guidelines for nutrition surveys - bangladesh

97
1 Guidelines for Nutrition Surveys - Bangladesh June 2015 Anthropometry, Mortality, IYCF, Food Security and WASH.

Upload: others

Post on 02-Apr-2022

7 views

Category:

Documents


0 download

TRANSCRIPT

WASH.
2
3
ACKNOWLEDGEMENTS
This National Guidelines for Nutrition Surveys in Bangladesh are designed to provide
clear instructions and guidance on area based nutrition surveys. I wish to convey my
gratitude to the Director DGHS and the director IPHN/NNS for the strong leadership in
preparing this guidelines. I do also acknowledge the contributions of the Program
Manager and deputy program manager - Nutrition In Emergencies – under IPHN/NNS
for their valuable administrative and technical support in the preparation of the
guidelines.
I gratefully acknowledge the contribution of the members of the technical committee
formed under IPHN to review the draft guidelines. Contributions by staff from the
following organisations is who were members of the technical committee is highly
appreciated.
- MOHFW
- NNS/IPHN
- UNICEF
- ICDDR,B
- Institute of Nutrition and Food Science (INFS) of Dhaka University.
I would also like to appreciate the time, review and useful recommendations provided by
the curriculum committee towards finalisation of the guidelines. Thanks too to all the
participants from government and development partners who attended and contributed at
consultative workshops at national level.
Regards.
Name
4
ACRONYMS
FCS Food Consumption Score
FGD Focus group discussion
GAM Global acute malnutrition
MAM Moderate acute malnutrition
NCHS National centre for health statistics
PPS Probability proportional to population size
RNA Rapid nutrition assessment
SAM Severe acute malnutrition
WASH Water Sanitation and Hygiene
WAZ Weight-for-age z-score
5
3.1 Decide whether to conduct a survey ................................................................................................ 12
3.2 Define survey objectives ................................................................................................................. 12
3.3 Define geographic area and population groups ............................................................................... 13
3.4 Meet local leadership and authorities .............................................................................................. 13
3.5 Determine the timing of the survey ................................................................................................. 14
3.6 Gather population data and other data ............................................................................................. 14
3.7 Select the sampling method, determine sample size ....................................................................... 14
3.8 Decide which data to collect ............................................................................................................ 14
3.9 Prepare supplies and equipment ...................................................................................................... 15
3.10 Select and train survey teams ....................................................................................................... 15
3.11 Collect data and manage survey teams .......................................................................................... 16
3.12 Enter and clean data....................................................................................................................... 16
4. SAMPLING ....................................................................................................................................... 18
4.3 Setting parameters for sample size calculation................................................................................ 20
4.5 Cluster sampling .............................................................................................................................. 26
5. SURVEY FIELD PROCEDURES ............................................................................................................. 30
5.1 Bias .................................................................................................................................................. 30
6
6.1 Selection of survey teams ................................................................................................................ 33
6.2 Training of survey teams ................................................................................................................. 33
6.3 Roles and responsibilities of team members .................................................................................. 34
6.4 Standardisation test .......................................................................................................................... 36
7. DATA COLLECTION .................................................................................................................... 39
8.1 Data entry ........................................................................................................................................ 54
8.2 Plausibility check ............................................................................................................................. 54
9.1 Data analysis .................................................................................................................................... 57
10. REPORT WRITING AND DISSEMINATION .............................................................................. 64
10.1 Preliminary Report ........................................................................................................................ 64
10.2 Final report .................................................................................................................................... 65
11.1 Decide whether to conduct a RNA ................................................................................................ 67
11.2 Define RNA objectives.................................................................................................................. 68
11.4 Meet local leadership and authorities ............................................................................................ 68
11.5 Determine the timing of the RNA ................................................................................................. 68
11.6 Gather population data and other data ........................................................................................... 69
11.7 Select sampling method, sample size ............................................................................................ 69
11.8 Decide which data to collect .......................................................................................................... 71
11.9 Prepare supplies and equipment .................................................................................................... 71
11.10 Select and train assessment teams ............................................................................................... 71
7
11.12 Enter and clean data..................................................................................................................... 73
11.14 Write and disseminate the RNA report........................................................................................ 75
Annex 2 Local calendar of events ......................................................................................................... 77
Annex 3 Referral form........................................................................................................................... 78
Annex 5 Mortality module .................................................................................................................... 81
Annex 6 IYCF module .......................................................................................................................... 82
Annex 7 Food security module.............................................................................................................. 84
Annex 8 Food security module.............................................................................................................. 86
Annex 10 Cluster control form .............................................................................................................. 93
Annex 11 MUAC screening form ......................................................................................................... 94
Annex 12 Key informant interview guidance sheet .............................................................................. 95
Annex 13 Focus group discussion guidance sheet ................................................................................ 96
Annex 14 Glossary of terms .................................................................................................................. 97
References ............................................................................................................................................. 99
Figure 2. Cluster sampling ..................................................................................................................... 19
Figure 3. Effect of changing estimated prevalence on sample size ...................................................... 20
Figure 4. Sample size calculation (SRS).................................................................................................. 24
Figure 6. Sample size calculation (cluster sampling) ............................................................................. 27
Figure 7. Modified EPI method ............................................................................................................. 28
Figure 8. Adjustment for small sample size .......................................................................................... 29
Figure 9. Standardization test ................................................................................................................ 37
Figure 10. Indices of nutritional status .................................................................................................. 39
8
Figure 12. Salter hanging scale ............................................................................................................. 42
Figure 13. Mother-to-child scale .......................................................................................................... 42
Figure 14. Height measurement ............................................................................................................ 43
Figure 15. MUAC measurement ........................................................................................................... 44
Figure 16a. Identification of nutritional oedema .................................................................................. 44
Figure 16b. Nutritional oedema grade 2 ............................................................................................... 44
Figure 16c. Nutritional oedema grade 3................................................................................................ 44
Figure 19. Conceptual framework of malnutrition................................................................................ 48
Figure 22. Flagged values on data entry screen .................................................................................... 52
Figure 23. SMART and WHO flags ..................................................................................................... 53
Figure 24. Check for double entry ........................................................................................................ 54
Figure 25. Plausibility report ................................................................................................................. 54
Figure 26. Results anthropometry ......................................................................................................... 57
Figure 27. Results mortality .................................................................................................................. 58
Figure 28. Food Consumption Score weights ....................................................................................... 61
Figure 29. Food Thresholds for FCS ..................................................................................................... 62
Figure 30. Timing of rapid assessments ................................................................................................ 69
Figure 31. MUAC data analysis ............................................................................................................ 74
LIST OF TABLES Table 1. Precision: anthropometry ........................................................................................................ 21
Table 2. Precision: mortality .................................................................................................................. 21
Table 3. Cluster sampling example ...................................................................................................... 26
Table 4. Standardization test ................................................................................................................. 36
9
Table 7. Mortality rate calculation ........................................................................................................ 47
Table 8. Plausibility report criteria ........................................................................................................ 55
Table 9. Vitamin A supplementation and measles vaccination............................................................. 59
Table 10. Morbidity results ................................................................................................................... 59
Table 11. Antenatal care and iron folate ............................................................................................... 59
Table 12. IYCF analysis ........................................................................................................................ 59
Table 13. Household analysis ................................................................................................................ 60
Table 14. Classification of severity of malnutrition .............................................................................. 62
10
1. INTRODUCTION.
According to the World Risk Report 2012, of 173 countries, Bangladesh is the 5th most disaster
prone country in the world, particularly susceptible to devastating tropical cyclones, storm surges
and floods. There are 27 million people across the 12 districts vulnerable to tropical cyclones and
storm surges. Forty percent of these 12 districts, covering around 10 million people, are considered
high risk areas and with an average poverty rate of 40%, people are particularly vulnerable to
natural hazards. Malnutrition has remained high in Bangladesh, although a decline has been noted
in stunting and underweight between 2004 and 2011 (Bangladesh 2011 DHS). Negative
consequences of a cyclone on other sectors may aggravate the already poor nutritional status of the
population. This in the background that under nutrition prevalence is chronically high in the
coastal areas as is in the entire country. The nutritional status of especially under five children as
well as pregnant and lactating women can deteriorate in quickly in the event of a disaster. It is
therefore important to regularly assess the nutritional status in Bangladesh, particularly in disaster-
prone areas. Currently, the Food Security Nutrition Surveillance Project (FSNSP), implemented by
BRAC University in collaboration with Hellen Keller International, is the only source of seasonal,
nationally representative estimates of malnutrition in Bangladesh. There is, however, scope to
increase the coverage of nutrition surveys, and the development of these guidelines is one
important step towards achieving the same.
