PREPARING A RAPID NUTRITIONAL
ASSESSMENT DURING
EMERGENCY
Nia Novita Wirawan Department of Nutrition
Faculty of Medicine University of Brawijaya
JENIS DARURAT
•Bencana alam terisolir (akses sulit) e.g. Banjir mentawai •Bencana alam tidak tidak terisolir e.g. Banjir jakarta •Kerawanan pangan, e.g.
•KLB gizi buruk
Warming up 2 (Group 2):
•Nutritional survey in emergency situation
•Alternative sampling desain for emergency situation •Survey Methodology
•Planning for
•training & supervision during survey
•data analysis & interpretation
•interpreting result & report finding
•Design data collection protocol
STEPS IN RAPID ASSESSMENT
DURING EMERGENCIES
•Define objective•Target population •Define geographical area •Planning a survey
•Gather information as many as possible from all possible resources
NECESSARY DATA
•the total number of displaced persons
•total affected population, and of average household size useful when planning an intervention, and when calculating needed quantities of food, water, etc. •Population figures are also needed to provide the
denominators for indicators (such as mortality rates) •Age and sex distribution of the population programmed
interventions can target specific groups, such as children under 5 or pregnant and lactating mothers
Source of Information
•Census and/or registration:
•In the case of displaced persons, it may be possible to carry out a systematic registration of persons as they arrive at the new site. This may be coupled with other aid activities, such as distribution of food cards, detection of malnutrition, measles immunization, etc.
•Exhaustive counting of habitats (or households)
•Habitats in the target area are counted one by one. The average number of persons per household is obtained from a sample of households, selected at random or through systematic sampling.
•The total population is then obtained by multiplying the total number of habitats by the average number of persons per household.
Source of Information
•Immunization coverage or programme activity data
•Suppose the immunization coverage rate age 6-59 months was 80% and that 10,000 children in this age group were immunized.
•The total children in this age group is therefore 10,000/0.80 = 12,500.
•Children in this age group ± 16 percent of the total population estimate total population at about 12,500/0.16, or about 78,000 persons
•Area sampling
•Delineate the boundaries of the target area in which people are living.
•Walk or drive along the boundary to identify key landmarks. Note their location, preferably using GPS.
•Draw a map, and calculate the total surface area.
•Draw a grid on the map, using squares of 25m x 25m or 100m x 100m, depending on the scale of the map.
•Randomly select a number of squares or GPS points, say 15.
•Count the number of people living in habitats within each square.
•Estimate the population by extrapolating the average number of persons per square, to the total number of squares counted for the full surface area.
Source of Information
•“Guesstimates”
•Key informants’ estimates, i.e. estimates by people and community leaders from the area,
•“Probability sampling means that every single individual in
the sampling frame has a known and non-zero chance of being selected into the survey sample.
•Non-probability methods of sampling such as quota or convenience sampling and random walk, may introduce bias into the survey, will throw findings into question, and are not accepted by WHO (their
emphasis).”
Source: WHO (undated)
•Define the objective
•Measurement unit vs Respondents (from whom we obtain the data)
•May be the same, may be different
•In case of the same measurement unit is HH
•Different if measurement unit is children U5, but respondents is the mother to obtain information of the child
HOW TO CALCULATE SAMPLE SIZE
•N = t2 x (p x q)/d2
N = sample size t= risk errors 1,96 atau 5%
p = expected prevalence of malnutrition as fraction of 1 q= 1-p expected non prevalence
d= level of precision
Simple random or systematic N=1,962 x (0.15 x 0.85) / (0,03)2
CAUTION
•If sample size is calculated for the number of e.g Children U5 HH to be contacted can be more
•Depends on estimate proportion of selected target grup •Number of family member/HH
OTHER APPROACH OF SAMPLING DESIGN
•30X30 DESIGN (widely used) •Alternative - LQAS:
•33X6
•67X3 DESIGN
•Sequential design
Sequential design
•a “look” at the data can be made after collection of each cluster
•If the data indicate a clear decision about the threshold level of interest (outcomes 1 or 2, above), data collection can stop. If a clear decision about the threshold level of acute malnutrition cannot be made, data collection continues.
