Background An unplanned hospital admission of a nursing home resident distresses the person, their family and nursing home staff, and is costly to the NHS. Improving health care in care homes, including early detection of residents’ health changes, may reduce hospital admissions. Previously, we identified four conditions associated with avoidable hospital admissions. We noted promising ‘within-home’ complex interventions including care pathways, knowledge and skills enhancement, and implementation support. Objectives Develop a complex intervention with implementation support [the Better Health in Residents in Care Homes with Nursing (BHiRCH-NH)] to improve early detection, assessment and treatment for the four conditions. Determine its impact on hospital admissions, test study procedures and acceptability of the intervention and implementation support, and indicate if a definitive trial was warranted. Design A Carer Reference Panel advised on the intervention, implementation support and study documentation, and engaged in data analysis and interpretation. In workstream 1, we developed a complex intervention to reduce rates of hospitalisation from nursing homes using mixed methods, including a rapid research review, semistructured interviews and consensus workshops. The complex intervention comprised care pathways, approaches to enhance staff knowledge and skills, implementation support and clarity regarding the role of family carers. In workstream 2, we tested the complex intervention and implementation support via two work packages. In work package 1, we conducted a feasibility study of the intervention, implementation support and study procedures in two nursing homes and refined the complex intervention to comprise the Stop and Watch Early Warning Tool (S&W), condition-specific care pathways and a structured framework for nurses to communicate with primary care. The final implementation support included identifying two Practice Development Champions (PDCs) in each intervention home, and supporting them with a training workshop, practice development support group, monthly coaching calls, handbooks and web-based resources. In work package 2, we undertook a cluster randomised controlled trial to pilot test the complex intervention for acceptability and a preliminary estimate of effect. Setting Fourteen nursing homes allocated to intervention and implementation support (n = 7) or treatment as usual (n = 7). Participants We recruited sufficient numbers of nursing homes (n = 14), staff (n = 148), family carers (n = 95) and residents (n = 245). Two nursing homes withdrew prior to the intervention starting. Intervention This ran from February to July 2018. Data sources Individual-level data on nursing home residents, their family carers and staff; system-level data using nursing home records; and process-level data comprising how the intervention was implemented. Data were collected on recruitment rates, consent and the numbers of family carers who wished to be involved in the residents’ care. Completeness of outcome measures and data collection and the return rate of questionnaires were assessed. Results The pilot trial showed no effects on hospitalisations or secondary outcomes. No home implemented the intervention tools as expected. Most staff endorsed the importance of early detection, assessment and treatment. Many reported that they ‘were already doing it’, using an early-warning tool; a detailed nursing assessment; or the situation, background, assessment, recommendation communication protocol. Three homes never used the S&W and four never used care pathways. Only 16 S&W forms and eight care pathways were completed. Care records revealed little use of the intervention principles. PDCs from five of six intervention homes attended the training workshop, following which they had variable engagement with implementation support. Progression criteria regarding recruitment and data collection were met: 70% of homes were retained, the proportion of missing data was < 20% and 80% of individual-level data were collected. Necessary rates of data collection, documentation completion and return over the 6-month study period were achieved. However, intervention tools were not fully adopted, suggesting they would not be sustainable outside the trial. Few hospitalisations for the four conditions suggest it an unsuitable primary outcome measure. Key cost components were estimated. Limitations The study homes may already have had effective approaches to early detection, assessment and treatment for acute health changes; consistent with government policy emphasising the need for enhanced health care in homes. Alternatively, the implementation support may not have been sufficiently potent. Conclusion A definitive trial is feasible, but the intervention is unlikely to be effective. Participant recruitment, retention, data collection and engagement with family carers can guide subsequent studies, including service evaluation and quality improvement methodologies. Future work Intervention research should be conducted in homes which need to enhance early detection, assessment and treatment. Interventions to reduce avoidable hospital admissions may be beneficial in residential care homes, as they are not required to employ nurses. Trial registration Current Controlled Trials ISRCTN74109734 and ISRCTN86811077. Funding This project was funded by the National Institute for Health Research (NIHR) Programme Grants for Applied Research programme and will be published in full in Programme Grants for Applied Research; Vol. 9, No. 2. See the NIHR Journals Library website for further project information.