(2004; 98 pages)
5.3 Purposeful sampling for qualitative studies
There are several strategies for purposeful sampling of information-rich cases. The methods most commonly used in qualitative studies are given here, including the purpose for which the method is especially useful and its disadvantages.
Convenience sampling is a method in which, for convenience sake, the study units that happen to be available at the time of data collection are selected in the sample. Many health facility or drug-outlet-based studies use convenience samples. If you wanted to study information provision on medicines in pharmacies, you could observe all client-drug-seller interactions during one particular day. This is more convenient than taking a random sample of people in the village and it gives a useful first impression. A drawback of convenience sampling is that the sample may be quite biased. Some people may be overselected, others underselected or missed altogether. In this example, the interactions observed may be biased because the pharmacist does not work on the day observed. You also miss the clients who obtain their medicines from other sources. Informal drug outlets in communities are often as important as pharmacies as sources of medicines. It is necessary to study interactions at those outlets as well, to get a good impression of the provision of information on drugs.
Maximum variation sampling
This sampling method aims to select study units which represent a wide range of variation in dimensions of interest. For example, the researcher may be interested in the reasons that people do not comply with antibiotic prescriptions, and assume that gender and socio-economic status are important background variables. The researcher is afraid to miss men, who are often not at home when researchers visit to conduct semi-structured interviews. Therefore, the researchers decide to conduct interviews during the day and in the evenings, and to ensure that at least 15 men and women are included in the sample.
Maximum variation can also be used as a strategy to select communities in which to do research. In the example, this would imply that the researcher selects one relatively rich and one poor community. Maximum variation sampling is also often used when deciding on which groups to involve in focus group discussions. Remember, the informants participating in each FGD should be relatively homogenous as far as key background variables for the study are concerned.
Snowball sampling is perhaps the most common sampling method used in qualitative studies. The researcher starts by identifying some (at least two) individuals who are relevant to the study, for example, women with pre-school children in a study on home-treatments in malaria, and then asking them to locate other useful informants, i.e. other mothers of pre-school children. The advantage of this method is that one informant refers the researcher to another, so that the researcher has a good introduction for the next interview. A disadvantage is that the variation in the sample may be limited because it consists of informants who belong to the networks of the index cases. This is why it is important to have at least two different additional entrances in the community.
Sampling contrasting cases
Comparative studies sampling will involve two or more population groups with distinct characteristics. This sampling method is useful in comparative studies that aim to explain problems by establishing which factors are associated with them or cause them. For example, in a study aimed at understanding why mothers do not use oral rehydration therapy (ORT) to prevent childhood death in diarrhoea cases, both women who use ORT and those who do not can be sampled and compared.
Contrast sampling can also be used in selecting research sites. For example, when evaluating a health programme, a research site can be selected where (according to statistical information) the programme has been successful (for example, in promoting ORT) and where this is not the case. Comparison can help in analysing which factors contribute to success and which factors constrain programme success. Contrast sampling can also be used to select participants for focus group discussions. Within each group the informants should be relatively homogenous in terms of the important dimensions of the study; but for the different groups you select contrasting cases (for example, men and women; younger and older; users and non users).
Qualitative sampling respondents for semi-structured interviews and FGDs
The qualitative methods presented in Chapter 3 for investigation on drug use include semi-structured interviews and FGDs. How can we sample respondents for these methods?
Sampling for semi-structured interviews
First you need to define whom you want to interview. If you are aiming to get an overview of drug use problems, it is best to select a wide range of individuals. If you are analysing a specific drug use problem, you concentrate on people who have direct experience with the drug use practice that is problematic, and people who are knowledgeable about it. Snowball sampling is the most common sampling method used in selecting respondents for semi-structured interviews. You can also decide to conduct contrast sampling. You can get an idea about which groups to select by reviewing your problem analysis diagram. Which socio-cultural factors seem to be related to the problem? Can we test these assumptions by comparing ideas and practices in different groups? It is also useful to contrast groups that use drugs appropriately with those who do not. Information on whether or not drugs are used appropriately can be obtained from the focused illness recalls. By conducting semi-structured interviews with both groups and comparing findings you can get an idea of the reasons for appropriate and inappropriate practices.
Sampling for focus group discussions
The main decision you need to take when planning focus group discussions is what focus you intend to have, and how many FGDs you intend to hold. FGDs are often used to contrast views of different ‘focused’ groups: for example, adolescents versus adults; or men versus women. Decide which population sub-groups need to be interviewed. Limit the scope of the study to those sub-groups which have direct experience with the problem. Usually local leaders are asked to select respondents for the focus groups. Aim for around 6-8 participants per group; and conduct at least two FGDs per population group involved. So, for example, two with men and two with women, or two with adults and two with adolescents. If the conclusions of the two groups are not in agreement you may need to hold a third FGD to further investigate the issues.