These guidelines are designed to provide clear instructions and guidance to survey managers on
nutrition surveys and rapid assessments by outlining steps and procedures to be followed in
planning, implementation and evaluation of nutrition surveys and rapid assessments. The
guidelines are divided into: a. Comprehensive nutrition survey guidelines, and b. Rapid nutrition
assessment guidelines. The development of these guidelines has been triggered by the need for
standardizing the approach and methodology used to conduct nutrition surveys and rapid
assessments for the purposes of comparability of results and compliance to international
standards. The methods and instructions are based on the SMART methodology
(www.smartmethodology.org), which is based on the assessment of malnutrition and mortality
to establish the magnitude of a crisis. The SMART Methodology draws from core elements of
several existing methods and current best practices. Recommendations are based on varying
degrees of evidence including methods for which there is clear scientific evidence to support its
recommendation. A practical consideration that initiated the development of the SMART
method, and guided the decision process in its development, is that partners should be able to
collect data in nutrition surveys with a minimum of added burden to their programs. In addition,
the method’s level of difficulty is a conscientious balance between technical soundness and
simplicity for rapid assessment of acute emergencies to obtain early, accurate, quantitative
profiles of a population’s nutritional status and mortality rate. For these reasons, the SMART
method is iterative, with continuous upgrading and building on this basic version, informed by
research, experience, and current best practices.
The current guidelines are to be used to undertake surveys in both emergency and none
emergency contexts. The timing of nutrition surveys during an emergency should be as per the
Joint Needs Assessment (JNA) Planning that propose detailed sectorial surveys at the 6-8 week
after a disaster. However, rapid nutrition assessments could be done immediately - 1st week -
after a disaster either as a part of the JNA or as stand alone. (Please see page 69 for the
diagram). In none emergency contexts, the guidelines should be used to undertake nutrition
surveys to provide nutrition status information of the population for specific programmatic
reasons.
2. BACKGROUND
In order to gain an understanding of the extent to which an emergency is impacting nutrition it is
important to analyse data on the affected population and area. Data relating to nutrition can be
collected, and existing evidence should be reviewed. Nutrition assessments are essential to guide
response during an emergency. There are three main methods used to assess the nutrition of
populations: rapid nutrition assessments, nutrition surveys and nutrition surveillance. In a chronic
or complex emergency the situation is ongoing and nutrition surveillance is carried out.
Anthropometric surveys are included as part of this; their purpose being to collect, analyse,
interpret and report on information about the nutritional status of populations over time and to
inform appropriate response strategies. However, in a rapid-onset emergency the priority is to
obtain a snapshot of the nutrition situation as quickly as possible and therefore rapid nutrition
assessments are carried out. The information may not always be representative and thus not
statistically valid, but the results from a rapid assessment can verify the existence or threat of a
nutrition emergency, provide an estimate of the numbers affected and establish immediate needs.
Rapid assessments are also done where there is poor security and very limited access. Data in
rapid assessments is collected directly from the field and is usually qualitative.
Acute malnutrition in children 6-59 months is closely linked with risk of death and is used to
draw conclusions about the situation of the health status of the whole population, not just young
children. Children aged 6-59 months are more vulnerable than other age groups to external
factors (such as food shortage or illness) and their nutrition status is more sensitive to change than
that of adults in many (although not all) populations. Mortality is the most critical indicator of a
population’s improving or deteriorating health status and is the type of information to which
donors and relief agencies most readily respond.
Nutrition surveys using a statistically representative sample of children remain the best method to
determine the magnitude of malnutrition in a population. However, there are certain limitations to
the use and interpretation of nutrition survey findings. Accurate population data is needed to list
the population in villages or population units. This may not be available in an emergency.
Additionally, the data cannot be disaggregated to produce statistically reliable results for
geographical sub-samples when cluster sampling is used. Surveys are also time and resource
consuming, but are often necessary to assess the anthropometric situation with accuracy.
Interpreting results of anthropometric nutrition surveys in relation to contextual factors and
interventions is also not straightforward and requires a wealth of information including food
security and public health.
3. STEPS IN CONDUCTING NUTRITION SURVEYS.
This section describes the steps to be followed when conducting a nutrition and/or mortality
survey.
12
2. Define survey objectives
4. Meet local leadership and authorities
5. Determine the timing of the survey
6. Gather population data and other data
7. Select sampling method, determine sample size
8. Decide which data to collect
9. Prepare supplies and equipment
10. Select and train survey teams
11. Collect data and manage the survey team
12. Enter and clean data
13. Analyse data and interpret results
14. Write and disseminate the survey report
3.1 Decide whether to conduct a survey It is always important to make sure that the decision on whether or not to conducts a nutrition
survey is made with consideration to the following points:
Avoiding overlap: the decision to undertake an assessment is usually made in
conjunction with the government, the nutrition cluster and other agencies so as to
prevent overlap by different agencies.
Are the results crucial for decision making? If a population’s needs are obvious,
immediate program implementation takes priority over doing a survey, and the survey
should be deferred. For example, after a natural disaster, such as a flood, where it is clear
that the population’s food stocks have been destroyed, the current nutritional status may
reflect the pre-disaster state.
It should be anticipated that the results will lead to action: there is little point of doing a
survey if you know a response will not be possible. If the agency cannot itself implement
a program where needed, the results must be useful in advocating for a response.
Is the affected population accessible? Insecurity or geographical constraints may result in
limited access to the population of interest. If this is extreme, a survey cannot be
conducted. 3.2 Define survey objectives It is important to be clear about what the survey seeks to achieve. In most cases, nutrition surveys
seek to quantify the level of malnutrition (and/or mortality) in a given population at a defined
point in time. Nutrition surveys also provide a baseline from which future trends can be
monitored. Nutrition and mortality surveys also provide opportunity for collecting additional data
on relevant interventions and nutrition-related variables such as food security. These include
immunization and nutrition program coverage, vitamin A, anaemia, or other micronutrient
deficiency and morbidity. However, caution must be exercised given that a survey presents a
greater likelihood of inaccuracy as more data is included.
The broad objective (aim) of a nutrition survey is to assess the current nutrition and health
status of a specific population. (The population may be the district, village, camp or urban
settlement, or even the region or country).
13
The specific objectives may vary depending on the interventions, situations
or circumstances in place or intended, and may include the following:
To determine the prevalence of malnutrition (wasting, stunting and
Under weight) among children aged 659 months.
To determine the nutritional status of a specific subgroup (e.g. women of
reproductive age, adolescents or the elderly).
To determine the coverage of health interventions (e.g. measles vaccinations,
Vitamin A supplementation and oral polio vaccine) among children aged 659
months.