•The sequential design thus allows for sampling to stop early
•Provided that the empirical data give a clear indication as to whether the prevalence of acute malnutrition is above or below the threshold level before data from the full 67 clusters have been collected.
•However, the drawback of the design is that it does not allow for point estimates of child- and household-level indicators unless the full sample size of 67 clusters, 3 observations per cluster (n=201) is collected
PRECISION
•When population-based surveys are used to obtain point estimates for indicators, it is important to consider the precision and accuracy of the estimates derived from the data.
•The precision of an estimate is a statistical quantification of the reproducability of the measurement. It is usually reported as a 95% confidence interval (CI) and interpreted as follows:
•Precision is different from accuracy. Whereas precision is a statistical quantification of the certainty of the
Lot quality assurance sampling (LQAS)
•The analysis approach is relevant for health programs since it is often useful to know whether a certain condition (e.g., the prevalence of a particular disease) in a given population exceeds a critical threshold level or if a program (e.g., immunization mop-up operation) has reached a certain target.
•LQAS analysis is also useful in emergency settings, where government and humanitarian agencies often need to know whether the prevalence of acute malnutrition has exceeded a certain threshold level or not. The threshold levels of 10%, 15%, and 20% acute malnutrition prevalence are often used to determine the severity of a situation
HOW TO SELECT SUBJECTS
TWO STAGE cluster sampling
•First step:
•Select the cluster or primary sampling units (PSUs) by Probability Proportionate to Size (PPS).
•With PPS, the probability of selecting a PSU for sampling is proportionate to the population size of the PSU
•A more populous PSU (cluster) therefore has a greater chance of being selected for sampling.
STAGE 1: STEPS 1-4
•Complete list of the PSUs (cluster) in the survey area along with the respective population size of each PSU. Generally, a random ordering of PSUs or ordering of the PSUs according to region is preferred (columns titled
“Village” and “Total Population”)
•Starting at the top of the list, calculate the cumulative population size and continue this process for the entire
PSU list (“Cumulative Population” and “Range” column
•Compute the Sampling Interval (SI) by dividing the total cumulative population by the total number of clusters to be sampled
•For the 33x6 design, divide the total population of Wobelleno, 123,498, by the 33 clusters: 123,498 / 33 = 3,742.36.
STAGE 1: STEPS 5-6
•A random number is then selected between 1 and 3,742.
•The PSU corresponding to where the random number falls is the first cluster selected for sampling.
•Assume the random number 1,820 is selected
•Dabi is the first of the 33 clusters selected for sampling. This is because the number 1,820 falls between the corresponding
cumulative population range of 1,409 and 2,758 in the “Range”
column.
•The procedure for selecting the remaining 32 clusters to be sampled is shown below. Notice the decimals are kept for cumulative addition, but not for cluster selection.
•Cluster 1 = 1,820.00
Cluster selected for sampling with PPS
CAUTION
•It may be the case that some of the clusters selected for sampling are later found to be inaccessible due to travel difficulties or security concerns.
•Should be avoided as much as possible as the inability to collect data from any of the original clusters selected for sampling can bias the results obtained.
STAGE 2: SELECTION OF SAMPLE
WITHIN SELECTED CLUSTER
• Random walk;•still widely adopted and accepted as an appropriate method to select observations within a cluster in emergency settings
•Compact segment sampling; •Simple random sampling
•The method is rarely used in an emergency setting, however, as it requires a complete listing (or enumeration) of all the observations (children or households) residing in the clusters selected for sampling
SIMPLE RANDOM SAMPLING
•Suppose a village consists of 100 households, and we want to interview 20 of them. We would do the following: •A listing of these 100 households, or a map showing the
location of these 100 households, would constitute a sample frame (Sample size is 20, and the sampling fraction is 1 in 5)
•To select a simple random sample (SRS), give each household a different number at @ paper
•shake the paper well, and draw out 20 pieces of paper. •Alternatively use a random number generator
•A numbered listing of all the 100 households is created, and an appropriate sampling interval (100/20=5) is worked out.
