium and large categories by key informants (with respect to the number of troupes using the water point at that time).
Troupe was identified as 'campement'; a group of pastoralists moving together. The average (mean) number of eligible children in each troupe was estimated to be seven, using information from key informants, local reports and observation. For the purpose of sampling, two neighbouring troupes were combined to form one PSU. The combined troupes would therefore include approximately 15 children. This was done to make the field work easier - it is more straightforward to locate two neighbouring troupes than two independently sampled troupes. The approach of combining troupes is valid statistically as long as the number of sampled PSUs is not less than about 25.
An organogram was constructed with the relative size of each water point with respect to the number of troupes that were using the water point at the time of the survey. The organogram was used to produce a sampling frame of troupe-pairs by water point and a systematic sample was taken to achieve the target sample size (Figure 3).
The target sample size was read off the appropriate nomogram assuming 10% prevalence, PSU size of 15 and a required precision of 3% (Figure 2). This gives the target sample size of 42 troupe-pairs. A margin of error should be added to this of 10- 15% to allow for overestimation of troupe size. In the Kidal survey our target sample size was only 40 troupes (this was lower than recommended due to an error in reading the nomogram and not adding the 10-15% for margin of error).
On each day of the field assessment, information was collected from a group of pastoralists at each water point. A simple map was drawn of all the troupes using the water point and their locations. The target number of troupe-pairs was selected randomly using this map to locate them (Figure 4). Each selected troupe was visited and all eligible children measured. MUAC case definition was used along with oedema assessment (Global Acute Malnutrition MUAC ? 125 mm or oedema and Severe Acute Malnutrition MUAC <110 mm or oedema). This made the measurements on children very rapid. The amount of time spent at each troupe was not greatly increased by the number of eligible children within each troupe. Most time was spent in travelling.
Data were computerised using an excel spreadsheet and analysed using purposewritten software (Figure 5).
No children were found with bipedal pitting oedema. The prevalence of global acute malnutrition by MUAC was 1.64% (95% CI 0.21-4.26%) and severe acute malnutrition 0.43% (95% CI 0.01-1.12%). All cases of malnutrition were recorded in children height 65-85cm (a proxy for age less than 24 months).
Validation of the method
Qualitative data were validated by comparing information collected through qualitative methods at the central location (Kidal town), from informants at the water points and observations from the field. This resulted in the following observations:
- Troupes were found within daily reach of water points and in locations predicted by key informants (within approximately 25 km of water points).
- Locations corresponded to the 'zones of occupation' available through local food security maps (Figure 6).
- Relative size of the water points (in terms of the number of troupes using each) corresponded well (Cohen's kappa test showed 'moderate inter-observer agree ment'). For this test we compared information given by key informants in town and key informants at the water point.
- The mean number of eligible children found in each troupe was similar to the number predicted by key informants (7 predicted, 5.4 observed)
Quantitative data were validated by comparing the precision of the estimates obtained in the field-trial to precision predicted by computer-based simulation. The prevalence of wasting was low at the time of the survey, so to test the method, we varied casedefining MUAC cut-points to produce a series of testable prevalences. This resulted in the following observations:
- Precision of the estimate of wasting from the field test was very similar to the precision predicted by computer-based simulations: 3.7% (actual) vs. 3.4% (predicted) using MUAC cut-point of 135mm and testable prevalence of 9.9% wasting.
- Using different cut-points of MUAC to simulate different prevalences of wasting, the predicted and observed precision were strongly related over the entire range of prevalences between 5% and 50% (Figure 7).
|Figure 3: Example of the sampling frame used for Pastoralist Survey Method in Kidal, Mali|
|Water Point||Relative Size||Assigned PSUs?||Cumulative PSUs||PSUs chosen?||No of PSUs|
|Ebelel||Large||25||45||24, 33, 42||3|
|Igouzar||Large||25||110||87, 96, 10||3|
|Djounhan||Large||25||150||132, 141, 150||3|
|Takalot||Large||25||205||186, 195, 204||3|
|Tassik||Large||25||255||231, 240, 249||3|
|Agharous Alkit||Large||25||300||276, 285, 294||3|
|Tahadjante||Large||25||325||303, 312, 321||3|
|Amassine||Large||25||360||339, 348, 357||3|
|Intibzaz||Large||25||385||366, 375, 384||3|
? Small water points were assigned 5 PSUs, medium water points were assigned 10 PSUs, large water points were assigned 25 PSUs based on information provided by key informants.