To determine the levels of retrospective crude mortality rates and age-
specific mortality rates for under5s in a specific time period.
To determine the incidence of common diseases (diarrhoea, measles and ARI)
among the target population, two weeks prior to the assessment.
To identify possible interventions that addresses the causal factors of
malnutrition.
Note that, in defining specific objectives, the target group must be specified where applicable.
The objectives must be measurable, and should be feasible within the context of a nutrition
survey, bearing mind the limitations of nutrition surveys due to their cross-sectional nature,
which does not allow for determining the cause-and-effect.
3.3 Define geographic area and population groups
In planning a nutrition survey, a decision must be made as to the area and population groups that
will be covered, with clear justification. It is useful to have a map of the selected area for
reference, and for inclusion in the final survey report. A survey should be conducted in an area where the population is expected to have a similar nutritional and mortality situation. If an area is assessed that has two or more very different agro- ecological zones, the results will be an average of the two zones and not give an appropriate perspective of either zone. Such heterogeneity can be resolved by doing separate assessments, although this usually increases the cost. In general, urban and rural areas, refugee/IDP, and resident populations should be assessed separately. If there are areas which are unreachable due to insecurity, these must be defined before the survey and must be reported as having been excluded from the survey. Anthropometric measurements and oedema assessments for children ages 6 to 59 months, and crude death rate (CDR) for the entire population (all deaths within a defined period of time) are the priority for nutrition and mortality surveys. The 6 to 59 month-old children are considered the most sensitive to acute nutritional stress and thus a proxy of the severity in the whole population. Globally, there is also more experience in collecting data from this age group.
3.4 Meet local leadership and authorities Meeting local leadership and authorities before a survey is important for the following
reasons:
• To obtain letters of permission from the local authorities addressed to the district or
village leaders, stating that you will be conducting the survey, stating the reasons for
the survey.
• It is necessary to agree with the community about the objectives of the survey. If the
population does not understand why you are doing an assessment they may not
14
cooperate during the survey.
• To obtain a map of the area to plan the survey.
• To obtain detailed information on population figures.
• To obtain information on security and access to the prospective survey area.
• To agree on the dates of the survey with the community and local authorities.
• To agree on how the results will be disseminated and used and, in particular, to discuss the
likely intervention if the situation is found to be as poor as expected.
3.5 Determine the timing of the survey The exact dates of the assessment should be chosen with the help of community leaders and
local authorities to avoid market days, local celebrations, food distribution days, vaccination
campaigns, or other times when people are likely to be away from home. Roads may be
impassable during the rainy season. In agricultural areas, women may be in the fields for most of
the day during ground preparation, planting, or harvesting. Wherever possible, community
leaders should inform the villages chosen to be surveyed in advance. In determining the timing
of a survey, it may be decided that a survey be conducted at the start of an intervention, and then
at the end, so as to determine the impact of an intervention. This may be challenging, as there
may be several factors contributing to the impact of an intervention and it may be difficult to
establish attribution. Nutrition surveys may also monitor trends of malnutrition and identify
possibly impact of a crisis, which is generally a more relevant rationale.
3.6 Gather population data and other data Before starting the survey, it is important to learn as much about the population as possible from
existing sources, including population characteristics and figures. Population figures are key for
sample size calculation. Data on previous surveys and assessments, health statistics, food
security information, situation reports (security and political situation), maps, and
anthropological, ethnic, and linguistic information is also important.
3.7 Select the sampling method, determining sample size
A decision must be made as to which sampling method to apply, based on knowledge about the
size of the population, the layout of households, and the presence of household lists. Simple
random sampling, systematic random sampling and cluster sampling are the three methods
recommended by SMART. With a small sample size, exhaustive sampling may be used. The
population data is then used to calculate the required sample size for the survey, which assists in
anticipating the expected duration and required personnel and equipment.
3.8 Decide which data to collect
The objectives of the survey will guide the decision on what data will be collected, and hence
what materials and equipment could be required.
Nutrition
For determination of the magnitude and related factors influencing malnutrition, data normally
collected includes:
15
1. age, in months (from a known date of birth or based on an estimate derived from a
calendar of local events)
6. mid-upper arm circumference (MUAC)
7. measles immunization
10. morbidity information
Mortality To estimate the mortality rate (and causes of death), the following information needs to be
collected:
1. total number (of all ages) currently in the household
2. number who were in the household at the start of the recall period
3. number of deaths
4. number of births
5. number who left the household during the recall period
6. number who joined the household during the recall period
7. age and sex of each household member
8. number of deaths of children below age 5
9. information about cause of death
Additional data to collect may include:
-Food security
3.9 Prepare supplies and equipment
It is essential to take measures to ensure that measuring material, including scales and height
boards, are procured in good time and are in good condition. During the survey, scales should
always be calibrated each day against a known weight. Faulty equipment should never be used
and there should always be spare equipment. Additional equipment and supplies include:
vehicles, fuel, paper and pens, questionnaires, and referral forms.
3.10 Select and training survey teams Team members do not have to be health professionals. In fact, anyone from the community can
be selected and trained. They need to be fit, as there is usually a lot of walking. They should
have a relatively high level of education, as they will need to read and write fluently, and count
accurately. Ideally, they will speak the local language. If not possible, there should be
interpreters as part of the survey teams. Women generally have much more experience dealing
with young children and should usually lead the interviewing of mothers/caretakers of children.
This is also important as some cultures do not allow women to be interviewed by men. The
gender composition of the team should conform to the local context.
The composition of survey teams depends on the data to be collected. Two people are required
16
for measurement of children (measurer and recorder) in addition to an interviewer. A team
leader is also required for quality control and leadership of the survey team. If there are
additional modules such as food security and water and sanitation, which are household
modules, an additional member may be required. Generally, four to six teams survey teams may be needed depending upon the number of
households to be visited, the time allocated to complete the survey, and the size and the
accessibility of the area covered. The number of teams should never be too many despite the fact
that the more the teams, the faster the data collection. The quality of the data deteriorates with too
many teams as it is much more difficult to train, supervise, provide transport and equipment, and
organize a large number of teams. Supervisors should be assigned to each team. If the teams are
to collect data in nearby areas, there may be a supervisor for two teams, but if they are far apart, a
supervisor may be required for each team. The supervisor must be experienced in undertaking
nutrition and mortality surveys, training team members, organizing logistics, and managing
people. Adequate training of the survey team members before the survey is crucial. All scheduled training
must be completed prior to data collection, and every team member should undergo exactly the
same training, whatever their former experience, to ensure standardization of methods. During the
survey the supervisor must continually reinforce good practice, identify and correct errors, and
prevent declining measurement standards.
3.11 Collect data and manage the survey team After having trained survey teams and assigning members to respective teams, the data
collection is ready to begin. Supervisors have the overall responsibility of management of survey
teams in the field. The supervisor must ensure that households are selected properly, ensuring
the equipment is checked and calibrated each morning during the survey and that measurements
are taken and recorded accurately. Unexpected problems nearly always arise during a survey,
and the supervisor is responsible for deciding how to overcome them. Each problem encountered
and decision made must be promptly recorded and included in the final report. The survey
supervisor is also responsible for overseeing data entry and for the analysis and report writing. The survey manager should organise a review session at the end of each day for a discussion on
the day’s progress and any possible challenges. Before leaving the field, each team leader should
review and sign all forms to ensure that no pieces of data have been left out. If there were people
absent from the house during the day, the team should return to the household at least once before
leaving the area.
It is also the duty of the supervisor to regularly supervise teams in the field. It is particularly
important to check cases of oedema, as there are often no cases seen during the training and some
team members may therefore be prone to mistaking a fat child for one with oedema (particularly
with younger children). The supervisor should note teams that report a lot of oedema, and visit
some of these children to verify their status.
