In deciding the form and content of a nutrition survey, the social structure of the community of an area must be considered. Social structures between people may be built on biological, economic, political, religious and/or geographical grounds. These social structures are assembled by various levels. Structures are built starting with the smallest survey unit (the individual) through families, communities, etc. up to the level of the population of a country or region (see table 2).

Great differences can also be found between urban and rural areas. Although in both urban and rural areas the smallest unit is the individual (survey level A), at the next higher level (survey level B) the social organizational structures start to differ from each another. In rural areas, the household unit is a family or extended family. In urban areas, households consist of single families or individual persons.

The complexity of the social structure is also important. The higher the survey level, the more complex the social organizational structure. In addition, these structures grow even more complex with increasing numbers of people in an interdependent living environment.

As shown in the following table, the quality of a higher survey unit is different from the sum of the subordinate survey units. In order to describe the quality of the whole, all levels must be described. If one wants to undertake a nutrition survey of a population, this principle also applies to the nutritional assessment of individuals. Not only should the sum of individual units be assessed, but the higher structural levels must also be assessed.

Therefore, a survey must allow for the special characteristics at the level of the individual, household, and village or suburb. The sum of information on individuals alone does not provide a complete picture of a family situation, and similarly the sum of information on families does not do so for villages or suburbs, etc.

Table 2. Observable characteristics related to the survey level
 

Survey level	Survey unit		Examples of observable characteristics
	Rural areas	Urban areas
A	Individual	Individual	Age, size, sex
B	Household	Household	Size, income, religion
C	Village	Suburb	Electricity supply
D	Valley (geographical unit)	City/town, part of a metropolis	Health structures, types of schools
E	District (political)	Metropolis	Climate

Information concerning survey levels A and B has to be obtained by direct interviews and data collection from the target group. Variables concerning individuals will be described in sub-chapters 3.2.10 - 3.2.13, and those concerning households in sub-chapters 3.2.1 - 3.2.9. Inquiries should be made at the appropriate agencies concerning the variables at survey levels C, D and E before household interviews are conducted. Part 2 discusses how to obtain this information.

In a nutrition survey the data collection is divided into

• interviews (e.g., socioeconomic and health related questions)
• observations (e.g., socioeconomic and health related questions)
• measurements (e.g., anthropometric and biochemical data).

3.1. Survey design
 

3.1.1 Initial interactions with community

To obtain an actual view of the nutritional situation of a target community and to lay the foundation for intervention, an essential requirement is that the survey is planned, carried out and evaluated only in cooperation with the community concerned.

This means that from the very beginning, a dialogue is sought with representatives of the target population. In this way the community will be informed as to the purpose of the survey and its form and contents. Before any planning of the survey method and its contents, time and cultural requirements must first be defined. If parts of the survey cannot be carried out due to the wishes of the community, and therefore, some information cannot be gathered, this objection must be accepted and respected. For example, taking blood samples from children or asking questions concerning income is often rejected by community members.
 

F	A basis of trust is a necessary starting point for any future cooperation with the community and paves the way for sustainable intervention.

During, or at the latest upon completion of, a survey a dialogue must be initiated with the local population concerning the results of the survey. At this opportunity

• the people should be informed of the results,
• their opinions on nutritional problems should be noted,
• their wishes and reservations should be considered in the formulation of intervention, and
• the potential for indigenous intervention should be identified.

An important principle in planning and evaluating a survey is to distinguish between the following community groups:

1. The total population, i.e., all people living in the project region.
2. The target group, i.e., those people for whom and with whom the project is planned and carried out.
3. The risk group, i.e., the section of the population with the worst nutritional conditions and in greatest danger of sickness and death from malnutrition.


A distinction has to be made between nutritional physiological risk groups and social risk groups. Physiological risk groups are, for example, small children, school children, nursing and pregnant mothers, elderly people, and men and women who perform physically hard work. Examples of social risk groups are the landless, low income or other groups displaying particular forms of behavior (e.g., nutritional or health practices).
 

4. The beneficiary group, i.e., those people who will benefit from the project intervention measures.
5. The survey group, i.e., those persons who provide personal data relevant to nutrition (e.g., weight, height, immunization status, diarrheal diseases).

During planning and evaluation the distribution of individual groups among the other groups needs to be defined. From the baseline survey the proportion of the risk group to the total beneficiary group can be estimated and the proportion of the beneficiary group to the total risk group can similarly be estimated.

This is illustrated in the following diagram:
 

		Risk group
		yes	no
Beneficiary group	yes	a	b
	no	c	d

1. Planned degree of intervention:

Proportion of total beneficiary group that is at risk = a / (a + b) * 100

With a known degree of intervention, the efficiency of the assistance provided can be determined; in other words, the proportion of efforts that reach those who need them will be established.

2. Planned coverage of intervention:

Proportion of the total risk group that receives benefits = a / (a + c) * 100

The coverage of intervention will determine the proportion of the population at risk that will benefit from intervention.

As described earlier, a project would be overtaxed both in terms of schedule and expertise if all risk groups were to be considered in a nutrition survey, therefore:
 

F	The nutrition survey methods discussed in this publication are restricted to observations on infants and small children up to 59 months.

The reasons for this are as follows:

1. Latent nutritional problems influence health-related infant and child (0-59 months) mortality. The nutritional condition of this risk group serves therefore as an indicator of the nutritional situation of the entire target group.

Additionally, the type, severity and causes of risks differ with each age group of children.

• In terms of nutritional physiology, the most vulnerable group is infants (0-11 months) because they do not have fully developed immune systems, are changing from breast milk to plant and animal foods, and are exposed to risks of infection due to their increased mobility.
• In terms of epidemiology, the most vulnerable group is children aged from 6 to 23 months. The type and cause of the risks can also differ within this group, e.g., between the 6-11 months age group and the 18+ month age group.

2. The direct and indirect causative factors for malnutrition based on intensive studies of this child group (0-59 months) have been established. Similarly, previous experience in intervention measures can also be applied to this group.