•An initial household is selected at random within the first sampling interval (let us suppose we selected the fourth household), and then the sampling interval is added to identify the remaining households: 4, 9, 14, 19, etc. •good to have the list running in a logical geographic order,
from one end of the village to the other
•Weakness: once we have made the first selection, the rest of the selections are predetermined, while those not selected had no chance of being picked.
STRATIFIED SAMPLING
STRATIFIED SAMPLING
•systematic sampling is combined with other methods of sampling
•E.g. Households close to a main road vs far from the road
Steps:
•Creating two sampling strata (one containing the houses in a predefined distance from the road and the other with the households that are further away)
•Selecting the samples separately within each stratum (2 sampling fraction: ratio sample size to total number population)
CLUSTER SAMPLING
•An approach in which each member of the population is assigned to a group (cluster)
•Clusters are randomly selected
•All members of selected clusters are included in the sample.
•Appropriate for situations in which there is no readily available sampling frame (such as a camp census list) but for which it is easy to obtain lists of subgroups or clusters of individuals, e.g. compounds or buildings or tents •Important design consideration is sample size sample •two key elements to this decision:
•how many clusters to take
•Binkin et al (2007) considers appropriate sample size for a nutrition survey in a situation of famine 30 clusters of 30 children should provide reasonably valid estimates of the prevalence of malnutrition with at least 95 percent confidence that the estimated prevalence differs from the true value by no more than 5 percent.
•The 30 x 30 approach has been used most frequently in emergencies and is known to provide reliable population estimates, however it is also time and resource intensive.
•Fanta-2, 2009 looks at alternative sampling designs which can provide reliable
•estimates on the prevalence of acute malnutrition
•33 clusters with 6 observations in each
•67 clusters with 3 observations in each
•a sequential design.
•Results: ’67 x 3' design provides estimates that are almost as precise as those
provided by the ’30 x 30' design, but requires only one-third to one-half of the field time to collect the data.
Multistage sampling:
•clusters are selected but this time sample members are selected within the cluster using simple random or systematic sampling, rather than taking the whole cluster.
COMPLEX SAMPLE DESIGN
•Sampling frame consisting of e.g. all the villages,
•Interest in selecting a sample of households across these villages.
•But rather than go to every village which would be very expensive, we might prefer to select a sample of villages, and then interview a cluster of households within the selected villages
STEPS:
•The villages might be grouped into strata according to their region
•These villages form the primary sampling units (PSUs), and can be placed in a logical geographic order, which would provide an element of implicit stratification.
•A common approach is to begin by listing the villages along with some measure of size (MOS) (e.g. number of households in them).
•Within a region, a certain number of villages are then selected systematically with probability proportional to this MOS, and within the selected villages a fixed number of households are selected.
Kabupaten
PSU Kecamatan 1 (region)
Village B, dst
Village is selected systematically with probability proportional to this MOS, e.g: Village A with 150 HH, with total population in Kabupaten 2500 HH and calculated minimal sample size is 1000. PPS is
150/2500 x 1000 = X selected HH number in village A
SAMPLING DURING IMMEDIATE
RESPONSE PHASE
•May not be possible to carry out a strict probability sample survey: access/mobility issues, time/resource factors and/or because the absence of good population data to create a suitable sample frame
•non-probabilistic sampling
SAMPLING DURING IMMEDIATE
RESPONSE PHASE
CONSIDERATION:
•Coverage: cover as wide a cross-section of the relevant population and geographical area as possible. Do not only include easy-to-reach elements.
•Sample frame: Establish clearly the sample frame is (list of people, villages or a map)
•Sampling methods: Use probability sampling if at all possible and justify if use non-probability sampling •Sample size: Use as large a sample as fit in with
resources. Better to visit more locations and interview less people in each, than vice versa.