The survey teams must be managed in such a way that they are not overworked, as this may
introduce bias due to short cuts and errors. This is achieved by the survey manager making a
realistic determination of the number of households which a team can realistically complete in
a day without fatigue. 3.12 Enter and clean data It is important to note that data cleaning in nutrition surveys begins from the moment data
17
collection begins, rather than at the end. By conducting data cleaning as data collection proceeds,
errors can be swiftly rectified to enhance the accuracy of data collected.
The process begins with the team leaders, who must check the questionnaires during the day for
errors, which may include omissions. Each evening, or during the next day while the teams are in
the field, the supervisor should arrange for data to be entered into the computer. Recording
errors, unlikely results, and other problems with the data may become clear at this stage. The
ENA for SMART software will automatically flag abnormal values as data are entered. Each
morning, before the teams set out for the day, there should be a short feedback session. If any
team is getting a large number of “flagged” results, the supervisor should accompany that team
the next day. If the results are very different from those obtained by the other teams, it may be
necessary to repeat the cluster from the day before.
3.13 Analyze data and interpret results ENA for SMART is recommended for analysis of anthropometry data, which may either be
entered directly, or copied from other software such as Microsoft Excel. Individual-level data on
additional indicators may be analyzed with other software such as EPI-INFO given the
limitations for ENA for SMART. Similarly, household-level data may neither be entered in nor
analyzed in ENA for SMART.
3.14 Write and disseminate the report The final part of a survey is report writing and dissemination. The results of the survey should
be presented in a standard format so that different surveys can be compared, and no important
information should be left out. After being introduced to the standard format, it is also becomes
much easier for readers to find particular pieces of information in the report. ENA for SMART
automatically generated a standard report format with standard headlines, which the survey
manager can build upon. This is very convenient for the survey manager.
Preliminary results from a nutrition survey should be available within approximately a week,
with the full report being available within a month, assuming that there are no unforeseen
problems. The survey report must clearly articulate the objectives, implementation steps and
findings of the survey in clear language. An important aspect of the report is recommendations
for possible intervention.
4. SAMPLING
This section will define and explain the different procedures and methods for sampling in
nutrition surveys, followed by a step-by-step description of procedures for sample size
calculation.
4.1 Defining sampling
If all the children aged 6-59 months from a given population were measured, we would get a
precise picture of the nutrition status of the population. This is called a census, or exhaustive
survey, and it is possible in a small population. However, an exhaustive survey is normally long,
costly and difficult to carry out in a large population. Instead of surveying all the children, we
normally survey only a sub-group of the population, called a sample, which “represents” the
whole population. Instead of interviewing all the households and measuring all the children, a
sample may be taken to represent the whole population. It is important that the sample be
chosen that indeed is representative of the whole population. This is done by choosing
18
households at random, whereby the selection of one household is independent of the selection of
another, so as to give each household and child in the population an exactly equal chance of
being selected into the sample.
4.2 Selecting the sampling method
Box 1 summarizes the decision-making process for selecting the suitable sampling method for a
nutrition survey.
No
Simple random sampling
When a complete and updated list of households is available and the population is relatively
small, it is recommended to use simple random sampling, whereby each household is randomly
selected using a random number generator.
Systematic random sampling
Where the population is relatively small and the numbers of households are known, households
may be arranged in a clear pattern as shown in Figure 1, with survey teams able to move
systematically from one household to another. In this case,
systematic random sampling is used. This method is a
variant of simple random sampling. In this method, a
sampling interval is determined by dividing the total
number of households by the required number of
households. A random number is then selected between 1
and the sampling interval to determine the starting point.
Cluster sampling
Systematic random
19
From the starting point, the sampling interval is applied continuously to select subsequent
households until the sample has been achieved. Simple or systematic random sampling is
normally useful in contexts such as small refugee camps and urban settlements.
Cluster sampling
In a relatively large population such as a district, cluster sampling is more preferable as it may
not be feasible for teams to travel long distances if households are to be randomly selected and
may be far apart. Another reason is that the likelihood of
having an updated list of all households in a large area is
very low. This method is implemented in two stages.
There are two stages, which are: 1. selection of clusters
and, 2. selection of households (Figure 2) Stage 1:The
whole population is divided, on paper, into smaller
discrete geographical areas, such as villages as in the
case of a district-level survey.
The population of each smaller area must be known or
page42 be estimated with reasonable accuracy. Clusters
are then randomly selected from these villages with the
chance of any village being selected being proportional to
the size of its population. This is called sampling with
“probability proportional to population size” (PPS).
Stage 2: Households are chosen at random from within
each cluster area or village.
4.3 Setting parameters for sample size calculation
In order to calculate the required sample size for a
nutrition survey, the following parameters are considered:
Estimated prevalence/death rate
Anthropometry
The estimated prevalence of acute malnutrition is estimated, and can be estimated from previous
survey data, or surveillance data. Emergency thresholds may also be used when previous data is
unavailable. It must be noted that, to determine the sample size it is recommended that the most
conservative value be selected, which is the prevalence as close to 50% as possible, given that the
required sample size increases as estimated prevalence increases up to 50% (Figure 3).
Figure 3 Effect of changing estimated prevalence on sample size
Figure 2 Cluster sampling
Mortality
The expected Crude Death Rate (CDR) in a mortality survey can also be estimated from previous
surveys or from discussion with key informants. It may also, in the absence of these sources, be
set as CDR of 2 deaths per 10,000 per day, which is the level that is often used to declare an
emergency (WHO, 1995).
Standard error, probability and sampling interval
Data gathered from a sample population only provides an estimate of the true population value.
Thus the true population value can only be calculated through exhaustive sampling (by measuring
every child in the population). Hence, whenever a sample is drawn, there is a risk that it will not
be truly representative and, therefore, that the results do not reflect the true situation. Inevitably,
if a second sample is drawn from the same population, slightly different results are likely to be
obtained. This risk is known as the standard error. In anthropometric surveys, the generally
accepted standard error is five per cent. That is to say that if a hundred sample surveys were
carried out on the same population, five would give results that were not representative of the
total population. When we undertake a survey, therefore, we calculate not only an estimate of the
rate of malnutrition but also the range of values within which the real rate of malnutrition in the
entire population almost certainly lies. This range is usually called the confidence interval (C.I).
In nutrition surveys we generally accept that a 95 per cent confidence interval is appropriate (5
per cent standard error). This means that we are 95 per cent certain that the true prevalence of
malnutrition lies in the range given. If a survey found the prevalence of global acute malnutrition
(GAM) to be 29.7% (23.8-36.4 95% CI), this would mean that we are 95% confident that the true
GAM prevalence lies between 23.8% and 36.4%.
The 95% is automatically calculated when using ENA for SMART software.
Precision
Precision measures the consistency of the results and is related to sampling error. The larger the
proportion of the target population that is measured, the lower this uncertainty becomes.
Therefore, the higher the sample size, the higher the precision. A larger sample size increases the
precision of the results.
prevalence Interval precision
40 30 – 50 10.0
There are recommended ranges for precision which are recommended by the SMART
methodology for different levels of prevalence of malnutrition and death rate (Table 1 and 2):
Table 2 Precision: mortality
Design effect
The design effect (DEFF) is a correction factor to account for the heterogeneity between clusters
with regard to the measured indicator. Therefore, it is only used to determine sample size in
cluster sampling. Generally, if there is no previous information about design effect, 1.5 can be
used as a default for GAM. DEFF depends on the prevalence and the size of the clusters. The
higher the expected prevalence, the higher would be DEFF. For example, if your expected
prevalence is around 10%, expected DEFF may be 1.5, whereas if expected prevalence is around
25-30% you would increase your expected DEFF to 1.7-1.8. The smaller the number of children
per cluster, the smaller the DEFF will be. For example, if you are measuring 15 children per
cluster, your DEFF may be 1.5, whereas if you plan to measure 25-30 children per cluster, you
would increase your expected DEFF to 1.7-1.8. If heterogeneity is expected to be high, the
maximum value used is 2. DEFF multiplies the sample size, meaning that a DEFF of 2 doubles
the required sample size. A higher DEFF than 2 would mean that more than one survey must be
conducted due to very high heterogeneity. In this case, stratified sampling may be considered.