3. Although it is impossible to recover fully from nutrition and health related stunting after a child has reached the age of two, children older than two should be surveyed for the following reasons:

• acute energy deficiency problems can be easily diagnosed and treated in this age group, and
• follow-up investigations of long-term interventions should be made after four years at the earliest because successful nutritional intervention from infancy should have manifested itself by then. Therefore it is important to have baseline data at the older age for comparison with the follow-up data for children who had participated in the project.

 
The broader coverage of age groups also has the advantage that more children can be covered for a given number of households surveyed. This produces a larger sample size without requiring extra effort to visit more households.

3.1.3 Cross-sectional vs. longitudinal surveys

The survey design needs to be tailored in a way that the maximum amount of information can be collected regarding the impact of the planned intervention. For example, the impact of a oil palm plantation project on the nutritional situation of its smallholders/producers is studied according to table 3.

Table 3. Survey design of nutrition surveys
 

LONGITUDINAL	CROSS-SECTIONAL
	Intervention Group	Group for comparison
Baseline survey (start of project)	(A)	(B)
1. Follow up survey	(C)	(D)
2. Follow up survey	(E)	(F)
etc.	...	...

A baseline survey at the start of the project (A) gives information about the nutritional situation of the project population. After several years a follow-up survey will be carried out within the same area or population (C) to study the project's impact on the nutritional situation. However, the comparison of the nutritional situation between (A) and (C) cannot be used alone to measure the impact of a project because several additional factors that have changed during the time of the implementation of the project may have influenced the nutritional situation of the population. It is difficult to attribute the project's impact to the change of the nutritional situation. If the general condition of the population in the area, but not in the project, deteriorated during the intervention period, for instance due to climatic or political factors, no change of the nutritional situation of those in the project area has to be regarded as a success because the nutritional status of the rest of the project's surrounding population has decreased. Therefore, it is recommended to include in the baseline survey a population that will not be included in the project (B). Often projects are carried out in phases and start with a smaller population group for a later expansion. The group of comparison (B) could be identified as a future expansion area. During the follow-up survey the population of the group for comparison will be surveyed again (D) and this survey can be used as their baseline survey. After several years a follow-up survey will be carried out within the same population (F) to study the project's impact on the nutritional situation of group (D).

If no comparison group is available, the results must be analyzed with extreme caution because it is difficult to separate changes due to the project from those independent of the project. If, for example, no improvement in the nutritional situation is observed within a specific time, it cannot be definitely concluded that the measures have not worked. It is possible that general worsening conditions have caused a general decline in the nutritional situation of the entire region. Under these conditions, if the original nutritional situation has been maintained, the results have been positive. Alternatively, observed improvements in the nutritional situation may not have much meaning by themselves as long as they are not compared to the overall situation. It is conceivable that improvements in nutritional indicators can be attributed to the general situation in the region and not to the measures undertaken in the project.

To further complicate matters, comparison groups are never fully identical to intervention groups. This problem can be reduced statistically by studying many individuals. However, this significantly increases the cost of a survey.

It can be said that in principle:
 

F	The nutritional situation of the target population should be assessed by case control studies through periodical surveys with comparison groups.

It has to be recognized that the assessment of the nutritional situation of population groups for the purpose of comparison may create ethical problems, since no immediate actions are planned to improve their situation (see section 3.1.3.3).

There is a special type of long-term study undertaken with the same individuals over the period of the survey. This survey method can be more economical, but in certain cases, more difficult to carry out. The primary advantage of this method is that a smaller number of individuals provides statistically valid results. Such a study cannot always be carried out because it is necessary to ensure that the same individuals will be available throughout the entire study. However, at the outset of an intervention this cannot be guaranteed for such reasons as migration, death, etc. This type of study also raises concerns about whether the results are due to the intervention or to the repeated surveys. Experience has shown that surveys carried out without any implementation of intervention measures can lead to an improved nutritional situation for the survey group.
 

3.1.3.1 Baseline survey

Each nutrition-related project should start with a nutrition survey. This first (baseline) survey is much more extensive than a follow-up survey, not only because the nutritional condition of the community has to be assessed, but also the possible causes (such as inappropriate weaning practices or child nutrition, frequent diarrheal diseases, etc.) and important determining factors must be identified. Only if the causes of undernutrition are known, can sensible and equitable intervention measures be planned and carried out. The success of intervention can be checked and evaluated by establishing the determining causes of the nutritional condition. The analyses of causes and determinants will be discussed later in further details (see Chapter 4.7).

Before each baseline survey there must be a comprehensive collection of background information (see part 2 of the guidelines). Because the surveys are carried out in developing countries, where it is often difficult to obtain scientific publications, literature research should be conducted in advance. In addition, at least one week should be allocated for the collection and analysis of information in the project country.

Most projects start with an orientation phase in which specific sector data for the project are assessed to analyze the starting situation. If during the orientation phase data are obtained which are also relevant to the nutritional baseline survey (e.g., socioeconomic data), perhaps both surveys can be combined. If the nutrition baseline survey is a part of an overall socioeconomic survey, the assessment of socioeconomic data can also be carried out by other persons in the survey team. Care should always be taken that the community is not overburdened with unnecessary and redundant questions.

The relationship between the amount of information collected and the quality of the information collected is not linear. Frequently, the amount of information climbs sharply with the first few variables, but the law of diminishing returns takes effect with less and less additional information obtained from increasing numbers of variables. Finally a saturation point is reached and from there on less information actually results from the growing number of variables. A basic underlying principle is that the patience of the interviewee and the accuracy of the enumerator declines with increasing numbers of variables, and thus the likelihood of clean analysis and interpretation of data is reduced. A decision must therefore be made at some point concerning the benefits in terms of information to be derived from further variables.

Only those variables that are relevant to the objective of the survey (e.g., selection of intervention measures, determination of intervention groups, measurement of project results) should be assessed. Therefore, not every variable that appears interesting should be included in the survey.
 

F	Data collection time should be limited to no more than three quarters of an hour spent with each family.

The timing of the baseline survey can be very important. In many rural areas there is a seasonal influence on the nutritional situation of children, because in agricultural areas the nutritional situation is often under greatest strain before the harvest season.

In urban areas a similar situation can be found with the families of employees. With incessant inflation, wages often do not increase gradually, but are only adjusted to the price of necessities after a long period. The nutritional situation is undoubtedly worse before such an adjustment in wages than afterwards.