•Secondary data: Make full use of secondary data
JUDGEMENT SAMPLING
•“experts” select the sample •extension of convenience sampling
•E.g: the expert may decide to draw the entire sample from one
•“representative” village, even though the target population
includes many villages.
•When using this method, the researcher must be confident that the chosen sample is truly representative of the entire population (extremely familiar with all villages to have this confidence)
•In reality, it is quite unlikely that the selected villages or households would be representative of all villages or households.
Purposive sampling
•Best choice of sampling in the immediate aftermath of an emergency when it is not possible to apply probability sampling.
•Within this approach, selection of the sample is done according to specified criteria to represent certain cases, e.g. the extremes or the norm
•stratify possible localities according to socio-economic or demographic criteria and visit diverse areas e.g. urban and rural areas, and with both residents and non-residents (displaced persons), higher/lower prevalence of chronic malnutrition, different ethnic groups, etc.
The criteria for site selection (
IASC, 2009
)
•Urgent need: First priority will be to assess areas in greatest need (Consider vulnerability suc as population size, density and influx, availability of water and food, reported epidemics or malnutrition.
•Accessibility: Where overall needs are urgent, widespread and unmet, it is justifiable to focus on accessible areas. However, where inaccessibility is a widespread problem or coincides with very urgent needs, the extreme rapid assessment – a two-hour visit – may be necessary to fill information gaps.
•Gaps in existing knowledge: Cover locations about which little is known or where key information is lacking, especially where no relief agencies are yet working.
•Worst-/best-case scenarios are often used to provide some reference for interpreting data.
Notes: sites selected are those most urgently in need of assistance
RANDOM WALK METHOD
•An approach which is often used in post-emergency surveys when complete data on the affected populations is still not available
•Ensuring that information is collected from households with different proximity to the village centre, roads, stream
RANDOM WALK METHOD
•Begin the sampling at some randomly defined
geographical point, and then follow a specified systematic path of travel in order to select the households to be interviewed.
•This might entail selecting every nth household, or else screening each household along the path of travel to locate the presence of the special target population such as children under 5.
•In the latter case, each qualifying household is interviewed, until the quota is reached.
References: (FANTA-2 Project, 2009) and (ICRC/IFRC, 2008).
RANDOM WALK METHOD
•Greet Community Leader and Seek Permission to Conduct Survey
•Explain the Random Selection Process
•It is recommended that the interview team requests to be accompanied by the community leader or another respected member of the community during data collection at the cluster site
•The team should describe to the community leader/member the importance that a random procedure be used to select the households to be sampled.
RANDOM WALK METHOD
•Spin the Pen
•Note of the direction the ball of the pen is pointing
•This is the direction that has been randomly selected for the interview team to walk in order to identify the first random household to be sampled in the cluster.
RANDOM WALK METHOD
•Map and Enumerate Households in Randomly Selected Direction
•walks from the center of the cluster site to the perimeter of the cluster site, in the direction indicated by the ball of the pen
•Households that lie approximately along the line extending from the center of the cluster site to the perimeter of the cluster site in the direction of the ball of the pen are mapped and enumerated.
•Even if walking in the indicated direction is difficult, this is the direction along which the households in the cluster need to be mapped and enumerated.
RANDOM WALK METHOD
•Select a Random Number to Identify the First Random Household to Sample in the Cluster
•The random number selected should fall between 1 and the total number of households enumerated during the walk to the perimeter of the village.
•If the random number selected is greater than the total number of households enumerated a new random number should be selected.
•Select Subsequent Households to Sample in the Cluster
BEYOND IMMEDIATE RESPONSE
•Probability sampling, e.g. Cluster sampling•Two important outcomes.
•Possible to derive estimates from the survey, and to say that the sample is representative of the target population.
•Possible to calculate sampling errors, and thus get a good idea of the precision of the survey estimates.
