NPT can help inform, guide or structure the way you code and analysis your data.  You could think of it in a similar way to one of Glaser's (1, 2, 3) coding families, or Strauss and Corbin's (4) single coding approach, in that all the data can be coded against NPTs four core constructs and the specific components.

Introduction

Coding and analysis is a craft. It involves a lot of time, energy, patience and above all perseverance. Qualitative researchers often use theory to inform coding and analysis. You can use NPT in various ways to help you make sense of your data.  You have three main options open to you:

1. Code your data directly via the headings of the various NPT constructs and components.
2. Code your data as you would normally, say via some version of the constant comparison method, or some variant of thematic analysis.  Then group your codes under the various NPT constructs and components.
3. Code your data as you would normally, say via some version of the constant comparison method, or some variant of thematic analysis.  Then re-code the data via the NPT constructs and components.  You can then compare your findings.

In relation to coding your data directly via the headings of the various NPT core constructs and specific components, let's look at a very brief example, taken from an interview with a health practitioner discussing the referral process between hospital departments.  Here's a section of un-coded transcript (See box A):

A. Un-coded Original

Health Practitioner:  a lot of (pause) kind of departments are somewhat obstructive when you are trying to pass patients on to them and if you know someone it's a lot easier … a lot of other departments would want you to do every test under the sun, confirm a diagnosis and then they'll see them.

We could then code this passage drawing on the four constructs of NPT (See box B).

B. Coding via NPT Constructs

Health Practitioner:  a lot of (pause) kind of departments are somewhat obstructive when you are trying to pass patients on to them and if you know someone it's a lot easier … a lot of other departments would want you to do every test under the sun, confirm a diagnosis and then they'll see them.

Collective action

Clearly, in describing the referral process, this practitioner is describing the practical work involved in making a referral, she is describing aspects of the work that 'defines and operationalizes' such referral work.  In this case, the code collective action seems appropriate. However, if you just work with the constructs, you are going to get quite a lot of data that's just fits into the four areas.  You end up having little detail in which to begin to think analytically.

Another option is to drop down to coding at the level of components (See box C).

C. Coding via NPT Components

Health Practitioner:  a lot of (pause) kind of departments are somewhat obstructive when you are trying to pass patients on to them and if you know someone it's a lot easier … a lot of other departments would want you to do every test under the sun, confirm a diagnosis and then they'll see them.

regional integration skill-set workability

In this case two immediately come to mind - skill-set workability - that a practitioner in order to enable a referral must be competent in providing adequate and accurate diagnostic reasoning and evidence offering just offering a 'hunch', alongside relational integration - in that clearly getting a referral requires some form of trust or ongoing relationship between the referring practitioner and the department.  Getting a diagnosis wrong and so wasting another time has implications for that practitioner's reputation and standing.  All this has implication for patient progression to suitable specialist care.

So, coding this way, via various NPT constructs and components, can enable you to begin to make sense of your data.  However, be aware that you can end up with a lot of large chunks of data under NPT headings like 'relational integration'.  You will then have to critically analysis and interpret the content and significance of data coded under a given heading (like relational integration) in relation to your aims, objectives, knowledge of the literature and so on.  Put simple, you will need to continue the analysis process by going through these large chunks and then draw out the various research-project specific issues at hand.

In relation to the other two options, let's look at an example of initially coding the data via some form of more traditional coding technique, say thematic analysis (5) or the first few processes of the constant comparison method (6). Initially, you might begin with some form of line-by-line, fine grained or in vivo coding (See Box D).

D. Initial in vivo coding

Such a style of coding is often found when you initiate a project, where you are trying to gain as much detail as possible.  You may start with such codes, they clearly report what is going on in the referral process.   After you initial rounds of coding, you might have refined your codes somewhat, so that you have more descriptive terms that reflect the issues you found over the data set. For example, in as shown in Box E (below), 'want you to do every test under the sun' is now part of a broader descriptive code, that off 'diagnostic tests'.

Health Practitioner:  a lot of (pause) kind of departments are somewhat obstructive when you are trying to pass patients on to them and if you know someone it's a lot easier … a lot of other departments would want you to do every test under the sun, confirm a diagnosis and then they'll see them.

Obstructive when try and pass on Knowing someone Every test under the sun

E. Coding via Descriptive/Focused Coding

Once you developed these more generic, descriptive codes, these can now be placed in the coding manual or coding sheets, or outlined in a memo (7), and then related to the relevant NPT core construct and components. So for example, your code book (if you use one) could look something like this  (See box F):

Health Practitioner:  a lot of (pause) kind of departments are somewhat obstructive when you are trying to pass patients on to them and if you know someone it's a lot easier … a lot of other departments would want you to do every test under the sun, confirm a diagnosis and then they'll see them.

Past contacts Diagnostic tests

F. Coding Sheet 1

NPT	Concept & Relevant Codes	Examples
	Concept 5: Referral Work This concept covers the ways that referral are undertaken. It outlines the local routines and practices around referral … etc. Code 5.1: Contacting departments Code 5.2: 'Difficult' patients/family …
Collective action • integration relational	Code 5.7: Past contacts Descriptions of the role of how prior relationship with other practitioners, be they negative or positive, be they from encounters as students or colleagues, face-to-face or by reputation alone, impact upon the referral process … etc.	'a lot of (pause) kind of departments are somewhat obstructive when you are trying to pass patients on to them and if you know someone it's a lot easier' (HP10: p7)
Collective action • workability skill-set	Code 5.8: Diagnostic tests Descriptions of the types and range of tests that practitioners are required or expected to conduct. Includes the type of test that practitioners think they should conduct as well those they are told to conduct when contacting prospective department to refer to … etc.	a lot of other departments would want you to do every test under the sun, confirm a diagnosis and then they'll see them.' (HP10: p7)

