Understanding delayed access to antenatal care: a qualitative study Rosalind Haddrill

3.2.1 Sampling

Sampling methods reflect the qualitative methodology and the research question, in order to gain a full range of women’s perspectives on the subject of late booking. Sampling decisions were made early on in the research project, after the focus had been decided, reflecting the influence of prior knowledge from theoretical and professional sources. However they evolved and were modified as the study progressed, reaffirming the inductive and iterative nature of the qualitative approach.

3.2.1.1 Sample size and saturation

The sample size must reflect the qualitative approach to data collection, aiming to capture a wide variety of experiences and influences in detail, rather than obtain a representative sample or generate statistically significant findings (Hall, McKenna and Griffiths 2012). The number of participants is considered less important than the richness of the data, recognising that large samples are not necessarily better and might result in unwieldy amounts of data and a loss of depth (Murphy et al 1998, Patton 2002, Liamputtong and Ezzy 2005). As Sandelowski (1995a) points out, the process of choosing a sample size is complex: "Determining adequate sample size in qualitative research is ultimately a matter of judgement and experience" (p179). Researchers need to evaluate the quality of the information collected in light of the uses to which it will be put, and the research method, sampling and analytical strategies employed.
As such the sample size was not fixed, but an initial sample of 25 women was suggested, reflecting common sample sizes for diverse groups in qualitative research (Holloway and Wheeler 2010). Twenty five women was anticipated to be achievable in terms of practical considerations such as interviewer availability, and aimed to be large enough to answer the research question convincingly, and ultimately give credibility to the research (Green and Thorogood 2014).
Ultimately the question of sample size was left open, influenced primarily by a desire to achieve a point of ‘saturation’, interviewing until no new themes are generated from the data (Hall, McKenna and Griffiths 2012). This is an established method used in qualitative research (Bryman, 2008). However, achieving the ‘theoretical saturation’ of themes/categories identified in the data does not simply consist of stopping data collection when the stories in the data become repetitive. It requires careful sampling and a flexible and inductive approach to interviewing, to create a “conceptually dense theoretical account of the field of interest in which all categories are fully accounted for, the variations within them explained and all relationships between the categories established, tested and validated” (Green and Thorogood 2014: 122, Thornberg and Charmaz 2011). As Liamputtong and Ezzy (2005) have argued, saturation is highly subjective and context-bound, and additional data can always be found if the researcher allows the topic of the research to expand unchecked.
With this in mind, decisions about when saturation had been achieved were made through analysis meetings and supervisory discussion and feedback, and were based on the breadth of views required for the study. The broad focus of the research on late booking, in all its forms, and the resulting heterogeneous sample, required a larger sample than research with a narrow focus in order to achieve saturation (Guest, Bunce and Johnson 2006). This was achieved using a combination of sampling methods, informed by the literature but balanced with pragmatic considerations.

3.2.1.2 Sampling methods

Strategies for sample selection, like all other aspects of qualitative research, must be determined and judged in the context of the study, its purpose and rationale (Britten 2006). In this study, quantitative methods such as random or probability sampling were therefore considered inefficient and counterproductive, in terms of answering the research question. Sampling is influenced by theoretical and ethical but also pragmatic decisions, as difficulties with recruitment, for example a reluctance on the part of participants, may necessitate different approaches. Murphy et al (1998) talk about the idea of two levels of sampling decisions, initial and further, and the need to present a clear rationale and show evidence of systematic strategies for both. This was reflected in a combination of an initial pre-determined (purposive) method based on the research question and initial literature review, and adaptive (theoretical) sampling methods, to obtain a broad range of study participants (Bryman 2008).
Purposive sampling seeks to create rich, in depth information, yielding insights and in-depth understanding from a small sample size, rather than empirical generalisations. The term covers a wide variety of approaches but can be defined as a non-probability form of sampling: the researcher samples with certain research goals in mind, and thus inclusion or exclusion criteria for the study are essential. It is not a random sample or a convenience sample (simply available by chance to the researcher) but involves the selection of people with direct reference to the research questions being asked, the identification of a specific group of people who share characteristics and/or personal experiences, in this case late booking. Purposive sampling aims to select information-rich cases for studying a phenomenon in depth, to examine meanings, interpretations, processes and theory. It is a strategic approach in which participants are selected to ensure a good variety in the sample, to provide the best information and to best generate the desired data. Purposive sampling is more systematic than convenience sampling, and as such offers greater credibility (Liamputtong and Ezzy 2005, Bryman 2008, Bowling 2009, Holloway and Wheeler 2010, Green and Thorogood 2014).
Most sampling in qualitative research entails some kind of purposive sampling. Some authors have argued that all qualitative sampling is purposive; Patton (2002) presents a taxonomy of 15 different types of purposive sampling. The logic and power of purposeful sampling derive from its emphasis on in-depth understanding. What would be ‘bias’ in statistical sampling and therefore a weakness, becomes intended focus in qualitative sampling, and therefore a strength. However careful analysis is essential, as Patton (2002) comments:

“Exercising care not to overgeneralise from purposeful samples, while maximizing to the full the advantages of in-depth, purposeful sampling, will do much to alleviate concerns about small sample size” (p246).

Theoretical sampling, whereby cases are chosen to explore theoretical issues and maximise relevance (in this case the existence of particular socio-demographic groups amongst late booking women), was used alongside purposive sampling. Theoretical sampling has been called a dynamic process of gathering data to fill out tentative ideas, which encourages the researcher to sharpen their reflections, ask increasingly focused questions, seek answers and test the robustness of ideas as the research progresses (Bryman 2008, Charmaz 2012). It is an integral part of the process of constant comparison in grounded theory research and necessitates the interplay between existing literature, data collection, and the analysis and reflection on such data (Corbin and Strauss 2008, Bowling 2009, Hall, McKenna and Griffiths 2012). Such modification of the sampling procedure is seen as a rigorous way of ensuring thorough data collection: a strength in qualitative research (unlike quantitative methods), enabling adjustments and maximising the diversity and relevance of perspectives as data is collected, to achieve theoretical saturation (Holloway and Wheeler 2010, Green and Thorogood 2014).