Dear Alison,

It might be beneficial in the first instance to take a step back and ask - what is the architecture of referral? Developing a middle-ranged framework for referral might become a valuable output of your research, so that service deliverers across healthcare and other sectors can consider the many factors at play (service level, clinical, social and psychological) in understanding referral success or failure.

you wrote:

I think it's justifiable to conceive of the referral process as the intervention you are investigating. There are many people with this kind of focus using realist evaluation to study service re-design/improvement. At the core of the intervention is the decision-making process: to refer or not, and/or to refer to an option among a variety of options. It would be useful to sketch out the C, M, O definitions as well: the mechanism response for example could be seen generally as referral decision making . The mechanism resource can include initiatives to improve referrals (e.g., training).
The context of referral may be aspects such as adequate staffing, awareness of available services, quality of doctor-patient communication etc. 
Outcomes would likely include the big picture outcomes of interest, e.g.,  clinical and health outcomes based on referral appropriateness/success etc.

Finally, you might consider a distinction between the referral decision making process and the referral pathway. The former is the mechanisms that is triggered in the minds of healthcare providers responsible for referring and the latter includes the wider context of the healthcare system that will include barriers and enablers in relation to referral decision making. 

best of luck,

Justin