E.g. Survey in a rural area, select 30 villages at random (preferably by sampling with probability proportional to size) from a list of all the villages in the affected area, and then pick a sample of households in those selected villages (e.g. By random walk).
If the region to be surveyed is very large or heterogeneous split into strata and 30 clusters selected from each stratum
e.g. RAPID ASSESSMENT SURVEY FOR EXPANDED PROGRAMME ON IMMUNIZATION (EPI)
•30 Cluster with probability proportional to the most recent census estimate of size, 7 children aged 12-23 months in each cluster. Total 210 children
•Selecting the children in each selected cluster by random walk
•Starts at a central point,
•Selects a random direction from that point (‘spinning the pen’),
•Choose a dwelling at random among those along the line from the centre to the edge of the community.
•All children in the household in the age range 12-23 months are selected and the mother or caregiver interviewed (In multi-household dwellings, all multi-households are visited.)
•Starting from this household, the next nearest household is visited in turn until at least seven children have been found.
ASSESSMENT OF FOOD SECURITY
DURING EMERGENCY
•Objectives, neccesary information and possible indicators for integrated rapid food and nutrition security
assessments (link)
DIFFERENCES HDDS
–
IDDS
HDDS IDDS
Respondent Orang yang bertanggung Jawab pada persiapan makanan untuk keluarga KEMAREN
Prepared in the home and consumed in the home or outside the home; or Purchased or gathered outside and consumed in the home
NOT INCLUDES:
Purchased outside the home and consumed
Outside**
All foods eaten by the
1 There is some evidence that women’s dietary diversity also reflects household
economic access to food.
** Those foods are not included because the respondent may not know which other household members purchase
Dietary Diversity Questionnaire
Form 1
Sarapa n
Snack Makan Siang Snack Makan malam Snack
Sebutkan semua makanan dan minuman (makan dan selingan) yang dimakan/diminum paling tidak oleh SALAH SATU anggota keluarga anda KEMAREN mulai pagi hingga malam hari baik di rumah maupun di luar rumah .
Jika yagn dikonsumsi berupa “mixed dishes”, tanyakan apa bahan-bahannya.
Jika responden sudah selsai menyebutkan, tanyakan makanan/snack yang kemungkinan belum disebutkan
[Households: include foods eaten by any member of the household, and exclude
foods purchased and eaten outside the home] Jika Sudah selsai isilah formulir 2 berdasarkan formulir 1.
Jika ada kelompok makanan (food group) yang belum disebutkan, tanyakan kembali kepada respodnen.
Number Food Group Examples No=0
1 Cereals
Corn/maize, rice, wheat or any other grains/foods made from these
(bread, noodles, mihun, bihun, porridge
2
White Roots and tubers
White potatos, white cassava, 'mbothe, talas, or other foods made from
roots
3
Vitamin A rich
vegetables and tubers Pumpkin, carrot, squash, sweet potato, 'benthoel', 'red talas',
4
Dark Green Leafy Vegetables
Cassava leaves, spinach, 'daun singkong', daun kacang, kangkung, sawi,
papaya leaves, etc
5 Other Vegetables Tomato, onion, eggplant,
6 Vitamin A rich fruit Ripe manggo, ripe papaya, and juice made from these
7 Other fruit
Star fruit, banana, orange, rambutan, avocado, and juice made from
these
8 Organ meat Liver, kidney, heart or other organ meats or blood-based food
9 Flesh Meat Beef, pork, lamb, goat, rabbit, chichen, duck, other birds, insects
10 Eggs Eggs from chicken, duck, or any other eggs
11 Fish and seafood Fresh or other dried fish or shellfish
12
Legumes, nuts and seeds
Dried beans, lentils, nuts, seeds or food made from these (sambal pecel,
tempe, tahu, etc)
13 Milk and Milk
products Milk, cheese, yogurt or other milk products
14 Oils and fats Oil, fats, butter added to food or used for cooking
15 Sweets
Sugar, honey, sweetened juice drinks, suggary foods such as chocolates,c
andies, cookies and cakes
16
Spice, condiments, beverages
Spices (black pepper, salt), condiments (soy sauce, tomato sauce,
ketchup), cofee, tea, alcoholic bvereages
Aggregation of food groups from the
questionnaire to create HDDS
1 The vegetable food group is a combination of vitamin A rich vegetables and tubers, dark green leafy vegetables and other vegetables.