Or your code book (if you are using one) could look something like this (See box G):

G. Coding Sheet 2

Concept	Collective Action
	Interactional Workability	Skill-set Workability	Relational Integration	Contextual Integration
Referral Work This concept covers the ways that referral are undertaken. It outlines the local routines and practices around referral… etc.	'Difficult' patients/family …	Diagnostic tests The types and range of tests that practitioners are required or expected to conduct impacts on the referral pathway. This includes the type of test that practitioners think they should conduct as well those they are told to conduct when contacting prospective department to refer to … etc.	Past contacts Prior relationship with other practitioners, be they negative or positive, be they from encounters as students or colleagues, face-to-face or by reputation alone, impact upon the referral process … etc.	Contacting departments …

The idea here is that you've started with your data - coded what is relevant to your emerging data and your interests (boxes d and e) - and then related those findings to NPT (boxes f and g).  In this way, especially with a large amount of data, you have a good sense of the range of issues you find under the broader NPT labels.

So why might such an approach be useful?  If you wanted to transform a specific aspect of referral for a specific disease, to introduce a new referral pathway, it is not enough to say that we need to address issues of 'relational integration'.  Instead you need to offer a much more practical, research site-specific answer.  In the example above, this practitioner informs us that there are specific referral work practices, almost, a referral 'etiquette', 'ritual' or 'order', so if we want to instigate a new referral procedure and we want it to normalize it, we would need to be aware of this and work with it.

However you choose to code your data, do not always expect your data to automatically fit nicely within the domains in the theory.

• When you first use it, it will take you some time to familiarise yourself with NPTs ideas, constructs and components.
• You may often spend time debating about which specific area of the theory that a specific section of data or code really belongs to.  That's not a problem and quite routine.  Some incident, situations or issues can fit within multiple area of the theory.  As with all qualitative analysis, as long as your colleagues agree and you can defend your choice to others, that fine.
• That your data makes sense within the theory is great, and that the theory helps you think through your data is often enough.  The exact location of your data in the theory is less important.
• Above all the theory is there for you to use, not to overly control and constrain your thinking.

A PRACTICAL EXPERIENCE OF CODING

Q: How did NPT impact on your coding?  What was it like?  How tough was it?

A: It was tough because we were not confident that we understood the intended meaning of the models constructs. We learned a lot about the constructs through an iterative process of coding and interpretation, checking the resonance of the data with its original meaning, our sense of its relevance to a particular construct, debating its relevance to this rather than that NPT construct etc etc. … I don't think it will be ever be so tough again because we have a better sense of what the NPT is about because we 'learned by doing'. This learning will transfer I am sure to other NPT analyses

Q: Did discuss/debate how to code with/via NPT?

A: Yes, A LOT

Q: Did you come to an agreement about how to enact a specific code?

A: We came to agreement through discussion and debate and asking each other loads and loads of questions about our data, the model, why we thought what we thought. Having the freedom to ask 'stupid questions' of each was very important!

However you are using NPT to shape and inform your coding and analysis, you need to be aware that certain topics, issues and concerns will always fall outside the theory.

• Good analysis means openness to emergent issues. This will generally mean that some issues that are relevant to your specific project will fall outside the theory.  That's fine, it's quite routine and normal.  You should never try and force data to fit a theory.
• If you are doing some form of coding via NPT constructs and components, you always need a code that is 'other', as in, issues and topics that do not clearly fit within the NPT. For some projects, where NPT is only a small part of the project, that might be fine.  However, when the ideas behind NPT are central to your research project, and the 'other' code is getting large, or many of your codes cannot be related to NPT, you need to ask various questions of the research project. 
  • Are the things being coded as 'other' central to answering your research questions?  Is NPT really helping you?  Should you be using another theory instead which helps you make sense of more of your data?  Are you being too rigid in the way you operationalize the theory?
  • Are you asking the right questions to the right people?  Are you observing the right sites?   Should you consider going back into the field to collect more data?

Things to consider

• To test out how useful NPT might be for you, think about doing a two stage analysis. First via a traditional method, then via NPT constructs and components.  Then compare them.  It may seem like a lot of work, but you know your data and it will force you out of some of the assumptions that you made the first-time around.  You do not have to do it on your whole data set.
• As with all qualitative research, work in a group, compare your coding, discuss your thoughts, refine (again and again) your coding frame/manual/book.

References

1. Glaser BG. (1978) Theoretical Sensitivity: Advances in the methodology of Grounded Theory. Sociology Press. Back to text
2. Glaser, B.G. (1998) Doing Grounded Theory - Issues and Discussions. Mill Valley, CA: Sociology Press.  Back to text
3. Glaser, B.G. (2005) The Grounded Theory Perspective III: Theoretical coding. Mill Valley, CA: Sociology Press.  Back to text
4. Strauss, A. & Corbin, J.M. (1990) Basics of qualitative research: Grounded theory procedures and techniques. Thousand Oaks, CA: Sage.  Back to text
5. Braun, V. & Clarke, V. (2006) Using thematic analysis in psychology. Qualitative Research in Psychology. 3: 77-101.  Back to text
6. Glaser, B (1965) The Constant Comparative Method of Qualitative Analysis. Social Problems, 12: 436-445.  Back to text
7. Miles, M.B., & Huberman, A.M. (1994) Qualitative Data Analysis: An expanded sourcebook (2nd edn.), Thousand Oaks, CA: Sage.  Back to text