BACKGROUND: Programmes have had limited success in improving guideline adherence for chronic disease. Use of theory is recommended but is often absent in programmes conducted in 'real-world' rather than research settings. MATERIALS AND METHODS: This mixed-methods study tested a retrospective theory-based approach to evaluate a 'real-world' programme in primary care to improve adherence to national guidelines for chronic obstructive pulmonary disease (COPD). Qualitative data, comprising analysis of documents generated throughout the programme (n>300), in-depth interviews with planners (clinicians, managers and improvement experts involved in devising, planning, and implementing the programme, n = 14) and providers (practice clinicians, n = 14) were used to construct programme theories, experiences of implementation and contextual factors influencing care. Quantitative analyses comprised controlled before-and-after analyses to test 'early' and evolved' programme theories with comparators grounded in each theory. 'Early' theory predicted the programme would reduce emergency hospital admissions (EHA). It was tested using national analysis of standardized borough-level EHA rates between programme and comparator boroughs. 'Evolved' theory predicted practices with higher programme participation would increase guideline adherence and reduce EHA and costs. It was tested using a difference-in-differences analysis with linked primary and secondary care data to compare changes in diagnosis, management, EHA and costs, over time and by programme participation. RESULTS: Contrary to programme planners' predictions in 'early' and 'evolved' programme theories, admissions did not change following the programme. However, consistent with 'evolved' theory, higher guideline adoption occurred in practices with greater programme participation. CONCLUSIONS: Retrospectively constructing theories based on the ideas of programme planners can enable evaluators to address some limitations encountered when evaluating programmes without a theoretical base. Prospectively articulating theory aided by existing models and mid-range implementation theories may strengthen guideline adoption efforts by prompting planners to scrutinise implementation methods. Benefits of deriving programme theory, with or without the aid of mid-range implementation theories, however, may be limited when the evidence underpinning guidelines is flawed.