Cover Image
close this book Measuring drought and drought impacts in Red Sea Province
close this folder 5. Nutritional status of children in Red Sea Province, November 1985 to November 1987. Mary Cole and Roy Cole
View the document Summary
View the document Introduction
View the document Methods
View the document Results
View the document Conclusions
View the document Discussion
View the document Future directions.
View the document Appendix 5.1. Data collection form, nutritional surveillance teams, Oxfam
View the document Appendix 5.2. Claasifications of coded variables.
View the document Appendix 5.3. Ecozones in Red Sea Province (from Watson, 1976).
View the document Appendix 5.4. Seasons by month and ecozone, Red Sea Province.
View the document Appendix 5.5. Classification of fled Sea Province into food security zones, 1987.
View the document Appendix 5.6. Locations of sampled sites, nutritional surveillance cycle 1.
View the document Appendix 5.7. Names of sampled sites, nutritional surveillance cycle 1.
View the document Appendix 5.8. Locations of sampled sites, nutritional surveillance cycle 2.
View the document Appendix 5.9. Names of sampled sites, nutritional surveillance cycle 2.
View the document Appendix 5.10. Locations of sampled sites. nutritional surveillance cycle 3.
View the document Appendix 5.11. Names of sampled sites, nutritional surveillance cycle 3.
View the document Appendix 5.12 Locations of sampled sites, nutritional surveillance cycle 4.
View the document Appendix 5.13. Names of sampled sites, nutritional surveillance cycle 4.
View the document Appendix 5.14 Locations of sampled sites, nutritional surveillance cycle 5.
View the document Appendix 5.15. Names of sampled sites, nutritional surveillance cycle 5.
View the document Appendix 5.16. Locations of sampled sites, nutritional surveillance cycle 6.
View the document Appendix 5.17. Names of sampled sites, nutritional surveillance cycle 6.


1. Methods used in the original nutritional surveillance.

This is not intended as an exhaustive review of the methods used in cycles 1-6, nor as a critique. Rather, it is intended to provide a framework within which to assess the results. A description of methods used and an assessment of the practical experiences of the nutritional surveillance teams in the field is planned for early 1990 (Fatima Gebreil).

a. Sampling.

The population targeted for surveillance was children in Red Sea Province less than or equal to 115 cm in height who were receiving food aid. The unit of comparison was the administrative district. The sampling frame was drawn from World Food Programrne/Oxfam lists of settlements receiving food aid, and included the number of families estimated to be associated with that settlement for the purpose of receiving food aid. Each of the seven administrative districts in Red Sea Province was surveyed every four months. Each four month province-wide survey was called a cycle. Six cycles were completed between November 1985 and November 1987.

The method of selecting the sample size varied for cycles 1 and 2. An attempt to weight the sample by district population from the 1983 census was abandoned. From cycle 3 onwards the sample size (n) was derived from a 95% confidence interval for an estimated percentage of malnourished children (p) of 22%, and with a precision of +/- 5% for the estimate of change.

C.I. = p ± 2 {p(100-p)/n}

This resulted in a sample size of 275 for each unit to be compared (the district). Because the proportion of the population to be sampled was small, it was assumed that the sample was drawn from an infinite population, and no correction factor was applied for actual population. The same number of children were therefore surveyed in each district, except in South Tokar, where the sample size was increased to 370 because it was considered to be densely populated.

The sample was divided into clusters of 30 children, a number selected as being both the number of children who could be measured in one day, and the minimum number required to be representative of the total settlement. The required number of children would therefore be selected from approximately nine settlements (12 in South Tokar). These were randomly selected from the lists for each district, after first being weighted for the number of families in each settlement. Settlements in the South Tokar mountains were excluded from the list because of inaccessibility. The sampling procedure was altered in cycle 6 because of complaints of excessive numbers of repeat visits. The majority of those settlements sampled in cycle 5 were excluded from the lists from which the random selection of the settlements to be visited in cycle 6 were made. The selection for cycle 6 was not therefore strictly random.

b. Sampling in the field.

After arriving at each selected site, the team leader met with the local community leader (shaykh) and mapped the approximate locations of the distinct clusters which made up the total settlement. These were then numbered, and another team member not involved in the mapping asked to choose a number. This was the cluster visited. In the majority of cases there would not be 30 children in the cluster, and the selection procedure would be repeated until 30 children had been measured. In larger clusters, individual houses were selected and all the children in them measured until the required number had been reached. Another technique involved collecting all the children in a central location before selecting and measuring them. It is possible that the technique used depended on the receptiveness and cooperation of the shaykh, and that house to house visits may not have been acceptable in some areas. This will be discussed in more detail in the pending discussion of field experiences.

c. Measurement techniques.