2 The fruit group is a combination of vitamin A rich fruits and other fruits. 3 The meat group is a combination of organ meat and flesh meat.
Question
Number Food Group Examples
Yes=1, No=0
1 Cereals
Corn/maize, rice, wheat or any other grains/foods made from these
(bread, noodles, mihun, bihun, porridge
2 White tubers Pumpkin, carrot, squash, sweet potato, 'benthoel', 'red talas',
3
Vitamin A rich vegetables and tubers
White potatos, white cassava, 'mbothe, talas, or other foods made from roots
4
Dark green leafy vegetables
Cassava leaves, spinach, 'daun singkong', daun kacang, kangkung,
sawi, papaya leaves, etc
5 Other vegetables Tomato, onion, eggplant,
6 Vitamin A rich fruits Ripe manggo, ripe papaya, and juice made from these
7 Other fruits
Star fruit, banana, orange, rambutan, avocado, jambu biji, and
juice made from these
8 Organ meat (iron rich) Liver, kidney, heart or other organ meats or blood-based food 9 Flesh meat Beef, pork, lamb, goat, rabbit, chichen, duck, other birds, insects 10 Eggs Eggs from chicken, duck, or any other eggs 11 Fish Fresh or other dried fish or shellfish
12
Legumes, nuts and seeds
Dried beans, lentils, nuts, seeds or food made from these (sambal
Aggregation of food groups from the
questionnaire to create WDDS
1 The starchy staples food group is a combination of Cereals and White roots and tubers. 2 The other vitamin A rich fruit and vegetable group is a combination of vitamin A rich vegetables and tubers and vitamin A rich fruit.
3 The other fruit and vegetable group is a combination of other fruit and other vegetables. 4 The meat group is a combination of meat and fish.
Specific food groups for IDD for children
•8 Food Groups in IDD for children, depending on the
importance of certain foods in children’s diets
•Score: 0-8
1. Grains, roots or tubers
2. Vitamin A-rich plant foods
3. Other fruits or vegetables
4. Meat, poultry, fish, seafood
5. Eggs
6. Pulses/legumes/nuts
7. Milk and milk products
8. Foods cooked in oil/fat
HDDS Calculation
•Step 1: Assign 1 if the food group/item consumed; 0 not consumed. Sum all the scores for various food groups. Sum will be between 0-12.
•Step 2: The average IDDS for the sample population
Sum HDDS
Total no. of households
‘Cut off points’ in DDS?
•Cut off points depend on local circumstances (prevailing food habits)
sampling approaches can be adopted in integrated rapid food and nutrition security assessments
•When anthropometric measurements are taken, sampling is an important issue for the following reasons
•minimum sample size is needed for drawing conclusions about the
population’s nutrition status that can be extrapolated to the larger population with a reasonable level of confidence.
•E.g. using a 30 x 30 cluster survey and a design effect of two needs 900 children under 5, confidence level of 95%, precision of 5%
•Large samples are usually needed to obtain precise estimates of malnutrition rates based on anthropometric measurements, while assessment of food availability and access, and information on the food security situation require smaller household samples.
Source: WFP. 2009
REFERENCES
•UNICEF. 2010. Rapid Assessent Sampling in Emergency Situation
•WFP. 2009. Emergency Food Security Assessments (EFSAs) Technical guidance sheet. Strengthening rapid food and nutrition security assessment
•Robert Magnani. Sampling Guide. FANTA
•UNICEF (2006), Multiple Indicator Cluster Survey Manual. http://www.childinfo.org/mics3_manual.html