? A sampling interval of 9 was used with an initial starting value taken at random from the range 1 to 9.
In this example the 'organising factor' was water-point and the 'troupe' was campement. 27 water points were identified and 385 campement-pairs (PSUs) assigned to the water points. Water points described as 'small' were assigned 5 pairs, 'medium' water points 10 and 'large' water points 25. These assignments were decided upon after qualitative data collection and were found to be reasonably accurate during the field trial.
The target sample size was 40 PSUs, hence the sampling interval 385/40, rounded down to 9. Random number chosen was 6. The systematic sampling method increased the sample to 43, and all 43 PSUs were included in the survey. The sample size of 40 was actually chosen in error. It should have been closer to 48 (see figure 2).
Discussion and Conclusions
The method is practical for use in pastoral populations; it is valid and is simple to apply. The quantitative data collection and data entry should present no difficulties for staff that are already familiar with cluster surveys. Many staff will be unfamiliar with collecting and analysing qualitative data and will require some training in the methods. Experienced staff may be needed during the early stages to advise on the choice of 'organising factor', for example. The delay between qualitative data and quantitative data collection needs to be as short as possible because seasonal changes and other movements of people affect the validity of the qualitative data.
The method could be used for any zone where population data are not well known in advance/ a population is moving, e.g. to escape conflict (although it may prove difficult to identify useful key informants to describe and locate 'communities' in such settings).
Pastoralist key informant at water point in Kidal.
The method of sampling can be used for collecting information on many different variables in addition to malnutrition.
Good relations, both within the population and between the population and the agency carrying out the survey, are required because reliable information must be gathered from the people in the area to be surveyed. In situations of conflict or other social divisions, it may be difficult or impossible to get unbiased information. As a general principle, there should be no survey without the provision of services to the community.
The urban and rural areas both need to be included to allow for migrations between the two zones. Each requires different methods of sampling. During the harsh season, the children often move to urban areas where prevalence of malnutrition may be much higher than in the rural areas. This was shown in the previous survey in Kidal14.
Further testing is necessary, especially in different pastoralist settings, and readers are invited to communicate with ACF about further trials. A field manual is being developed to assist the future development of the method.
1Myatt, M, Duffield, A, Seal, A, and Pasteur, F. (2009). The effect of body shape on weight-for-height and mid upper arm circumference based case-definitions of acute malnutrition in Ethiopian children. Annals of Human Biology. Also see: (2008) Effect of body shape on weight-for-height and MUAC in Ethiopia. Emergency Nutrition Network. Field Exchange 34.
2(2005). Measuring Mortality, Nutritional Status and Food Security in Crisis Situations: SMART METHODOLOGY. Version 1 UNICEF and USAID
3A spatial sampling method that uses a systematic spatial sample from a defined geographic area.
4Myatt, M, Feleke, T, Sadler, K., and Collins, S (2005). A field trial of a survey method for estimating the coverage of selective feeding programs. Bulletin of the World Health Organization 83, 20-26
5Myatt, M, Mai, NP, Quynh, NQ, Nga, NH, Tai, HT, Long, NH, Minh, TH, and Limburg, H (2005). Using lot quality assurance sampling (LQAS) and area sampling to identify priority intervention areas for trachoma control activities-Experiences from Vietnam. Bulletin of the World Health Organisation 83, 756-763
6ACF (2006). Enquete nutritionelle et de mortalite. Commune de Kidal, Mali 16-27, Dec 2006
7Vincent, E, and Salse, N (June 2008). Methodology for a nutritional survey among the nomadic population of northern Mali. Emergency Nutrition Network. Field Exchange 33, 14
8See footnote 1.
9See footnote 1.
