Adolescents's Knowledge,Attitude and Practice Concerning HIV/AIDS in Sierra Leone

The sampling unit of the survey is the household. The recently updated sampling frame recommended by the Central Statistics Office (CSO), used in recent surveys, was used for this survey. The Central Statistics Office has a list of Enumeration Areas (EAs) that covers the entire country. The EAs served as clusters and the households as sampling units.

The strata to be used in the sampling process are derived from the administrative structures of the country with special reference to the following administrative areas: Western Area in the West, Port Loko district in the North, Bo district in the South, and Kenema district in the East.

First stage of sampling

The list of EAs were ordered and stratified according to the following variables: location (rural/urban), District, Chiefdom, and Population size. In each of the four regions, one displaced camp was selected at random and added to the list of EAs in the area of existence. About 5% of the floating population was also targeted.

Second stage of sampling

In the second stage of sampling, the EAs were classified as urban and rural, and listed accordingly for each district. The actual number of EAs selected, for each district, within the urban/rural areas was proportional to the population sizes of these urban/rural areas.

An every nth systematic sampling was used to select one hundred and forty (140) EAs all in all. The number of urban and rural EAs selected in the sample are listed in Appendix 1

Third stage of sampling

In each of the EAs chosen in each district a locality was randomly selected and twenty (20) households in turn randomly chosen from this locality. In the few cases where the chosen locality had less than the required number of twenty (20) households, locality was then immediately replaced by a bigger one that was randomly selected from the same EA.

For the purpose of cost reduction and efficiency, the sample design just described is simpler and more convenient than the more tedious method of first dividing each EA into equal segments of twenty (20) households and thereafter, selecting at random one segment from the EA.

The choice of either sampling design is a matter of preference, but in any case, the over all conclusions remain unchanged.

One adolescent (alternating between male and female), chosen at random was interviewed in each household. In households where large extended families or groups of people live together in different units within the same compound, the recommended procedure was to randomly choose a study subject from among those found in a given unit and complete the interview for only that respondent.

The list of EAs and the corresponding localities from which the twenty (20) households were chosen are given in Appendix II.

The population of each IDP camp was ascertained from the UNOCHA database of IDP populations. The database is updated monthly, and considered to be reasonably accurate. One displaced camp was randomly selected in each district and hundred (100) individuals interviewed in the selected camp.

For the floating population, two hundred (200) individuals (50 per district) were targeted and interviewed.

The list of the chosen displaced camps is also given in Appendix II

Sample size

Decisions regarding sample size were made based on a number of factors that included;

(i) The probability of making Type 1 error that the survey was willing to accept. This probability “r” generally set at 0.05 (or 5%), is the probability that the true population value for a given indicator might fall outside of confidence limits (in this case the 95% confidence limit), that surround the estimate of the indicator;

(ii) The design effect (or deff) of the survey. This provides a correction for the loss of sampling efficiency resulting from the use of a more complex (cluster) sampling design, instead of simple random sampling. The default value of 2 is commonly used. Assuming that cluster sample sizes can be kept moderately small, the use of a standard value of deff = 2.0 should adequately compensate for the use of cluster sampling in most cases;

(iii) The average number of people per household (nh). Available data suggest that the average household size in rural areas of Sierra Leone was seven (7), whilst in urban areas it was approximately six (6). An average value of nh = 6.5 was accepted;

(iv) The estimated non-response rate (nr) was 10% and 1.1 was the factor necessary to raise the sample size by 10% for non-response;

(vi) The desired precision “d” of the estimate of key variable, set at d = 0.05;

(vii) the proportion “p” of the total population that the target population group 12 - 21 years comprise; set at p = 0.2 (or about 20%); and

(viii) the predicted or anticipated prevalence for the key indicator, set at v = 0.01 (or 1%). The formula used in estimating the sample size is given in Appendix III.