It is very informative if a survey can be carried out at a critical time such as the period before harvest, before an expected general increase in wages, or after a period of drastic increases in food prices. Such timing will undoubtedly bias results negatively, however, it is far better to obtain information on the situation during "bad" than during "good" times, as it is not the "good" times one needs to worry about. Also, in an agricultural community, the period before harvest is a desirable time to carry out a baseline survey because family members are free from work in the fields and therefore are available to spend time participating in a survey.

As the timing of the survey depends on the progress of the project (see chapter 1.4), it is not always possible however to carry out the survey precisely during a critical phase.
 

3.1.3.2 Follow-up Survey

If a follow-up survey for evaluation is to be carried out, the relationship between the effect of the project on the nutritional status and the contributing poverty situation of the target group should be assessed.

A follow-up survey is faster than a baseline survey because most of the parameters, such as the survey area, the beneficiary group, the infrastructure, etc. have already been established in the baseline survey. It is also unnecessary to conduct any preliminary sampling as the questions used in earlier survey(s) should be used again. Ideally, the former survey staff will be available to conduct the follow-up survey and therefore can apply their earlier training. Consequently, only a short retraining period will be necessary.
 

F	It is essential that in the follow-up survey the same method of assessment is used as in the baseline survey.

This means for example, variables with the same coding as in the baseline survey must be used, and that as far as possible the survey design (such as sampling procedure) remains constant. To achieve this, the report of the baseline survey must be carefully documented and reviewed as it must serve also as a basis of the follow-up survey..

Occasionally a project or intervention area will extend into a former comparison area. In this case the follow-up survey in the former comparison area can also serve as the baseline survey for a future evaluation phase. A new comparison area should be found.
 

3.1.3.3 Groups for comparison

As described at the beginning of this sub-chapter, when possible and ethically justifiable, comparison groups (or "control groups") should be employed in nutrition baseline surveys. A comparison group differs from the survey group only in that no project intervention measures have been carried out there. The comparison area should be selected so that it lies outside the influence of the activities of the project as much as possible.

There is an ethical aspect in comparison group surveys. Because no intervention is carried out on the comparison group these individuals are left with their problems for the sake of comparison. Before arriving at a survey scheme, this point has to be discussed in depth with as many project representatives as possible and decided in all seriousness.
 

F	The interests of individuals have priority over the accuracy of survey implementation. No survey without service!

This means, for example:

• When disease is identified in an individual, medical treatment must be provided.
• When individuals, or sections of the community, seek help to solve major problems, it must be decided whether the project can contribute to the solution of the problem.

3.1.3.4 Frequency of the surveys

Determination of the frequency of surveys is a compromise between the requirements of accuracy in the M+E (monitoring and evaluation) system and the completion of the project within the deadline.

The interval between the baseline survey and the follow-up survey, as well as the frequency of the follow-up surveys, depends on:

• The type of problems surveyed in the risk groups,
• The type of target, risk and beneficiary groups,
• The type of intervention, and
• The expected impact and results.

3.1.4 Sampling

As a rule, it is impossible and unnecessary to investigate all members of the target community to determine the prevalence of malnutrition is present in the community. A representative sample is studied and the results are extrapolated to the entire community. A sample is not representative if sampling systematically favors certain groups of people or makes systematic errors. As this can happen unknowingly, unintentionally, and is only detected during the analysis after the completion of the survey or even never, random sampling is essential. This means that all eligible persons have an equal chance to be included in the survey.

It is rare that nutritional deficiencies are evenly distributed between men and women or among age groups. When one wishes to determine the frequency of a certain problem, one should also define and describe the groups to be investigated, whether by age, sex, place of residence or social status, etc. In the selection of a sample, this means that the sampling should only be done in these groups.

In large survey regions it is difficult to secure that every person in the group under investigation will be equally likely to be selected as part of the sample. To ensure this, a complete list of all inhabitants and their characteristics would be necessary. Since such a list is rarely available, a solution to this problem is the division of the area into clusters (sampling units). Clusters of individuals often arise naturally (e.g., classrooms) or they may be formed artificially (e.g., geographic clusters). The clusters are then selected at random and within each cluster investigations are carried out on every member of the community concerned. If the cluster is still too large, further random sampling can then be done.
 

3.1.4.1 Sample size

Calculation of sample size depends on the objective of the nutrition survey. If the sampling is to be done for a survey in a nutrition-related project, in which the prevalence of undernutrition is to be established in a baseline survey, where specifically aimed nutritional intervention is not to be undertaken, the necessary sample size for the survey of anthropometric data can be derived using the following formula:

n = (4 x p x (100 - p)) / 25

where: p = expected prevalence of undernutrition.

EXAMPLE:

If the prevalence of undernutrition in a survey area is estimated at 40%, the sample size (n) should be:

n = (4 x 40 x 60) / 25 = 384

In other words, the survey must be conducted on 384 individuals.

The estimated prevalence of undernutrition can be derived from literature study, data from health services, or ideally from the pilot study. The formula can only be used if:

• estimated prevalence of undernutrition is greater than 5%, and
• the sample is representative of the whole population to be surveyed. This means that each individual, as well as each household, has an equal chance to be sampled, and the selection of either individuals or households is not affected by other individuals or households.

In self-standing nutrition projects, where intervention is to be undertaken, a different method is used to determine sample size. The following information must be available to ascertain the required sample size for comparison with anthropometric data in a follow-up survey:

1. the percentage of the comparison group and/or intervention group estimated to have nutritional problems before intervention (prevalence), and
2. the percentage of the intervention group estimated to still have nutritional problems after intervention (prevalence).

The formula used to calculate the sample size is relatively complicated and cannot be understood without an extensive knowledge of statistics. Sample size can more easily be determined using graphs (see the following page), based on Fleiss' illustrated calculations (Fleiss, J.L. (1981) Statistical methods for rates and proportions. Second edition, John Wiley & Sons, New York, pp 38-42).