Children were weighed to the nearest 0.1 kg using Salter hanging scales. The scales were calibrated using 1, 5 and 10 kg weights at the start of each weighing session. Stales were replaced after a maximum of two cycles, or whenever required. The scales were held by one female team member while the child was weighed by another. Very small children were suspended in a nylon sling, while older children held onto the balance. Children were not required to remove light clothing.

Height or length was measured to the nearest 0.1 cm using a locally made board with a metal tape attached, and with a sliding head board. Children who were too young to stand upright, or who were too weak, were measured lying down. There were no quarditative cut off points for a length versus height measurement; In many cases mothers objected to young or weak children being laid horizontally, a position associated with being laid out after death. These children were supported in a vertical position against the height board.

Sex, height and weight, together with comments on the child's condition (fever, marasmus, other illness, etc.) were recorded on prepared forms (Appendix 5.1). Children with oedema were marked as having kwashiorkor, and their weight for height set at less than 7096 irrespective of the actual measurements.

d. Analysis.

Raw data were brought back to the office for analysis. Percent of median weight for height or length was calculated for each child, using the NCHS/CDC/WHO normalized reference of February 1982, taken from "Refugee Community Health Care" (Ed S.Simmonds, P.Vaughan and S.W.Gunn; Oxford University Press, Oxford 1983). Mean percent weight for height, percent less than 80% weight for height, and percent less than 70% weight for height were calculated for each settlement and for each district. Hand calculators were used throughout.

e. Reporting.

Cycles one to five were written up as individual reports and circulated to the Relief Section of Oxfam Port Sudan, Oxfam Khartoum, the Sudan Desk and Health Unit Oxfam House. Raw data and district summaries were made available on request to the Relief Section in Port Sudan.

2. Methods used In the present study.

The present study is a reanalysis and a synthesis of the nutritional surveillance data collected from cycles 1 to 6. The original raw data were entered into a spreadsheet. For each case the following variables were recorded:

1. Case number.

2. Cycle number.

3. District.

4. Health status of child (good/bad).

5. Height of child.

6. If height or length of child were measured.

7. If child's age was less than one year (mothers' report).

8. Month of survey.

9. Settlement name.

10. Sex of child.

11. Weight of child.

12. Year of survey.

These variables were collected in the original surveys. The variables "less than one year by mothers' report" and "health status" were subsequently dropped because of poor quality and inconsistent reporting, particularly in the later cycles. In addition to the above variables, new variables were added:

1. Average daily ration of relief grain per family.

2. Ecozone.

3. Food security zone.

4. Season.

5. Settlement type.

The classifications used for each variable are given in Appendix 5.2. The first three additional variables were based on a mapping of all the places surveyed in cycles 1-6. Based on latitude, longitude and the major khor system with which it was associated, each settlement was placed within an ecozone. These were taken from Watson's 1976 aerial survey of livestock and human population in Red Sea Province (Appendix 5.3). A season (major wet, minor wet, dry and harvest) was assigned to each ecozone by month (Appendix 5.4). A season was then assigned to each case according to ecozone in which the settlement was located and the month of the survey. "Food security zone" refers to a classification of Red Sea Province carried out by the Research Section, Oxfam Port Sudan, in which zones of high, medium and low drought impacts and food insecurity were identified based on a composite of variables that included rainfall, floods, agriculture, livestock, grazing and economic opportunities (Appendix 5.5). Settlement type was either railway, town, rural or camp. The average ration of grain delivered per family to each settlement was calculated from Oxfam and W.F.P. distribution schedules. Actual quantities and days between deliveries were used rather than allocated quantities and proposed delivery dates. The figure calculated represented the average amount of grain per day delivered to each family for the distribution cycle during which the nutritional surveillance took place. The limitations of this calculation are discussed below.

After completing the data set as outlined above, sex specific percent weight for height/length and z-scores were calculated for each case using WHO/NCHS/CDC references taken from "Measuring change in nutritional status" (WHO 1983). These tables were entered on the spreadsheet and the calculations made using the "Lookup" function. Percent weight for height and z-scores were rounded to two and three decimal places respectively. Each case was coded as being in one of seven height categories and eight percent weight for height categories.

