Background Therapy is key to effective stroke care, but many patients receive little. Objectives To understand how stroke therapy is delivered in England, Wales and Northern Ireland, and which factors are associated with dose, outcome and resource use. Design Secondary analysis of the Sentinel Stroke National Audit Programme, using standard descriptive statistics and multilevel mixed-effects regression models, while adjusting for all known and measured confounders. Setting Stroke services in England, Wales and Northern Ireland. Participants A total of 94,905 adults admitted with stroke, who remained an inpatient for > 72 hours. Results Routes through stroke services were highly varied (> 800), but four common stroke pathways emerged. Seven distinct impairment-based patient subgroups were characterised. The average amount of therapy was very low. Modifiable factors associated with the average amount of inpatient therapy were type of stroke team, timely therapy assessments, staffing levels and model of therapy provision. More (of any type of) therapy was associated with shorter length of stay, less resource use and lower mortality. More occupational therapy, speech therapy and psychology were also associated with less disability and institutionalisation. Large amounts of physiotherapy were associated with greater disability and institutionalisation. Limitations Use of observational data does not infer causation. All efforts were made to adjust for all known and measured confounding factors but some may remain. We categorised participants using the National Institutes of Health Stroke Scale, which measures a limited number of impairments relatively crudely, so mild or rare impairments may have been missed. Conclusions Stroke patients receive very little therapy. Modifiable organisational factors associated with greater amounts of therapy were identified, and positive associations between amount of therapy and outcome were confirmed. The reason for the unexpected associations between large amounts of physiotherapy, disability and institutionalisation is unknown. Prospective work is urgently needed to investigate further. Future work needs to investigate (1) prospectively, the association between physiotherapy and outcome; (2) the optimal amount of therapy to provide for different patient groups; (3) the most effective way of organising stroke therapy/rehabilitation services, including service configuration, staffing levels and working hours; and (4) how to reduce unexplained variation in resource use. Funding This project was funded by the National Institute for Health Research (NIHR) Health Services and Delivery Research programme and will be published in full in Health Services and Delivery Research; Vol. 8, No. 17. See the NIHR Journals Library website for further project information.