Recall period
In mortality surveys, a recall period, in days, is applied, and is defined as the interval over which
deaths are counted. It is determined by looking at the period most relevant to the purposes of the
survey, the risk of mortality being measured, and the context of the study. To improve the
accuracy of mortality estimates in cross-sectional surveys, the beginning of the recall period
should be a memorable date known to everyone in the population. For example, the start of the
recall period may be a major holiday or festival (Christmas, beginning of Ramadan, etc.), an
election, an episode of catastrophic weather or other remarkable event. The end of the recall
period should be the interview date. The recall period is commonly set at around 90 days.
Average household size and percentage of children under 5 years
In nutrition surveys, although children are the primary target, it is households which are selected,
hence the necessity for calculating the number of households and estimating the number of
children and vice versa. This calculation requires knowledge of the average household size and
the proportion of children below 5 years.
22
Non-response rate Non-Response Rate (NRR) accounts for households that could be either absent, not accessible,
refuse to be surveyed, or any other reason that prevent survey teams from surveying a selected
household. The sample size is accordingly inflated using this NRR.
Procedures for sample size calculation
In calculating the sample size for anthropometry, the following formula is applied by ENA for
SMART in automatic calculation to determine the number of children required:
Anthropometry

where:
z =critical value for the normal distribution at 95% Confidence Interval (C.I) = 1.96
p:=estimated prevalence (as a decimal)
q=1-p
Cluster sampling
Sample size (n) = t² x p x q X DEFF

t=2.045 (Note that the t-distribution is used for cluster sampling).
The number of children is then converted to the number of households using the following
formula, which assumes that children 6-59 months constitute 90% of all children below 5 years:
Sample size households (N) = Sample size children (n)
(Average household size x % children under 5 x 0.9)
Example of determining the number of clusters.
The sample size for anthropometry was done on ENA for SMART software planning
screen and was based on the following assumptions; .
i) Average Household size 6.1 (SHHS 2006)
ii) Under five children 16.7%. = 1.02 under fives in each household
iii) Children 6-59 months 17.9% = 1.09 under fives in each household.
iv) Estimated prevalence – 20% Realistic estimate based on Rapid assessments
throughout 2006 to 2011. Couldn’t use SHHS 2006 since the time is too long before.
v) Desired precision = 4; Reason: There is a plan to repeat the survey during the post
harvest of 2012 for comparison purposes.
23
vi) Design effect = 1.5 (DEFF for malnutrition usually falls between 1.4 and 1.8)
vii) Percentage of none response households = 3 based on reports of rapid assessments
that have shown zero refusals.
viii) Therefore anthropometric sample
Calculation of cluster size.
i) Working day from 7 am to 7 pm = 12 hours
ii) Time taken to and from the field on average = 1.5 hours (90minutes).
iii) Time to sample and identify 1st household = 30minutes
iv) Time for lunch and breaks = 1hour (60 minutes)
v) Time to interview once household = 25 minutes.
vi) Time to walk from one household to another = 5 minutes.
vii) Calculation.
- Total working hours = 12 x 60 = 660 minutes.
- Total time spend in the field undertaking various survey tasks = 90min +
30min = 120minutes.
- Time available for the teams to undertake the survey = 660 – 120 =
540minutes
- Therefore, number of households to be visited / day = Time available for the
team divided by time spend interviewing and moving from one household to
another= 540/30 = 18 households.
viii) Calculation of number of clusters per survey = Sample size1 (Household) ÷ size of
cluster = 713/18 = 40 clusters.
Note: This is an hypothetical example and parameters may differ case by case depending on the
situation.
Mortality
Systematic random sampling
The sample size (households) is calculated as follows, whereby CDR is the estimated Crude
Death Rate:

Cluster sampling
11 The higher of the sample among Anthropometry and Mortality is used for the calculation of number of clusters per
survey. In this case 713 households for the mortality.
24

To calculate the sample size, the parameters explained in Section 4.4 are entered into the planning
screen of ENA for SMART. For ease of understanding, we will use a single hypothetical
population and apply the three sampling methods separately to show how the sample size will be
calculated.
50,000 people and 10,000 households:
Estimated prevalence of GAM: 20%
Estimated CDR: 2/10,000/day
Desired precision- 5%
246 children for anthropometry and 337
households (corresponding to 1,518 population)
for mortality (Figure 4). Given that, in reality, the
anthropometry and mortality data is collected
from the same households in a single survey, the
higher number of households is taken as the
sample size (337 in this case) to ensure that the
sample is sufficient for both. It is important to
note that survey teams should collect data from
all 337 households even if the 246 children are measured before all the households are completed.
This ensures that survey teams do not select households only with children, which can introduce
bias. This method, known as the fixed household method, is also useful as nutrition surveys
frequently include the collection of household-level data such as food security, which requires all
households to be included for a more representative picture.
Simple random sampling
In terms of implementation, with simple random sampling, the next stage is to randomly select
the 337 households from the 10,000 households, using the random number generator, also found
on the ENA for SMART planning screen, by entering the range of required households (1-
10,000) and the number required (337) and clicking “Generate Table” (Figure 5) to produce a list
of randomly selected households. These numbers should then be sorted in ascending order. From
the list of households, the households are then selected. If, for example, the first number on the
list is 964, meaning that, from the list of households, the 964th household will be the first to be
selected.
25
Systematic random sampling
The total number of households (10,000) is divided by the sample (337) to determine the
sampling interval (30 rounded to the nearest whole number in this case). A random number is
then selected between 1 and 30 to determine a starting point (using the procedure described
above). Assuming the random starting point is 5, the survey team will go to the 5th household as
per the layout. This is the first household that will be interviewed. The next household will be
determined by adding the sampling interval, which will be the (5+30)th household, which will be
the 35th household. The next household will be the (35+30)th, which is the 65th household. This
process is repeated until the 337 households are achieved.