First one looks for the exponential curve that corresponds to the expected rate of nutritional problems for the surveyed community before intervention. Then on the lower horizontal axis the expected prevalence of malnutrition after intervention is identified and a perpendicular line is followed from this point to the exponential curve. From where these lines intersect one goes horizontally to the left or right vertical axis of the graph, and one can find the respective number of persons needed for sampling. If the prevalence of undernourished children under five is 30% before intervention and the desired goal is to halve this number, i.e., to 15%, slightly fewer than 300 children should be investigated before and after the intervention.

Figure 2. Expected prevalence after intervention
(Sample size for the comparison of proportions, a = 0.01; 1-b = 0.05!)
 

3.1.4.2 Implementation of sampling

Two methods are available for selecting the clusters to sample. The selection of the method depends on the availability of maps of the area.

1. If the map of a rural project area shows all villages, or if the city or urban map identifies districts, the sampling can take place using the grid method. A grid is placed over the map, and the resulting rectangles are numbered. The rectangles should be small enough that in only a few cases will more than one village or urban district fall within a rectangle. The specific clusters (rectangles) in which a survey is to be carried out are selected by a draw or by using a random number table (Appendix 6.5). The number of clusters (rectangles) necessary for a survey is determined by the overall sample size and the number of children estimated to live in a typical village or urban district.

EXAMPLE:

If 500 children are to be surveyed and in each village there are at least 30 families with an average of 1.5 children under 6 years (i.e., 45 children/village), then 500/45 = 11 clusters must be selected. If there is more than one village or urban district within a single rectangle, one village or urban district in that rectangle must be selected by draw. This can be done by drawing lots, throwing dice, or tossing a coin.

 

1	2	3	4	5	6	7	8	9	10
11	12	13	14	15	16	17	18	19	20
21	22	23	24	25	26	27	28	29	30
31	32	34	35	36	36	37	38	39	40
41	42	43	44	45	46	47	48	49	50
51	52	53	54	55	56	57	58	59	60
61	62	63	64	65	66	67	68	69	70
71	72	73	74	75	76	77	78	79	80
81	82	83	84	85	86	87	88	89	90

2. Another option is to construct a table listing all clusters, e.g., the urban districts or the villages in the project area. In this table, the total number of inhabitants for each cluster should be included in a column. An additional column should tabulate the cumulative number of inhabitants in the project area (see table on next page). Using literature data or pilot study data, the percentage of the target group in the total population can be estimated.

EXAMPLE:

It is known that in developing countries the under-five age group of children generally makes up 18% to 22% of the total population. Consequently, the number of children in each cluster can be estimated. In the following table there are least 77 inhabitants per cluster (Mayama, the least populated cluster) or at least 15 children under five years of age. If a total of 300 children is to be surveyed, we can calculate that 20 clusters need to be surveyed, i.e., 300/15 = 20.

EXAMPLE:

The total number of inhabitants in the survey area is divided by the selected number of clusters and the mean number of inhabitants is calculated (In the example from table 4: 99,756/20 = 4,988). Now, a number below the mean is randomly selected from a table (e.g., 507). With this random number a series of numbers is constructed by addition of the mean to this randomly selected number and subsequently to each sum (in the example: 507, 507 + 4,988 = 5,495; 5,495 + 4,988 = 10,483; 10,483 + 4,988 = 15,471, etc.). Using the second column of the cluster list (cumulative population) this series of numbers can be used to identify the villages to be surveyed, that is, all villages that against the selected cumulative number of inhabitants have the lowest difference from the numbers in the series.

Table 4. List of communities
 

Cluster or Community	Number of Inhabitants	Cumulative No. Inhabitants
Moutebe Vouka Missiele Tofi Simba N'Zororo Mungundou Sur Manbangua Boupanda Youlandzambi Biyanba N'Zima Mayama Mouyamba Bac Louesse Boungoto Dounguila Mianga Kanga Ibeni Popo ... ...	1 363  605  926  1 167  703  491  573  1 012  387  155  1 109  821  77  904  210  427  1 269  1 101  2 174  1 129  371  ...  ...	1 363  1 968  2 894  4 061  4 764  5 255  5 828  6 840  7 227  7 382  8 491  9 312  9 389  10 293  10 503  10 910  12 199  13 300  15 474  16 603        16 974 ... ...
T o t a l		99 750

In our example the clusters to be surveyed are:

1. Moutebe
2. N'Zororo
3. Mouyamba
4. Mianga
etc.

After the individual clusters, i.e., urban districts or villages, are selected, the households within the clusters need to be selected. Again, either of the two methods described above can be used for selection of households.

If an accurate city map is available, in which all residential areas are shown, the grid method can be used. The size of the rectangle should be selected so that it contains about 4-8 houses. All rectangles are numbered and the numbers of the rectangles to be sampled are selected at random. In each of these areas (rectangle) one household with at least one child will be surveyed. If individual houses are shown on the city map, the houses in the sampling area should be numbered at random and assigned numbers in sequence starting at number 1. The survey team starts the survey at house 1. If there is no family with children under 5 years, they go on to house 2. If there is still no household with children under 5, the team continues to house 3 etc. When a family with at least one child under 5 is found, the survey is completed in that area and the survey team proceeds to the next sampling area.

If there is no city map, or the map is very unreliable in showing houses, the following option can be used: a central location is chosen in the selected area (intersection, marketplace, etc.), and a coin is tossed twice, once to select north/south and once for east/west, or a number between 1 and 4 is picked from a randomized number table (see chapter 6.5) to choose north/south and east/west. From the same table a number can than be selected between 1 and 25, and the survey will commence at the house corresponding to this number. From this house, sampling will be undertaken at the next 15 houses with children under five years. If the edge of the village is reached before there is the opportunity to visit 15 houses, one starts again in the central location to continue the investigation on the next street in the clockwise direction. A third possibility arises if there is a list of households with children in a village, the listed children can then be selected by draw.

This type of sampling can be undertaken when there is a homogeneous target community. If a survey has to be undertaken with a heterogeneous community group, however, clusters representing certain population groups are the basis of selection. In this way the specific nutritional conditions of that group can be determined. In this case, however, it is not possible to make a statement on the nutritional situation of the total population.

In rural areas village communities do not always understand why only some of their community should be surveyed. It may therefore be necessary to conduct the survey including all families, and consequently correspondingly fewer villages will be surveyed, or more time will be needed to complete the survey.
 

