Survey on Living Conditions Among People with Activity Limitations 2003-2004

A two-stage cluster sampling procedure was applied using the National sampling frame in each country, in close collaboration with the National statistical offices who also did sample size calculations to ensure representativity at regional/provincial level. A required number of geographical units (often called Enumeration Areas, EAs) are thus sampled, with all households in these areas included in the first stage of the sampling. Then follows screening where all households in the selected areas are interviewed (normally the head of the household) using the WG 6 screening instrument.

Sampling in Malawi:
The sample size was worked out noting that in a survey of living conditions of people with disabilities, the data user would want to know the estimates of proportions of respondents sharing respective views on issues relating to disability. The characteristics requiring respondents' views in this study are many and each characteristic would have its own proportion of respondents responding in a particular manner. In this regard, the proportion would vary from characteristic to characteristic. Determination of sample number of respondents that would give a national estimate of the proportion at a given level of precision depends on the variance of the proportion and the sample design adopted. A characteristic with a proportion having a large variance would require a larger sample to arrive at an estimate of the proportion at national level at a given acceptable level of precision than that with a smaller variance. In order to avoid having varying sample sizes for given characteristics of people with disabilities under the study, the largest possible sample number of people with disabilities based on the largest possible variance that a proportion can have at a given level of precision under given sample design was calculated. The variance of a proportion being highest when the proportion equals 50%, the required sample number of disabled persons was calculated based on the assumption that the estimated proportion would take that value with a margin of error equal to plus or minus 3.5 percent at the 95 percent level of confidence. Since the sample, as will be illustrated later, was to be drawn in stages, the design effect was assumed to be equal to 2. The design effect is the effect on the variance of adopting a sampling procedure other than Simple Random Sampling (Bradley and South, 1981).The national sample size derived was made up of 1570 respondents.

The sampling frame that was utilized in this survey was obtained from the National Statistical Office (NSO). This frame was developed by NSO through the operations of the most recent population Census in Malawi conducted in 1998. Through a mapping exercise prior to the census, a total of 9206 Enumeration Areas were demarcated covering the whole country. The boundaries of these areas followed physical features such as rivers/streams, roads/paths, galleys, etc. and these enumeration areas were demarcated in such a way that during the census an enumerator would enumerate all the persons in a given enumeration area within maximum of 21 days. Each enumeration area is estimated to have approximately 300 households or an estimated 1,000 individuals. During the operations of the census, the number of persons as well as the number of households found to exist in each one of the enumeration areas was recorded. However, no list of names and location of the households within the respective enumeration areas were made. This was due to the problems which are inherent in Malawi as well as most developing countries in giving information leading to the location of a household especially in the rural areas. Malawi has a total of 28 Districts divided into Traditional Authorities (TAs). In rural areas, the Traditional Authority is the lowest units for which maps showing boundaries of the enumeration areas are available while in the cities areas called Wards are the lowest unit for which enumeration area maps are available.

Iit was calculated that a sample of 1570 persons with disabilities would be adequate to provide estimates of acceptable precision at the national level and the terms of reference dictated that there should be complete enumeration of all the people with disabilities in the sampled enumeration areas. The lowest level for which the available frame had information, as discussed above, was the enumeration area and the information comprised of only totals of persons and households. In addition, there was no information on the prevalence of persons with disabilities at the enumeration area level.

The study conducted by SINTEF Health Research and the University of Zimbabwe using the ICF definition of disability (Eide, Nhiwatiwa, Muderezi & Loeb, 2004) estimated the proportion of those disabled to be 1.9%; while the one conducted in Namibia (Eide, van Rooy & Loeb, 2003) estimated proportion of disabled in that country to be 1.6%. Lessons learnt from Namibia and Zimbabwe indicate, therefore, that utilizing the ICF definition, the prevalence of disabled persons in Malawi may be closer to the 2.9% estimate of 1983 (NSO, 1987). In the absence of information on the prevalence of disabled persons in Malawi at enumeration area level, it was assumed that the prevalence of disabled persons in each enumeration area would be 3%. Hence, in order to be able to sample and budget for the study, it was assumed that an enumeration area would contain on average 3% of its total number of households having at least a member with a disability. Based on this assumption and considering an average of approximately 300 households per enumeration area, it was calculated that the household with at least one disabled person would on average equal to 10 in an enumeration area. Considering the coverage of 1570 disabled persons, and that an enumeration area would contain on average 10 households with at least one disabled member, a sample of 157 enumeration areas were planned to be covered in the study within which all persons identified to have a disability were to be interviewed.

Each one of the districts (Likoma Island was excluded for logistical reasons) as well as each of the three cities in Malawi formed a stratum. The total sample of 157 enumeration areas was allocated to the respective strata in proportion to the population of the stratum and the distribution thereof.
The selection of the allocated number of enumeration areas within each stratum was done with probability proportional to size prior to the commencement of the data collection exercise. The size measure was the human population of the enumeration areas as found in the 1998 population census.

Apart from enumerating all households having at least a person with a disability in a selected enumeration area (Cases) a similar number of households (designated as minimum 10 per enumeration area) without any disabled persons (Controls) should also be interviewed. The selection of the enumeration areas was done in the office. In the absence of households sampling frame within enumeration areas, the selection of Controls could only be possible after enumeration teams were deployed and sent to the field where they would conduct a household listing exercise. However, it was decided that a systematic sample of households equal to the Cases listed (assumed to be 10 on average) should be selected adopting Systematic Sampling procedures.