Figure 5 Random number generator
Random Number table
Range: 1 to 10000, Number: 337
964 5461 1320 4544 8169 5635 1270 1456 7280 7644 2105 2098
5471 6759 4630 298 8404 1340 8435 412 7369 2932 4219
8311 1124 9839 3543 9534 3528 9653 3363 1319 556 8712
5178 7722 2795 8945 1144 9559 2006 5082 1993 3377 2864
7696 4101 6176 3444 35 5556 5104 571 9130 3555 3844
8612 4137 7989 9897 2810 7386 8555 9512 755 6097 5227
9443 8079 5245 4455 7279 6978 4535 5906 4558 8645 4958
7115 9381 3221 2891 5874 7747 3338 2220 2718 5324 9870
7449 8330 3885 968 3256 3230 9118 7985 8489 9581 9028
4373 5036 7462 4659 6157 4569 435 7344 5078 8615 7780
9216 7986 9921 3207 6175 8597 5597 2422 5059 188 4981
4359 344 7726 3451 9202 1327 44 2761 7828 2182 9603
329 1272 1165 2036 3577 470 9584 4521 9113 9511 650
4421 7778 256 9433 3432 7283 4890 2516 226 7277 640
8437 7039 4189 3361 2517 8195 5693 9226 2169 1009 7128
6225 1277 2064 4735 9700 4740 4310 3037 8629 6777 2965
6216 1887 9088 682 5826 6187 7458 4953 9689 9938 7901
9319 9725 5949 3659 5709 8636 5153 2099 5835 8154 3425
8066 4718 7831 8278 5398 381 9466 4589 5412 1558 4999
959 3082 481 9852 3273 5609 5922 6549 4078 1623 953
8672 3502 2111 6500 8335 5226 2199 1735 5741 4691 3089
983 880 9445 4963 4899 3245 5551 8599 7149 7585 6318
5879 1348 454 4436 7347 3198 2790 4236 393 5801 8358
26
3301 4173 8621 417 5140 9843 3046 1427 7140 7218 42
3175 525 1434 7846 7126 1536 5717 183 7381 8848 7964
1051 8354 8419 9949 3413 1296 3473 318 7688 9541 2627
3650 1777 5128 6258 254 6804 9067 3866 7335 4813 9193
191 3955 5456 2978 6180 262 4019 7045 3411 3961 5629
7583 9482 2993 8389 7827 7496 5761 8917 7558 6017 1949
9001 9896 8306 1394 1393 3403 2250 6468 81 6207 6896
1930 4011 6249 5552 9829 8451
4.5 Cluster sampling
ENA calculation
With the same example, to determine the sample size for cluster sampling, a DEFF must now be
entered in the planning screen of ENA. The population sizes of the smaller units must also be
entered to generate clusters. Let us use a DEFF of 1.5 and the sub-populations in Table 3.
Table 3 Cluster sampling example Geographical unit Estimated total population
Location 1 5,000
Location 2 6,000
Location 3 6,500
Location 4 6,200
Location 5 7,000
Location 6 5,900
Location 7 7,100
Location 8 6,300
Total 50,000
A decision must then be made as to the number of clusters to be used for the survey. The higher
the number of clusters, the higher the probability will be that the sample will be truly
representative of the population from which it is selected. This is because, the more clusters there
are, the smaller the confidence interval will be around the estimate of the prevalence of
malnutrition and the more accurate our estimate of malnutrition will be. According to SMART
guidance, the number of clusters are recommended to be 30 or more, and cannot go below 25
under any circumstanced. In this example, let us assume that we will use 30 clusters. The output
from ENA for SMART, using the same assumptions above, gives 551 households and 401
children. In this example, we will use 30 clusters. After entering the variables in ENA for
SMART, and clicking “Assign Cluster’, the clusters are selected from the population sub-units.
By clicking the icon , the table with the selected clusters is copied to Microsoft Excel (Figure
6).As shown, the 30 clusters have been selected, meaning that location one will contain clusters 1
to 3 and so on. Note that the software also generates replacement clusters (RC) which are
substitute clusters to be used in the event that 10% or more of the clusters are unreachable, either
due to insecurity of other reasons. In such circumstances, all the RCs will be used.
27
Output:
Geographical
unit
Population
Location 1 5000 1,2,RC,3
Location 2 6000 4,RC,5,6
Location 3 6500 7,8,9,10
Location 4 6200 11,12,13,14
Location 5 7000 15,16,17,18,19
Location 6 5900 20,21,22,23
Location 7 7100 RC,24,25,26,27
Location 8 6300 28,29,30,RC
Having selected the clusters, the next stage is to select households. In this example, the sample
required is 551 households, meaning that there will be 551/30=18.4 children per cluster. This will
be rounded up to 19 (it is advisable to round up rather than down). The 19 households within each
cluster must be randomly selected. The recommended approach is to use simple random sampling
of households from a list, or systematic random sampling if a list is unavailable and households
are arranged in a clear pattern. It may be that villages are not very large but village leaders are
still unable to list all households. In such situations, survey team members can walk around the
village and identify all households, by writing a number (starting at 1 to the total number of
households in the village) with a chalk on their door, for example. If clusters have been selected
from villages, the list will be obtained from the village leader. Only in the event that the list of
households is unavailable, the modified EPI method is used.
Step 1
Step 2
Figure 7 Modified EPI method
When the team arrives at the village that will contain the cluster, the following procedure should
be followed after discussions with the village leaders (Figure 7).
Go to somewhere near the centre of the selected cluster area.
Randomly choose a direction by spinning a bottle, pencil, or pen on the ground and
noting the direction it points when it stops.
Walk in the direction indicated, to the edge of the village. At the edge of the village
spin the bottle again until it points into the body of the village. Walk along this
second line counting each house on the way.
Using a random number list select the first house to be visited by drawing a random
number between 1 and the number of households counted when walking. For
example, if the number of households counted was 27, then select a random number
between one and 27.
The team will not have a computer in the field, so each day before setting out, the
supervisor should print out the list of random numbers. If the number 5 was chosen,
go back to the fifth household counted along the walking line. This is the first house
you should visit.
Go to the first household and interview all children aged 6–59 months in the household
for the nutritional survey and complete the mortality questionnaire.
The subsequent households are chosen by proximity. In a village where the houses are
closely packed together, choose the next house on the right.
29
Continue in this direction until the required numbers of households are interviewed.
Continue the process until the required number of children has been measured.
The modified EPI method understandably gives less representativeness as children will be
selected with close proximity to each other, which has potential for bias. Additionally, the
selection of a household is actually determined by the selection of another, which contradicts the
principles of probability sampling.
Segmentation
In some cases, villages selected randomly to contain a cluster might be very large or households
very dispersed and sample selection can then become very tedious; teams will have long distances
to walk and not enough time to complete one cluster per day. In those scenarios (approximately
more than 100 households in the village), segmentation can be used in order to reduce the area
that will be covered by the survey teams. The objective of this procedure is to divide the village
into smaller segments and choose one segment randomly to include the cluster. This division can
be done based on existing administrative units, such as natural landmarks (river, road, mountains,
etc.) or public places (market, schools, churches, mosques, temples, etc.) The segments should be
preferably be of approximately equal size, whereby the team will randomly select one segment to
be the cluster. This should be accompanied by a sketch map.
4.6 Adjustment for small sample size
If the target population (number of children 6-59 months) is below approximately 10 000, the
“Correction for small population size” box is clicked in ENA for SMART (Figure 7 and 8). This
is because, If the target population is small, a smaller sample size would be needed to achieve the
required precision. ENA for SMART calculates the target population from the total number
entered in the cluster selection table and % of Under-5 children entered into the calculator. For
example, if the total population size in the cluster selection window is 40,000, and % of under-5
is 15%, ENA for SMART would assume that there are 40,000x0.15x0.9=5,400 children aged 6-
59 months in this sampling universe, and use this number for adjustment for small population size
Figure 8 shows the effect of adjusting for small population size. With the same assumptions, the
required sample size has reduced from 496 to 485 households for anthropometry.
5. SURVEY FIELD
This section will highlight ways of
ensuring that field teams are well managed, and will discuss measures for reducing bias during
data collection. Special field circumstances in sampling and selection will also be explained.
Figure 8 Adjustment for small population size
30
5.1 Bias
Bias is anything other than sampling error that causes the results of the survey to be different
from the actual population prevalence. Bias cannot be calculated nor its effect upon the result
assessed, but is the main reason for inaccurate survey results. It is important for survey teams to
understand potential sources of bias and to minimise them.
Common sources of bias in nutrition surveys include:
Systematic errors due to faulty weighing equipment or measuring techniques.
Non-calibration of weighing equipment.
Recall error: Respondents often fail to recall all deaths during a given recall period.
Infant deaths, in particular those within a short time after birth, are particularly under-
reported. Respondents may also misreport ages, dates, and salient events.