Height categories:

Approximate age equivalent:

<= 55.0 cm

< 3 months

55.1 - 65.0 cm

up to 6 months

65.1 - 75.0 cm

6-12 months

75.1 - 85.0 cm

1-2 years

85.1 - 95.0 cm

2-3 years

95.1 - 105.0 cm

3-4 years 3 months

105.1 - 115 cm

>4 years 3 months

Percent weight for height categories:



















1 10


1 19.9




The data set was then edited by sorting on z-scores and rechecking extreme values with the raw data. Spot checks were also made through the rest of the data. Because of the size of the data set (over 14000 cases and 250000 individual pieces of data), all character based variables such as district and settlement names were converted to numbers, and a code book developed. The data set was then converted to an American Standard Code for Information Interchange (ASCII) file for reading into a statistics package. Copies of the raw data in ASCII format and the code book are available on request.

Although z-scores are a more accurate measure of nutritional status than percent weight for height, percent weight for height is the measure presented in this report. Percent weight for height was chosen because of a general unfamiliarity with the concept of z-scores, unfamiliarity with the relation of z-score cut off points to degrees of malnutrition, and because percent weight for height has historically been the measurement of choice in reporting nutrition surveillance data in Red Sea Province.

The data set was divided into two groups; children 75 cm and under, and children 75.1115 cm. The 75 cm cut off point was selected as approximating the height of a one year old. A height cut off was selected because mothers' reporting of children as being under one year old was inconsistent, particularly in the later cycles. After examining the raw data, and the heights/lengths of children who were recorded as being under one by mothers' report in the early cycles, the WHO recommended cut off of 80 cm for one year olds seemed inappropriately high for this sample. The division of the sample into those less than and equal to or more than 75 cm was carried out for three reasons:

1. Children under one year old tend to be breast fed. Their nutritional status is therefore not as reflective of the food stresses faced by a community as the nutritional status of weaned children, and should be reported separately.

2. There are some technical problems with the data collected on children less than one year old:

a. Sampling of under one year olds may not have been consistent. The policy as to whether children under one should be included in the sample changed between cycles 1 and 2. After cycle 1, any child selected who was less than one year old by mothers' report was measured, but another child more than one year old was added to the sample to make up a quota of 30 children over one year old. The intention was to report the nutritional status of less than one year olds separately, but reports from cycles 1-5 appear to include all children under 115 cm. In the later cycles the sample was restricted to 30 children irrespective of whether they were less than or more than one.

b. A large number of under one year olds were measured in a vertical position, being supported by a team member. It has not been possible to validate this procedure.

The sampling method used in cycles 1-6 had the district as the unit of analysis. The same number of children were sampled in each district. In order to obtain representative provincewide statistics, it was necessary to weight each district statistic by the proportion of the total population of Red Sea Province in that district. These population proportions were obtained from the Oxfam/ERGO 1989 low level aerial survey of Red Sea Province. These were considered the most accurate population data avail-able at the time of the analysis. Figures for population density and inhabited area of each district were taken from the March 1989 survey. This was considered to be when the population was at its' annual peak. Population per district was obtained by multiplying population density per square kilometre by habitable area in square kilometres. Uninhabited areas of Halaib and Haya Districts were excluded from the total area of those districts. In addition, 25,000 square kilometres were excluded from South Tokar District. This was the estimated area of the mountain zone, which was not included in the sampling frame for cycles 1-6. The percentage of population per district was then used as a weighting factor for district statistics when being summed to arrive at province wide figures (Table 5.1). In applying these weighting factors to data from earlier years, the assumption was made that despite population growth, the proportions of the population in each district have not changed significantly.

Table 5.1. Population densities, inhabited area and proportion of population per district, Red Sea Province.









































* Coastal strip only (excludes mountains).

Source: ERGO, Preliminary report of phase one aerial and ground survey, February April 1989.

In addition to cycle by cycle results, annual averages were calculated in order to make comparisons of change over time without the influence of intervening variables such as season. Cycles 1-3 and cycles 4-6 were averaged to correspond approximately to 1986 and 1987 respectively.

Statistical analysis was carried out in two statistical packages; Systat and SPSS-PC+. Significance testing was limited to simple l-tests for comparing two means, ANOVA for comparing groups of means, and confidence intervals to compare proportions. The relationship between the relief ration and percent weight for height was examined using Spearman's rank correlation. A 95% confidence level was used throughout; results were considered significant only if there was less than a 5% probability they could have occurred solely by chance.