3.1.5 Training and supervising of survey personnel

Beyond generating information on the nutritional situation, the implementation of surveys on nutritional status also has a second important task. The importance of nutritional problems as a major constraint of living conditions in developing countries should be emphasized by involving technicians and communities.

The training of technical personnel for nutrition survey implementation carries a special importance beyond that of the survey itself.
 

F	Adequate time and funds should be made available for planning and implementing training.

 

3.1.5.1. Job description of enumerators and supervisors

There are two distinct groups of technical personnel for conducting the survey:

• enumerators and
• supervisors.

Supervisors are responsible for the technical quality of the surveys for which either they or the survey teams under their charge are conducting. Furthermore, they function as a liaison between institutions with an interest in the survey and representatives of the target groups.

Knowledge of local government structures and sociocultural characteristics are prerequisites for effective coordination. Only if the supervisor has the trust of the target groups, their representatives, and local authorities, can the survey proceed as planned. Efforts should be made to draw survey personnel from the health sector, or from the agricultural or educational sectors. Thus, it is possible for the respective sectors to gain greater understanding of, and motivation for intervention measures.

If the survey teams are required to investigate families in their own homes, and the families are widely scattered over the project area, the supervisors will be able to oversee only a small number of survey teams. The assignment of survey teams to a supervisor must be done randomly. The randomized number tables can be used for this purpose (see Appendix 6.5).
 

A supervisor should be in charge of approximately four to no more than six survey teams.

The duties of a supervisor include advice and control of the accuracy of the survey data.
 

To reduce the incidence of error, the supervisor should check various measurements, such as the height and weight of the children, as well as check three to five answers on the survey questionnaire of one out of every ten or fifteen families.

If differences are evident between the data collected by the supervisor and those collected by the enumerators, then obviously the causes must be found and specific measures taken to prevent further discrepancies.
 

F	The supervisor should meet with all surveyors at least once a week during the survey implementation to discuss and exchange experiences.

It is important to maintain constant feedback with representatives of the target community about problems arising during implementation in order to prevent failure and friction.

The organizational structure of a survey is presented in figure 3. The survey leader should be chosen from a sector in charge of the nutritional situation of the population in that country. This is normally the health sector. The training workshop is carried out by the survey leader. The survey leader is responsible for hiring the supervisors, and the supervisors for hiring the survey teams. In smaller surveys, the survey leader may be able to undertake direct supervision of the survey teams, thus eliminating one organizational level. One or two additional persons are responsible for data entry into the computer. Data entry should occur during the implementation of the survey. This has the following advantages:

• time can be saved in the analysis of the survey;
• the quality of the questionnaires can be checked, and if necessary, the survey can be modified to avoid gathering unusable data;
• the completeness and accuracy of the finished forms can be rechecked, and inaccurate enumerators can be instructed accordingly.

Figure 3. Organizational chart of field personnel participating in a nutrition survey
 

Survey Leader

¯

¯

¯

Supervisor 1

Supervisor 2

Supervisor 3

¯

¯

¯

¯

¯

¯

¯

¯

¯

¯

¯

¯

S01 S02

S03 S04

S05 S06

S07 S08

S09 S10

S11 S12

S13 S14

S15 S16

S17 S18

S19 S20

S21 S22

S23 S24

Data entry personnel (S01 - S24: Enumerators)
 

3.1.5.2 Personnel recruitment

The progress of a survey depends just as much on human qualities as on the technical qualifications of the personnel taking part in the survey. Completion of elementary education is a minimum requirement for an enumerator. An enumerator must have reading and writing capabilities, as well as the mastery of basic types of calculations. Previous work experience can replace the minimum formal educational requirements.

Besides education and work experience, an accurate knowledge of the local language is required. People who speak the same language, know the local customs and are familiar with the problems of the target community will more easily win the confidence of the target groups necessary for conducting interviews. As far as possible, the enumerators should originate from the same sociogeographical and cultural area as the target groups. However, it is also possible that foreign enumerators are more accepted than local interviewers. Therefore, in each survey the possibility that distinctions such as religion, ancestry, caste or sex may influence the result of the survey has to be examined.

In all societies there are varying degrees of distinct roles for the sexes. These distinctions have to be considered in the recruitment of enumerators. As the responsibility for household nutrition generally lies with women, it is recommended that women be engaged as nutritional enumerators.

Personality attributes should be considered when selecting an enumerator. Therefore, it is important to study the personalities of the enumerators during the training program to identify potential problems. In addition, potential enumerators should be assessed in relation to the following easily recognizable attributes:

• Motivation: Only a motivated enumerator will work convincingly with the target groups and win their confidence. Motivation is influenced by the work itself and the leadership style of the surveying organization. A positive motivation can be expected if achievements by the enumerator are recognized by his colleagues and leaders, and also by the target groups.
• Communication skills For a survey to be conducted successfully, the enumerators must be able to maintain communication in the field, even under difficult conditions.
• Reliability: The results of the interview depend on the reliability of the enumerator and on the care taken in measurements and completion of the questionnaires.
• Initiative: The difficult working conditions of a survey require that an enumerator has initiative and self-reliant problem solving abilities.
• Willingness to learn: During the training program the future enumerator will become familiar with interview techniques, measurement techniques and forms of communication. He or she must be prepared to learn these subjects and then to apply them.
• Physical strength: Health and a robust physical constitution are requirements for enumerators as all will work under difficult climatic, topographical and hygienic conditions.
• Mental health: Work in poverty stricken or abandoned rural areas can be psychologically demanding. The work requires patience and a friendly nature as well as the ability to bear disappointments and frustrations.

The requirements set for enumerators apply also to supervisors, but also additional educational qualifications should be required. This could include education at a vocational school, or studies continued toward a specific goal after elementary education. Once again, successful previous work experience should be rated over formal education.

Finally, personnel engaged in a survey should also be motivated through an adequate level of remuneration.
 

3.1.5.3 Training of survey personnel
 

Enumerators must be trained intensively to carry out carefully the following tasks:

• anthropometric measurements,
• clinical and biochemical diagnosis,
• interviewing using predetermined wording,
• completion of the questionnaires, and
• communication with the community.