"Calendar" error: Respondents may report events as happening within the recall period
when they did not (or vice versa) due to lack of clarity about dates.
Age heaping/digit preference: Respondents may round ages to the nearest year i.e. 12,
24, 36 and 48 months.
Sensitivity/taboos about death: In general, the death of a household member is not a
subject discussed readily with strangers.
Deliberate misleading: In some populations, with experience of relief operations, some
respondents may deliberately give incorrect answers in the expectation of continuing or
increased aid.
Interviewer error: Enumerators may ask questions or write down answers incorrectly,
skip questions, assume answers, or rush respondents in an effort to complete the
interview quickly.
The best way to minimise bias is to thoroughly train and supervise teams, ensuring that all
procedures outlined in section 6 are strictly followed. Additionally, the following minimise bias
and enhance accuracy of data:
Ensure errors in the field are minimised by using good quality equipment that is regularly
calibrated.
Check the questionnaires and control forms for blank entries at the end of each day to
make sure no data is left out. The team leader should review all questionnaires before
leaving an area in order to make sure no pieces of data have been left out. If there are any
problems the team can return to the household and correct any identified error.
Check for data collected. Each evening, or during the next day while the teams are in the
field, the supervisor should arrange for data to be entered into the computer. Recording
errors, unlikely results, and other problems with the data may become clear at this stage.
ENA software will automatically flag abnormal values as data are entered.
Each morning, before the teams set out for the day, there should be a short feedback
session. If any team is getting a large number of flagged results, the supervisor should
accompany that team the next day. If the results are very different from those obtained by
the other teams, it may be necessary to repeat the cluster from the day before.
5.2 Supervising data collection team
Field supervision is important in ensuring valid data collection and minimising bias. The
supervisors should:
Make frequent unannounced spot checks on the teams in the field.
Ensure that the methodology is closely followed and document any deviations.
31
Check all questionnaires and control forms to ensure that all sections are accurately
completed.
Ensure that all instruments to be used the survey teams are calibrated every day.
It is particularly important to check cases of oedema, as there are often no cases seen
during the training and some team members may therefore be prone to mistaking a fat
child for one with oedema.
5.3 Special circumstances
There are certain circumstances which field team may encounter, which must be anticipated.
Some households or children may be absent or refuse to participate, and in some cases there may
be need for teams to return to the households later in the day. All these must be documented, and
it is strongly recommended to use a cluster control form for this purpose (Annex 10).
Impossible to visit a selected household
In the event that a household cannot be interviewed, either due to refusal or lack of access, the
team should continue to the next household according to the sampling procedure. The households
that are impossible to visit have already been accounted for in the planning stage by inflating the
sample size with the non-response rate.
No children in the household
Not all households will have children. All applicable modules/questionnaires should be
completed if there are no children. Excluding households without children from selection will
introduce serious selection bias in measuring household-level and other non-child variables (e.g.,
mortality, WASH, food security).
Absent household
The survey team may find all household members absent. After confirming with neighbours, this
should be recorded on the cluster control form. The team should return to absent households
before leaving the village, to see if residents are back. If not, this should be reported on the
questionnaire and control form. As explained above, absent households are not replaced.
Absent household
This is a household which has had no one living there for a long time. Such households should
not be included in the list of households used for household selection.
Absent household
If a child lives in the household but is not present at the time of the survey, this should be
recorded on the household questionnaire and control form, and the household should be revisited
before the end of the day. The rest of the information (age, sex, feeding practices, immunizations,
etc.) can still be filled completed.
Child with disability
Some disabilities might not allow you to take all anthropometric measurements needed or might
lead to a biased measure. For example, the weight of a child missing a limb will not be very
meaningful when comparing it with the standard population. All other data that is not influenced
by the disability should be collected such as sex, age, oedema (if the child has both feet), etc
should be collected.
Such children should be recorded as absent.
6. SELECTION AND TRAINING OF SURVEY TEAMS
This section will outline guidance for selecting survey teams, and the aspects to be included in
training of enumerators.
6.1 Selection of survey teams
Survey team members do not necessarily have to be health professionals. However, They need
to be fit, as there is usually a lot of walking. They should have a relatively high level of
education, as they will need to read and write fluently, and count accurately. Ideally, they will
speak the local language. If not possible, there should be interpreters as part of the survey teams.
Women generally have much more experience dealing with young children and should usually
lead the interviewing of mothers/caretakers of children. This is also important as some cultures
do not allow women to be interviewed by men. The gender composition of the team should
conform to the local context.
The composition of survey teams depends on the data to be collected. Two people are required
for measurement of children (measurer and recorder) in addition to an interviewer. A team
leader is also required for quality control and leadership of the survey team. If there are
additional modules such as food security and water and sanitation, which are household
modules, an additional member may be required.
Generally, four to six teams survey teams may be needed depending upon the number of
households to be visited, the time allocated to complete the survey, and the size and the
accessibility of the area covered. The number of teams should never be too many despite the fact
that the more the teams, the faster the data collection. The quality of the data deteriorates with too
many teams as it is much more difficult to train, supervise, provide transport and equipment, and
organize a large number of teams. Supervisors should be assigned to each team. If the teams are
to collect data in nearby areas, there may be a supervisor for two teams, but if they are far apart, a
supervisor may be required for each team. The supervisor must be experienced in undertaking
nutrition and mortality surveys, training team members, organizing logistics, and managing
people.
6.2 Training of survey teams
Survey team members must receive adequate training prior to conducting a survey, even if they
have prior survey experience. Each survey has its unique challenges, and survey work requires
constant re-training so as to standardize methods and techniques as well as to update knowledge.
The survey manager must come up with a survey training schedule and organise a suitable
training venue, where there is sufficient space. The equipment and materials for the training need
to be procured and organised in advance of the training.
The main topics to cover in training of data collectors include:
Introduction to nutrition surveys: To introduce team members to the rationale behind
surveys and the objectives.
Sampling procedure: Defining sampling, explaining why sampling is used and how it is
applied. Describe the rationale and importance of representativeness and outline the
sampling method to be used for the survey.
Interviewing techniques and questionnaires: Explain the best practice in terms of
interviewing so as to prevent bias. Go through each survey question to give guidance on
33
suggestive questioning but probe where necessary.
Measurement techniques: Introduce the teams to the different measurements to be used in
the survey and the equipment to be used, and explain procedures for each. This is
followed by a standardisation test exercise at the end of the survey for anthropometry.
Composition of survey teams, roles and responsibility of team members: Discuss and
agree on the composition of the different teams and assign team leaders and supervisors,
clearly explaining the responsibility of each.
Survey field procedures: Go through the procedures to be followed by teams before going
to the field, whilst in the field, and after leaving the field on a daily basis.
Survey logistics: Outline the plans for the survey regarding: materials and equipment to
be used by each team, communication, travel, food and accommodation in the field as
well as allowances (if applicable).
6.3 Roles and responsibilities of team members
The roles of the different members of the survey team are generally as follows:
Survey manager (1 per survey)
1. Gathering available information on the context and survey planning.
2. Selecting team members.
3. Training team members.
4. Supervision of the survey: Taking necessary actions to enhance the accuracy of data
collected.
5. Visiting teams in the field and making sure that supervisors are following up team leaders.
6. Ensuring that households are selected properly and, that the equipment is checked and
calibrated each morning during the survey, and that measurements are taken and recorded
accurately.
7. Deciding on how to overcome the problems encountered during the survey. Each problem
encountered and decision made must be promptly recorded and included in the final report, if
this has caused a change in the planned methodology.
8. Organizing data entry into ENA for SMART and checking any suspect data every evening,
by using the appropriate sections of the SMART plausibility report.