10Myatt, M, Khara, T, and Collins, S. (2006). A review of methods to detect cases of severely malnourished children in the community of their admission into community-based therapeutic care programmes. Food and Nutrition Bulletin: SCN Nutrition Policy Paper 21, S7
More like this
By Ernest Guevarra, Saul Guerrero, and Mark Myatt Ernest Guevarra leads Valid International's coverage assessment team. He has formal training as a physician and a public...
FEX: RAM-OP: A rapid assessment method for assessing the nutritional status, vulnerabilities, and needs of older people in emergency and development settings
By Pascale Fritsch (HelpAge International), Katja Siling (VALID International), and Mark Myatt (Brixton Health) Dr Pascale Fritsch is an experienced public health specialist....
I would like to know the most appropriate and cost effective sampling system which is relevent to apply the pastoral communities like Somalia. The article describing the Mali...
In looking at materials on CSAS, there are several references to S3M but I can't find anything that actually explains it. Does anyone have any resources for that? Hi Mark, I...
I am nutrition student looking to do a study on CF practices in an arid area (predominant pastoralist). Cluster sampling seems to be the main method in use. However, the...
By Eva Vicent and Núria Salse Eva Vicent has a background in nursing studies in the University of Valencia and is currently working for Action Against Hunger as Nutrition...
FEX: Assessment of the PROBIT approach for estimating the prevalence of acute malnutrition from population surveys
Summary of research1 Location: N/A What we know already:Prevalence of GAM is normally estimated using two stage cluster sampled surveys using the SMART method. The PROBIT...
I have got problem in reconciling between manual calculation and the ENA for SMART 2011 (version November 16th 2013). Here are the assumptions; GAM 17.5%, precision 4, DEFF...
en-net: SQUEAC and Nomads
Dear all, We have recently completed a second SQUEAC survey in a region in Chad. We found a coverage 5 points lower than the previous coverage. We hypothesized that one of...
en-net: Quick and (not) dirty
What would be the most adequate and practical rapid nutrition assessment tool to be carried out in a huge emergency e.g. Pakistan floods if circumstances and time do not allow...
Now days, the term "Small scale SMART survey, SSSS" is becoming common. In principle, is it different from that of normal SMART survey procedure? To make it clear, does SSSS...
By Johannah Wegerdt, Monica Zanchettin and Mark Myatt Marking the midpoint during MUAC measurement Monica Zanchettin is a coverage surveyor for Valid International. She has...
en-net: LQAS types
I have a chance to read number of assessment reports using LQAS but they used different terminologies/methods regarding LQAS. these are the conventional LQAS, cluster LQAS and...
By Mark Myatt Mark Myatt is a consultant epidemiologist and senior research fellow at the Division of Ophthalmology, Institute of Ophthalmology, University College London....
Summary of Field Exchange 49 article by Sophie Woodhead, Jose Luis Alvarez Moran, Anne Leavens, Modibo Traore, Anna Horner, Saul Guerrero. In 2014, the Mali Ministry of...
FEX: Stunting and wasting in children under two years old in a semi-nomadic pastoralist population in Kenya
By Amelia Reese-Masterson, Masumi Maehara and Mark Murage Gathii Amelia Reese-Masterson is Research Advisor in the Nutrition, Food Security and Livelihoods Unit at...
By Sophie Woodhead, Jose Luis Alvarez Moran, Anne Leavens, Modibo Traore, Anna Horner, Saul Guerrero Based within the ACF-UK Operations Team, Sophie Woodhead coordinates all...
en-net: SMART HHs Sampling
Dear Colleagues, We are planning to conduct a SMART survey in one area in Sudan. This area is a mixture of IDPs camps and host communities with no specific arrangement of HHs...
By David Rizzi David Rizzi graduated in Pharmacy and later took an MSc in Development at the Rome University La Sapienza, Italy. He holds a second MSc in Public Health...
Hello, Id like to know what is the best recommended methodology for SMART Survey in Urban Setting with large population (total population: 500,000p, pop/cluster >2000...
Reference this page
Anne-Marie Mayer, Mark Myatt, Myriam Ait Aissa and Nuria Salse (2009). New method for assessing acute malnutrition in nomadic pastoralist populations. Field Exchange 35, March 2009. p30. www.ennonline.net/fex/35/newmethod