The training should stimulate interest through teaching aids. To achieve this, the following materials should be used:

• Charts and diagrams describing the various forms of malnutrition allow accurate and rapid transfer of information.
• Audio recordings taken during practice interviews provide the trainees with actual examples of the types of questions asked and allow detailed discussions of mistakes.

Patience is one of the most important qualities of an enumerator. As target groups are often not accustomed to being asked questions, interviewees can take a long time to answer, and the answers may be limited by the vocabulary of the interviewee and can be very ambiguous.

The communication between the "survey partners" will be enhanced if the enumerator looks at the interviewee with a friendly expression, uses encouraging words, and does not stare incessantly at the questionnaire.

Training content:

The first concern of the survey leader is to instill in the enumerators the importance of their future activities.

When the person in charge of training has made sure that the enumerators have grasped the importance of the survey for the subsequent program or project, the next step is to make the participants aware of the inherent risks of mistakes caused by:

• erroneous,
• incomplete, and
• inaccurate

To assure standardization, the individual questions on the questionnaires should be asked using a consistent predetermined wording. This can be reinforced through practice with different interview partners during training. The other enumerators and the supervisor can then make critical observations of the dialogue and discuss its strengths and weaknesses. These exercises will introduce the enumerators to a variety of interview situations and help them to gain confidence in asking questions and in completing the questionnaires. The amount of time needed during an average survey can be estimated by timing the practice interview session. However, in a field study, it is necessary to allow additional time for the enumerator to establish contact with the target person.

Next the anthropometric measurements and investigation methods relevant to nutritional physiology (e.g., measurement of blood hemoglobin levels, examination for xerophthalmia and goiter) will be practiced, and the data entered on the questionnaires. At this point, there should be a knowledgeable person supervising the measuring in order to eliminate potential errors. Measurement of the same person by several enumerators can help to check for uniformity of measuring results.
 

F	Plenty of time must be made available for surveyors to receive proper training.

A proposed schedule for the training of enumerators is shown on the next page. At the end of the training period, a pilot study must be carried out (see subchapter 3.1.6). Therefore, additional time has to be made available.

Adequate space and audiovisual materials are necessary for the training of survey teams.

Table 5. Proposed training schedule for enumerators
 

Day 1	Enumerators arrive, register and are briefed, etc.
Day 2	Morning: Introductions between the enumerators and training personnel, time to become acquainted, introduction of the training program and discussion. Lecture on the general concept of the project, the survey objectives, and the consequences for the target groups and the entire project. Introduction of the survey plan.
	Afternoon: Introduction of the questionnaire. Step-by-step presentation and memorization of the individual questions (in the predetermined wording). Interview practice in pairs with survey colleagues (e.g., through role play) followed by discussion of positive and negative types of behavior during the interviews.
Day 3	Morning: Short lecture: 1. Malnutrition and its consequences on the target group 2. Anthropometry - what for? Demonstration of anthropometric measurement techniques
	Afternoon: Practice with anthropometric measurement techniques. Determination of intra- and inter-observer errors through repeated measurement of same individual at given intervals by both same and different enumerators  Analysis of measurements with the enumerators, checking for errors, and discussion of any unclear points
Day 4	Morning: "General test" - completion of three questionnaires (including anthropometric measurements) with different survey partners .
	Afternoon: Analysis of the surveys with the enumerators, checking for errors, and discussion of any unclear points.
Day 5	Payment of daily allowances. Departure.

 

3.1.5.4 Procedure for the collection of information in a household
 

In the training session the following items should be emphasized:

• When a survey team visits a household the first time, a brief introduction should be made in which the enumerators

- introduce themselves with their names and the name of their professional institution,
- explain the purpose and the importance of the survey, and
- ask for the necessary support from the household.
• Many questions are related to values and norms and may involve very personal matters. Therefore, the presence of another person beside the enumerators and the interviewee may be distracting and responses may be altered.
• Although it is important to stay on schedule with the survey, some interviewee may have difficulties hearing and understanding the questions. For the sake of accuracy it is then necessary to slow the interview and repeat the questions.
• At the end of the survey, the enumerators should ask the interviewee about their impression of the survey and should thank them for their support.

3.1.6 Pilot testing

Shortcomings and unforeseen factors are always discovered during a survey, despite careful planning. Such unforeseeable deficiencies can be considerably reduced by conducting a pilot survey.
 

Pilot testing must be conducted before any baseline survey.

The pilot survey has the following objectives:

• Training the enumerators.
• Testing the instruments to be used (e.g., questionnaires, measuring devices) for reliability.
• Estimating the incidence of nutritional problems in order to recheck the sample size (see section 3.1.4.1).
• Fine-tuning the scheduling and organizational planning of the main survey.
• Determining intra- and inter-observer errors, since it tells whether enumerators need more training and identifies error limits (see also chapter 4.6 and annex 6.7).
• Determining the mean weight of a child's clothing under circumstances where clothes cannot be removed either because of climate or local customs.

• testing the variables for relevance (irrelevant variables must be deleted from the questionnaires, while new variables may be added to the questionnaires),
• checking the variables for cultural acceptance in the community,
• rechecking the survey type of each variable (question, observation, or measurement),
• transforming open-ended questions into closed form (creating answer categories; see sub-chapter 3.1.9),
• testing the wording of the questions (are they clearly understood by the interviewees),
• adapting the questions to the local language,
• rechecking the measuring devices, such as scales, height measuring devices, and devices for determination of blood levels, for usability and accuracy,
• defining the time required for data collection per family in order to estimate the total time required,
• discovering and correcting possible shortcomings in the enumerators,
• training the enumerator groups and the community in organizational coordination.

After the completion of the pilot survey, a final consultation is held with the enumerators. The aim of this discussion is to clarify any remaining obscure points and to reach an agreement on the final procedure and survey to be used.
 

Reproduction of the questionnaires can only take place after the final discussion on the experiences of the pilot survey.

 

3.1.7 Scheduling of a survey

The scheduling of the survey is subject to factors outside the control of the planning but which nevertheless have to be considered. In table 6, some circumstances are listed under which a survey needs more or less time.

