Development of complex interventions for management of chronic conditions has become increasingly common, with guidance now provided. Fidelity (whether the intervention is designed, delivered and received as intended) is critical to understanding if, and how an intervention works (or not). However, methods for achieving this are still evolving. This study describes the methods used in the TANDEM trial - a large multicentre study evaluating the impact of a cognitive behavioural intervention preceding routine pulmonary rehabilitation for people with chronic obstructive pulmonary disease and anxiety and or depression. Results for enhancement and training aspects of fidelity, are presented. Using the National Institute for Health Behaviour Change Consortium (NIH BCC) framework of fidelity, a set of enhancement strategies and a fidelity measurement strategy were developed with input from a multidisciplinary team. The Cognitive First Aid Rating Scale (CFARS) was used to assess Facilitator (the respiratory professional delivering TANDEM) therapeutic competence at the end of the initial training and throughout treatment delivery (on a randomly selected set of cases). A TANDEM specific treatment adherence measure was developed following previously recommended procedures. Together these (the CFARS and adherence measure) comprised the TANDEM treatment delivery fidelity tool. Hiring of respiratory professionals to the initial training programme was successful, with 44% of those expressing initial interest in being a Facilitator successfully completing the process. Video recordings of potential Facilitators conducting standardized patient role plays at the end of the initial training demonstrated fidelity of training. Addressing fidelity in complex intervention trials is a time and resource intensive process but has significant potential to increase understanding of results and strengthen the evidence base for effective interventions. By defining a full fidelity assessment method prior to analysis we aimed to minimise bias when interpreting results. ISRCTN59537391 . Registered on 20 March 2017. Trial protocol version 6.0, 22 April 2018. Process evaluation protocol version 4.0, 1 November 2020.