3. Theoretical Perspective
5.5 The research sample and sample size
Morse and Field (1996) suggest two principles that guide qualitative sampling: appropriateness and adequacy. Participants should be able to inform the purpose of the study as determined by the research question and stage of the research, and should provide enough data to develop a full and rich description of the phenomenon (Morse and Field 1996, Spradley 1979). If these criteria are not met, qualitative results are thin and the trustworthiness of the study is threatened (Morse and Field, 1996).
In a critical analysis of purposive sampling, Patton (2001) identified sixteen types of sampling in qualitative research, and there are others not included. Patton (2001) concluded that, despite the different language used to define variations of sampling, all sampling in qualitative research is purposeful sampling, in that it is conducted with a purpose.
Kvale (1996), Hammersley and Atkinson (1995) as well as Spradley (1979) emphasise the notion of locating „a good informant‟, which will give a rich description of the phenomenon under study. In this study, sampling decisions were made at two levels. Initially, a decision had to be made about the group to be studied. A second decision was made after a group selected for the study had withdrawn. Originally, women who had suffered with postnatal depression were included and midwives and obstetricians who were working for the particular hospital.
The University Hospital where the study conducted serves the Northwest region of Greece and disposes a number of seven hundred and fifty beds. The annual rate of deliveries reaches one thousand and five hundred (National Statistical Service at Greece 2008). Midwives and obstetricians were required to have at least two years of working experience either in the antenatal clinic, or the postnatal ward or the labour ward of the particular hospital, which makes them according to Spradley (1979), „enculturated informants‟. This might enable them to give thorough insights of the particular setting in order to gain more accurate information. No other criteria applied in terms of gender, background or age. The informants had a mixture of
varied qualifications and experience, as is the case in any similar setting. The health care providers were recruited after contacting the Head of each group respectively in the particular hospital.
Women taking part in the study were women who suffered postnatal depression after their latest pregnancy. These women had to have been clinically diagnosed with postnatal depression and were fully recovered. These women had to have delivered in the particular hospital during the previous four years until approximately the end of 2003. This was to ensure that women, who had given birth recently and may currently suffer with postnatal depression, were excluded. Women who suffered with postnatal depression more than four years ago were also excluded as they might have forgotten several parts of their pregnancy and the care provided. There were no inclusion or exclusion criteria in terms of parity, age, nationality, or the type of delivery. However, women that had serious obstetric complications during pregnancy and were hospitalised for a long period of time were excluded.
The primary selection of informants was going to be made by the outpatients department of the Psychiatric Unit of the hospital. However, the gatekeepers were hesitant to reveal the names of their patients. Therefore, a secondary selection of participants was used according to Morse (1996) and Denzin and Lincoln (1998). In case a similar problem arises, they recommend advertising as a solution to the problem (Morse 1991, Denzin and Lincoln 1998). Therefore, the researcher
advertised it in the local newspaper and nursery schools to enable her to recruit participants fitting the criteria posed above.
Patton (2001) refers to the sample size as the number of participants necessary to reach data saturation. And the number of participants depends on the purpose of the study, what the researcher wants to know, what is useful and credible and the resources and time available. Lincoln and Guba (1985) recommend redundancy as a sample selection criterion.
Though, as practical problems might arise when no idea of sample size is given, Patton (2001) suggests, that a minimum sample should be recommended. Therefore, in the current study a minimum sample of thirty participants was the intended target to be recruited, with ten participants for each group respectively. The hospital employs twenty five midwives and fifteen obstetricians. Out of the twenty five midwives twenty were fitting to the inclusion criteria. From the remaining twenty, three refused to be interviewed for personal reasons. Obstetricians were all fitting the criteria, however, five refused to participate for personal reasons. Hence, ten obstetricians and seventeen midwives were recruited. This makes a heterogeneous sample which might provide in-depth information and demonstrate both a variety and commonality of experiences. Following Lincoln and Guba (1989), after saturation of the data, meaning that no new information was coming up, the researcher stopped recruiting any additional informants. The researcher, following the advertisement technique, managed to recruit five women. Given that this is an exploratory study,
the researcher wanted to ensure that she had included „information-rich cases and describe the current status of the care provided.
A while after their interview was completed, three women wanted to withdraw from the study for personal reasons. This left the researcher with two women only. This was a very small number to be considered as part of the data analysis process. after extensive discussion with her supervisors, a decision was made not to include them in the study.
The study clearly has limited generalisability considering the small sample size and the context of the study. For example, all participants were selected from the same hospital. However, the goal of a qualitative inquiry is to understand rather than generalise (Denzin and Lincoln 2000). Hence, the inclusion of all possible participants offers a fairly representative sample of the population of the area of study.