Table 6. Factors influencing the time needed for a survey
 

Less time spent on the survey	More time spent on the survey
Nutrition-related project End of the project Follow up survey Experienced survey team Part of a socioeconomic survey Family travels to the enumerator Good transportation infrastructure Closed village settlements and urban areas Functioning group structures Time schedules not interrupted for religious or cultural reasons	Self-standing nutrition project Beginning of the project Baseline survey Inexperienced survey team Isolated survey Enumerator travels to the family Poor transportation infrastructure Open village settlements and nomadic settlements Non-functioning group structures Time schedules interrupted for religious or cultural reasons

The chronological progression of an isolated nutrition survey is presented in table 7. In this example, the schedule is based on the maximum time required for a purely nutrition project - estimated to be about 9 weeks (2 months). However, the time taken may be considerably less, depending on local circumstances. The search and study of literature, as well as further statistical analysis, which do not have to take place in the project area, are not included in this table.

Table 7. Example of scheduling of a baseline survey
 

A C T I V I T I E S	1	2	W 3	E 4	E 5	K 6	7	8	9
Review of scientific literature Review of reports from national and international organizations Preparation of the training program for the enumerators Gathering information at local institutions and from local persons  Collection of structural data  Initial calculation of sampling size  Design of questionnaires  Organization of the survey  Recruitment of enumerators  Talks with target groups and representatives of local organizations and institutions First test of the questionnaire  Training of the enumerators  Preliminary study and determination of the final version of the questionnaires (pilot study) Final determination of the sample size  Data collection in the communities  Data entry into the computer  Initial analysis of the survey  Informing the target groups and technical personnel of the survey results More detailed analysis (analysis of causes and determinants)	#   # # # #	#             # # #   #	#                               #	#                                   #   #	#                                             #	#                                             # #	#                                                 # #	#                                                     #	#

 

3.1.8 Resources needed

The following list is a checklist for quick reference to see if the necessary resources are in place for a nutrition survey. No quantities are given, as these depend on the sample size. The resources in parentheses are not essential, but they may be needed depending on the specific situation.

Personnel

• Survey leader
• (One supervisor per 4-5 survey teams)
• Enumerators (usually two enumerators per team)
• Driver
• 2 persons for data input and correction

Equipment and supplies

• For scheduling and geographical planning, extensive map materials will be needed (maps, road maps, city maps, etc.)

• Production of the questionnaires will require the following:

- Computer + printer
- Paper, printing ink or toner
for photocopying:
- photocopy machine
- toner, paper
- Mimeographing: - mimeograph machine  (!!!!!)
• Each survey team of two should be provided with the following:

- 1 weighing scale (e.g., UNIscale, UNICEF)
- 1 anthropometer (length board for length measurement of infants, AHRTAG)
- 1 microtoise measuring tape (UNICEF)
- 1 tape for mid-upper-arm circumference measurement (UNICEF)
- questionnaires
- pens and pencils
• For data processing the following will be required:

- Computer and printer
- Software: Nutrition Baseline software and word processing program
- (statistical program, e.g., SPSS for Windows, SAS)
- (graphics program, e.g., Excel, Prism,...)
- Diskettes, paper and printing ink or toner.

• Field photometer for hemoglobin determination (e.g. Compur, Bayer)
• Test tubes
• Batteries
• Autoclick (e.g. Boehringer)
• Disposable lancets (e.g. Boehringer)
• Receptacle for used disposable lancets
• Protective gloves (e.g. latex, rubber, plastic).

3.1.9 Questionnaires

Despite all desires for standardization, it is necessary for each survey to develop its own questionnaire to be completed by the enumerators with data obtained from the field.

There are three distinct types of questionnaires:

1. Questionnaire for collecting individual data
2. Questionnaire for collecting household data
3. Questionnaire for collecting structural data (data at community level - village or city suburb, district level, etc.)

Variables applying equally to individuals within a household, such as the size of the family or the amount of living space, should be recorded on a separate questionnaire for the family (household). Similarly, observable variables applying equally to all families in a village, such a climatic data, should not be recorded on a household questionnaire, but should be entered on a questionnaire prepared for a village or city suburb.

The questionnaires should be standardized to simplify their completion and subsequent use. For this purpose, the following rules should be explained on the following example of a page from a questionnaire (figure 4).

Figure 4. Example of a page from a questionnaire for a nutrition survey

1. Each page of a questionnaire should begin with a header that states

 
• the name of the responsible institution,
• the name of the survey (i.e., Nutrition survey),
• the household number,
• the type of questionnaire (i.e., Household, Child, Supervisor) with its page number.

2. A three or four digit field for the household number should be identified at the top of the questionnaire (as part of the header). Before commencing the survey a unique household number should be assigned to each questionnaire and the assigned number should be written on each page of the questionnaire. In this way the survey will be protected against accidental exchange of pages between households, and no two households will be given the same identification number (household number).

 

A household or family is defined as the group of people for whom food is prepared using the same cooking facilities, who eat together and who sleep under the same roof at least three times a week.

As only households with at least one child will be surveyed, each household number must be entered as well on the individual questionnaires before the survey begins, which will save some work during the survey. If a family has more than one child under five years, during the survey the corresponding household number must be entered on additional individual questionnaires (one questionnaire for each additional child). If more than one family lives together, but the families use separate cooking facilities, each family is assigned its own household number.

In some cultures, men may have several wives. If a husband has several wives with children, an additional variable, "family unit," must be included in the survey. Thus all family units with the same male head of family are assigned the same family unit number, but individual mothers are assigned unique household numbers.

Example:
 

Household number (Mother with cooking facilities)	Family unit (Father)
001 002 003 004 005 006 007 008 etc.	001 001 001 002 002 002 002 003 etc.

In summary, each page of a questionnaire must bear a household number to avoid accidental exchange during subsequent analysis.
 

3. Each variable shall be assigned a sequential number (variable number).

4. The order of the variables should follow a logical sequence related to the topic and to the procedure of the data collection. The survey should start with the questions and end with the measurements. At the end or the very beginning the observations may be carried out so that the interviewed person is not burdened with waiting.

5. Within the individual or household questionnaires the order of the variables to be surveyed can be determined newly in each baseline survey. Because each variable has a unique variable code (see point 18 below) there is no need to follow the order shown in the guidelines. However, a logical sequence of the questions should be maintained.