9. Organizing an evening or morning “wrap-up” session with all the teams together to discuss
any problems that have arisen during the day.
10. Ensuring that the teams have enough time to take appropriate rest periods and has
refreshments with them. It is very important not to overwork survey teams since there is a lot
of walking involved in carrying out a survey, and when people are tired, they may make
mistakes or fail to include more distant houses selected for the survey.
11. Analyse and write the report.
Survey supervisor (1or 2 per team depending on survey)
1. Visiting teams in the field and making sure that before leaving the field, each team leader
reviews and signs all forms to ensure that no pieces of data have been left out; making
sure that the team returns to visit the absent people in the household at least twice before
leaving the area.
2. Checking cases of oedema, as teams are prone to mistaking a fat child for one with
oedema (particularly with younger children).
3. Ensuring all food/refreshments are ready at the start of the day.
34
4. Participating in deciding on how to overcome the problems encountered during the
survey. Each problem encountered and decision made must be promptly recorded and
included in the final report, if this has caused a change in the planned methodology.
5. Assisting with organizing data entry into ENA for SMART and checking any suspect
data every evening, by using the appropriate sections of the SMART plausibility report
and other checks.
Team leader (1 per team)
1. Ensuring that all questionnaires, forms, materials and equipment are ready at the start of
the day.
3. Organising briefing meeting with team before departure in morning.
4. Speaking with representatives to explain the survey and its objectives.
5. Using a local events calendar to estimate the age.
6. Checking if the child is malnourished and making a referral if necessary.
7. Supervising the anthropometric measurements.
8. Ensuring that households with missing data are revisited before leaving the field the same
day.
9. Ensuring that all equipment is maintained in a good state.
10. Managing time allocated to measurements, breaks and lunch.
11. Ensuring the security of team members.
12. Noting and reporting problems encountered to the supervisor.
Measurer (anthropometry) -1 per team
1. Measuring weight, height, weight and MUAC.
2. Assessing the presence of oedema.
Assistant (anthropometry) -1 per team
1. Using local events calendar to estimate age.
2. Recording age.
3. Ensuring each child is in the correct position for measurement.
4. Recording the measurements on the questionnaire.
Interviewer -1 per team
6.4 Standardization test
The standardization test consists of all the members of the teams measuring 10 (or more)
different children twice, with a time interval between individual measures. The size of the
variation between these repeated measures is calculated to assess how precisely each person
measures the children (repeatability of measurements).
The standardization exercise is performed with a group of children whose ages fall within the
range for the study (6–59 months). Before carrying out the exercise, the supervisor carefully
measures each child without allowing the trainees to see the values. The supervisor is
automatically given the ID number 0, and should start by filling in the form. It is important that
the supervisor undertakes the exercise as well as the team members. The supervisor’s data may
35
be assumed to be of higher quality than the trainees; however the actual values should relate
closely to the mean value for all the teams.
Each team member is also given a unique ID. Each child that will be measured is also be
given a child-ID, starting from 1. For the exercise, each child, with his/her mother, remains at
a fixed location with the ID number clearly marked. The distance between each child should
be far enough to prevent the trainee from seeing or hearing each other’s results.
At the beginning of the exercise, each pair of trainees starts with a different child. The supervisor
instructs the measurers to begin the measurements. The trainees should carefully conduct the
measurements and clearly record the results on the second, third and fourth columns of the
standardization form next to the child’s identification number (Table 4a and 4b).
Table 4 Standardisation test
Enumerator name....................ID......1st measure Enumerator name....................ID......2nd measure
37
Each pair of measurers should have their own form to complete, and each should take turns
taking measurements. When each member of the pair has done the measurement, they should
move on to the next child. At the end of the process, the sheets are handed in and a second sheet
is taken. The teams then take a break (lunch). The whole process is then repeated after the break.
Thus, without seeing the measurements they previously made, each enumerator measures each
child twice.
The equipment used in the exercise should be the same equipment used to measure children in
the survey itself. The team members will rotate but the equipment should not, so that each child
is always measured with the same equipment (the team is being tested not the equipment). Only
one pair of measurers should be with a child at any one time. Talking between pairs of trainee
measurers during this exercise should not be allowed.
Upon completion of the exercise, the data is entered into the “Training” screen of ENA for
SMART, as shown in Figure 9. On clicking , a report is
generated in Microsoft Word, giving the accuracy of each measurer either as “good” or “poor”.
Each team member’s measurements are compared to the mean of the whole group to assess how
accurately the measurements are made. Each team member is then given a score of competence
in performing measures. Any misunderstandings or errors in technique are corrected during the
training. Any team member unable to make the measurements sufficiently well should be
replaced or given a different job in the survey that does not require taking the primary
measurements.
39
6. DATA COLLECTION .
This section will explain the procedures for data collection for anthropometry, mortality as well as
additional indicators, which include morbidity, infant and young child feeding.
7.1 Anthropometry
Anthropometric measurements (measurements of body proportions, such as weight and height) are
used to give an approximation of the nutrition status of a population, or to monitor the growth and
development of an individual. At the individual level, anthropometric data is used to determine
whether or not an individual is malnourished. In turn, this information may be used to decide whether
or not the individual should be included in a supplementary feeding programme, or treated for severe
malnutrition. The information is also used to decide when to discharge the individual from a feeding
programme. At the population level, anthropometric data is used either in a one-off survey to assess
what proportion of a population is malnourished, or as a surveillance tool to follow the nutritional
situation of a population. Collectively, the anthropometric measurements of children aged 6-59
months may be used to compare different populations, or to make comparisons of the same population
over time. Anthropometry data is mandatory for all nutrition surveys as it forms the basis for
determining the magnitude of malnutrition.
Nutrition indices and indicators When body measurements are compared to a reference value they are called nutrition indices. Three
commonly used nutrition indices are weight-for-height (WFH), weight-for-age (WFA) and height-for-
age (HFA). The indices are shown in Figure 10. As an alternative to weight-for-height (WFH),
wasting can also be measured by Mid Upper Arm Circumference (MUAC), which is relatively easy to
measure and a good predictor of immediate risk of death for children 6-59 months. It is used for rapid
screening of acute malnutrition. MUAC can be used for screening in emergency situations but is not
typically used for evaluation purposes
Figure 10 Indices of nutritional status. Source: ENCU/PPDA (2002).Guideline on Emergency Nutrition
Assessment
Nutrition indicators are an interpretation of nutrition indices based on cut-off points. Whereas indices
are simply a figure, indicators represent an interpretation of the indices. For example, WFH is an
index of nutritional status, whereas low WFH is an indicator. Anthropometric indices or indicators are
most commonly expressed as z-scores. A z-score is a measure of how far a child’s measurement is
from the median value of the reference distribution.
For example, weight-for-height z-score (WHZ) is based on:
a) The child’s weight
b) The median weight for children of the same height and sex in the reference population
c) The standard deviation for the distribution of weights in the reference population for
40
children of the same height (because the standard deviation of a distribution increases as
children get older, you need to use the standard deviation for the reference distribution of
children of the same height).
WHZ = actual weight – median weight
standard deviation for reference population
The median values and standard deviation are contained in the reference population. The
recommended reference to use is the In WHO 2006 reference (Annex 1), although when data is
analyzed by ENA for SMART, results are also presented using the NCHS 1977 child growth
standards for purposes of comparison only in the annex.
Classification of nutritional status
Chronic malnutrition, underweight and wasting
A z-score below -2 for any indicator defines moderate malnutrition, whilst a z-score below -3 defines
severe malnutrition. For example, a WHZ below -2 is classified as moderate wasting, whilst a WHZ
below -3 is classified as severe wasting.
Acute malnutrition
Acute malnutrition is