Hi all. Doubtless this has been asked before but I wanted some thoughts on my particular situation. I'm conducting research into stroke care. The bulk of my data comes in the form of semi-structured interviews with clinicians (but also from a few nurses/allied health professionals). The clinicians are all consultant level (some combine their clinical activities with academic ones, so are Profs. as well as medical doctors) i.e. leaders in their field.

The telephone interviews last anywhere from 25-45 minutes. Given the seniority of the interviewees, what number would be sufficient for a PhD thesis? I will be transcribing the interviews and then manually coding them/looking for themes etc.

Thank you!

Would 20 be enough?

======= Date Modified 20 Jul 2012 05:23:58 =======
Hi Whoknows,

You mention the seniority of the people you hope to interview as part of this study. By that comment, are you implying that the population size is small? 

I also conducted semi-structured interviews which lasted between 7 mins and 1.5 hours. Both of my supervisors (one comes from a qualitative stance and the other has a quantitative background) were keen for me to conduct more than the initial 30 interviews proposed. In the end I completed 43 interviews with 3 different population groups. Their reasoning behind carrying out more interviews than the initial 30 was in case someone coming from a quantitative background was present during my viva (as you probably well know, statisticians are partial to large sample sizes!). But of course issues such as the estimated population size, issues around access, etc., have to be considered as part of the methods design.

thanks for the reply RLD. Yes, I was hinting that the sample size is small when one considers the seniority of the interviewees. Both my supervisors are qual. by trade so I've had little "statistical input" - from your perspective what percentage of a given sample ought to be interviewed to ensure "validity" etc.? I'm struggling to get more than around 20-25 due to the access isssue you mentioned - will this be problematic in your view? Also, you say that you used three separate sample populations - would you suggest the number of interviewees should be (roughly) equally distributed among these?

Hi, I would say a lot depends on your methodology. With some approaches, like Interpretative Phenomenological Analysis, you only need a few interviewees (less than 10 in the case of IPA). Others keep doing interviews until they reach saturation. It also depends on how many participants you manage to recruit! I'm really struggling with this, I'll be very happy if I end up with 10 - but then I am using IPA, so it should be fine. My research is also longitudinal, so I'm doing a number of interviews per participant.

======= Date Modified 21 Jul 2012 05:36:49 =======
Unless for exceptional methodological reasons you should continue to collect and annalise more data until you reach saturation. Quite when this is is subjective, however broadly if you have a tightly described purpositive sample and are looking to answer a narrow question then a smaller sample may be adequate with some studies having as few as five interviews. Others have many many more. There is evidence that many PhD theses studies use inappropriate criteria to set sample sizes. More including disproportionately large samples.

The seniority of your interviewees does not change the validity or otherwise of the data from their interviews. Nor does the time for transcription or coding - Continue until saturation...

I have a bunch of references on this please PM me for more details.

Very useful article here

http://www.qualitative-research.net/index.php/fqs/article/view/1428/3027

Whoknows it feels like there are several issues here. My background is qualitative recearch and i have spent the last few years looking into qualitative approaches and I think most reserchers would say the lack of statistical input is irrelevant in an interview based qual study - we are looking to robustly describe the points that are important in our sample group as a whole rather than 'prove' that this applies to everyone else who would meet the studies entry requirements. A follow on quant study based upon the themes/codes from the qual results would be needed for this 'proof'.
The sample population selected is usually defined so the study findings may be considered representative of those who would be included under our research question.
If your population is genuinely very small with your question defining them tightly as in 'What are the experiences of specialist consultant grade doctors in providing care to stroke patients at St Gregories Hospital' the same principles apply, however it is acceptable to have more limited study subjects for obvious practical reasons. You will want to acknowledge this in your discussion and reflect on its potential impact on your conclusions.
If you have multiple ’subgroups' whose results you want to consider seppertely, then each idealy needs to be large enough to reach saturation. I am interviewing two groups of patients defined by how reassured they feel after a clinical encounter and will start with analysing five transcripts from each group. Further will be added until no significant new themes emerge. I am hoping, like you that as these are patients with tightly defined characteristics that I will not have to do too many however I will not know until I get there.

Thank you for all your responses - the feedback has been insightful.

======= Date Modified 23 Jul 2012 10:06:37 =======
============= Edited by a Moderator =============
*Removed by PostGrad Forum team - not relevant*

Hi Whoknows,

I found an article online that may of interest to you regarding qualitative sample sizes and practices/concerns surrounding saturation:

http://www.qualitative-research.net/index.php/fqs/article/view/1428/3027

Cheers,
RLD

Thanks RLD - I think that is the same one that Clementine posted, right? It's a good one though, so perhaps worthy of a second posting :) cheers!

Having read a few articles about this topic, I'd say around 20 interviews is the minimum, unless you are working with a particular methodological/analytical approach which you have good evidence requires less. It also depends what other data you're collecting (e.g. are you collecting quantitative data as well, or doing statistical analysis on other people's data-sets), as this will give you other data to expand your findings chapters. A good argument for collecting the number of interviews you collect is saturation (i.e. you are finding no new new major issues/topics/sub-topics during each subsequent interview you carry out). Obviously the quality of the data collected as well will suggest whether you have enough to write up a decent PhD, so 20 good interviews which allow you to answer your research questions well, will be more convincing to examiners than 20 interviews that don't seem to give you enough data to sufficiently answer your question.

Thank you for your thoughts. Yes a post doc I spoke to also mentioned 20 as a min. I think the quality issue you raise is an important one too :)

Supervisors thinks I need 40 to 50 though! Don't know where I'm going to get that many from! Seems like overkill to me ...