6. A variable can be recorded as a response to a question, observation, measurement, or calculation.

 

 

EXAMPLE: The age of a child can be ascertained by asking the mother during the interview or by reading a birth certificate. The data obtained will likely be more reliable in the case of reading the birth date from the birth certificate.

For standardization, the type of recording of each variable must be defined and stated after the number of variable.

Variables that result from calculations carried out during the data analysis by computer (e.g., persons per living room, z-scores, or BMI) do not appear in a questionnaire.

7. The text of questions to be asked during the survey must be stated in full on the questionnaire, so that the enumerator has no doubts about the wording during the interview. This means that the questions on the questionnaire should be stated as briefly but as precisely as possible. The questions must be stated so that the answer is not obvious from the question (leading questions must be avoided).


The local language of the target community often differs from the national language. In these cases, the questions and answers in a questionnaire should be written in the language that the people of the target community use to communicate among themselves. If the questions were initially written in the national language, the quality of the translation into the local language should be checked by reverse translation back into the national language.

8. Questions with sociological content are easier to handle with open questions which permit greater flexibility, and there is no risk of loss of information. However, using closed questions in a survey allows for faster and more objective data collection, particularly if a survey relies on several enumerators.


With closed questions, possible answers are predetermined during the pilot study, and therefore, there is the risk of loss of information. Nevertheless, using standardized questions with corresponding answer categories allows for a rapid and meaningful analysis of the entire data set, which outweighs the disadvantages of any possible loss of information. This is particularly true when several enumerators are used.

For these reasons, where the purpose of a nutrition survey is to identify problems and suggest solutions and not to carry out basic research, closed questions must be used.
 

9. As for the questions, all possible responses are listed on the questionnaire. The answers must not be read by the interviewed person.

10. The response categories for closed questions shall be coded (e.g., male = 1; female = 2). The answer categories on the questionnaire shall be completed with the corresponding code for each question. In cases of size measurements (such as height, age, weight) the appropriate unit (such as cm, months or kg) shall be stated on the questionnaire.

11. On the right side of the questionnaire a field for each variable should be identified. The number of spaces in the field should correspond to the maximum number of digits in the anticipated response - either a code or a measurement. The general organization of the fields on the right side of the questionnaire simplifies filling in the form and later reading the data for data entry. Codes of responses will be entered in these fields. For example, 3 fields should be used if body height (measured in cm) where the measurement is to be given to the nearest centimeter only.

12. For ease of subsequent reading of the data, the fields should appear in vertical order as shown above, and not next to each other.

13. For easier data analysis, all response codes should be expressed as numbers.

14. The following response code numbers should be reserved:

7 or 77 when an answer is given which has not been provided for in the coding, it should be recorded as "other"

8 or 88 when the interviewed person answers "I don't know," or in the case of an observation when it was not possible to identify a clear symptom

9 or 99 when the interviewed person gives no answer or it was not possible to carry out an observation or measurement

15. Several variables have the same responses independent where a survey will be carried out. Suggestions for responses and its corresponding code numbers are presented in the following chapter 3.2. It is recommended to use these suggested codings. This would allow that several studies can be analyzed easier cross-sectionally.

16. All response code fields must be filled in. This helps to determine whether an enumerator forgot a variable during the survey.

17. There should be no additional response coding or other forms of recording during the survey, as


- this adds considerably to the workload of data entry and the later analysis, and
- when several enumerators are engaged, it is impossible to maintain standardization.

However, it may be extremely valuable for the improvement of the survey procedure and the later data analysis if additional, qualitative observations are written down. For this purpose, all enumerators should have some additional sheets of paper where notes can be reported.

18. Each variable must have its own, unique code name (variable code) that should consist of not more than 8 digits. The code name of each core variable is mentioned later in the description of the variables (a list of all variable codes used here is presented in Appendix 6.2). Uniform variable codes are necessary for merging the data of different surveys


e.g., Variable code: GENDER

In contrast to the individual, household and structural questionnaires, the variable codes of the supervision questionnaires can be selected within each country since the reproducibility of the collected data should be calculated within each country separately.

19. Some data needed for analysis are based on further calculation of variables. For example, the body mass index is calculated from body stature and weight. As already mentioned, these variables do not appear in the questionnaire but will be analyzed later during data processing automatically by the Nutrition Baseline software. However, it is also necessary to standardize the variable codes of these variables. These variables are marked in chapter 3.2 "Contents of a Survey" of these guidelines by square brackets [ ].

Samples of questionnaires can be produced automatically by the Nutrition Baseline software.
 

3.1.10 Ethical considerations

The implementation of surveys has ethical implications that need to be considered thoroughly as early as the planning stage. Although a baseline survey is intended to establish the basis for interventions to improve conditions and has no scientific research component, ethical implications must be considered. Ethics are strongly associated with cultural norms and values that may differ from society to society; however, there are some universal issues that have to be respected.

• Ethical evaluation of survey design and procedures by local professional experts and community representatives
• Efficiency of the survey

- information is collected with the least burden on the expected respondent
- survey expenditures are justifiable considering the resources available for later activities/ interventions
- information collected does not duplicate information already available
- information is wanted by communities and organizations dedicated to improving the nutritional and general basic needs situation
- commitment that information is used later to improve the nutritional situation and living conditions of the target population
• Information to all surveyed persons

- honest explanation of the survey purpose and procedure
- description of benefits and any risks
- offer to answer all inquiries
- affirmation that persons and communities are free to withdraw consent and to discontinue participation without any form of prejudice
• Informed consent

- individual consent
- community agreement
• Privacy and confidentially of collected information

- collection of information in a way that makes it impossible or at least very hard to identify the respondent
- destruction of questionnaires that contain information that could be used to identify the surveyed household or persons (after the survey)
- instruction of all personnel working with data collected in the survey about their obligation and responsibility to respect the privacy of the respondents and the confidentiality of the data.
• Authorship and property of a survey

- all information collected and analysis carried out is available to relevant institutions of the respective country
- all main contributors to the survey are listed as authors and consensus on authorship is reached during the planning stage
- all contributors and supporters who are not authors are mentioned in the acknowledgment