Background: The effective management of agitation and other neuropsychiatric and behavioural symptoms in people with dementia is a major challenge, particularly in care home settings, where dementia severity is higher and there is limited training and support for care staff. There is evidence for the value of staff training and the use of psychosocial approaches; however, no intervention currently exists that combines these elements into an intervention that is fit for purpose and effective in these settings based on evidence from a randomised controlled trial. Objective: The objective was to develop and evaluate a complex intervention to improve well-being, reduce antipsychotic use and improve quality of life in people with dementia in care homes through person-centred care, management of agitation and non-drug approaches. Design: This was a 5-year programme that consisted of six work packages.Work package 1 consisted of two systematic reviews of personalised psychosocial interventions for behavioural and psychological symptoms for people with dementia in care homes. Work package 2 consisted of a metasynthesis of studies examining implementation of psychosocial interventions, in addition to developing a draft Well-being and Health for people with Dementia (WHELD) programme.Work package 3 consisted of a factorial study of elements of the draft WHELD programme in 16 care homes.Work package 4 involved optimisation of the WHELD programme based on work package 3 data.Work package 5 involved a multicentre randomised controlled trial in 69 care homes, which evaluated the impact of the optimised WHELD programme on quality of life, agitation and overall neuropsychiatric symptoms in people with dementia.Work package 6 focused on dissemination of the programme. Setting: This programme was carried out in care homes in the UK. Participants: Participants of this programme were people with dementia living in care homes, and the health and care professionals providing treatment and care in these settings. Results: Work package 1: Reviews identified randomised controlled trials and qualitative evidence supporting the use of psychosocial approaches to manage behavioural symptoms, but highlighted a concerning lack of evidence-based training manuals in current use.Work package 2: The meta-analysis identified key issues in promoting the use of interventions in care homes. The WHELD programme was developed through adaptation of published approaches. Work package 3: The factorial trial showed that antipsychotic review alone significantly reduced antipsychotic use by 50% (odds ratio 0.17, 95% confidence interval 0.05 to 0.60). Antipsychotic review plus social interaction significantly reduced mortality (odds ratio 0.36, 95% confidence interval 0.23 to 0.57), but this group showed significantly worse outcomes in behavioural and psychological symptoms of dementia than the group receiving neither antipsychotic review nor social interaction (mean difference 7.37 symptoms, 95% confidence interval 1.53 to 13.22 symptoms). This detrimental impact was reduced when combined with social interaction (mean difference –0.44 points, 95% confidence interval –4.39 to 3.52 points), but with no significant benefits for agitation. The exercise intervention significantly improved neuropsychiatric symptoms (mean difference –3.58 symptoms, 95% confidence interval –7.08 to –0.09 symptoms) but not depression (mean difference –1.21 points, 95% confidence interval –4.35 to 1.93 points). Qualitative work with care staff provided additional insights into the acceptability and feasibility of the intervention. Work package 4: Optimisation of the WHELD programme led to a final version that combined person-centred care training with social interaction and pleasant activities. The intervention was adapted for delivery through a ‘champion’ model. Work package 5: A large-scale, multicentre randomised controlled trial in 69 care homes showed significant benefit to quality of life, agitation and overall neuropsychiatric symptoms, at reduced overall cost compared with treatment as usual. The intervention conferred a statistically significant improvement in quality of life (Dementia Quality of Life Scale – Proxy z-score of 2.82, mean difference 2.54, standard error of measurement 0.88, 95% confidence interval 0.81 to 4.28, Cohen’s d effect size of 0.24; p = 0.0042). There were also statistically significant benefits in agitation (Cohen-Mansfield Agitation Inventory z-score of 2.68, mean difference –4.27, standard error of measurement 1.59, 95% confidence interval –7.39 to –1.15, Cohen’s d effect size of 0.23; p = 0.0076) and overall neuropsychiatric symptoms (Neuropsychiatric Inventory – Nursing Home version z-score of 3.52, mean difference –4.55, standard error of measurement 1.28, 95% confidence interval –7.07 to –2.02, Cohen’s d of 0.30; p < 0.001). The WHELD programme contributed to significantly lower health and social care costs than treatment as usual (cost difference –£4740, 95% confidence interval –£6129 to –£3156). Focus groups were conducted with 47 staff up to 12 months after the end of work package 5, which demonstrated sustained benefits.Work package 6: The outputs of the programme were translated into general practitioner workshops and a British Medical Journal e-learning module, an updated national best practice guideline and a portfolio of lay and care home outreach activities. Limitations: Residents with dementia were not involved in the qualitative work. Conclusions: The WHELD programme is effective in improving quality of life and reducing both agitation and overall neuropsychiatric symptoms in people with dementia in care homes. It provides a structured training and support intervention for care staff, with lower overall costs for resident care than treatment as usual. Future work: It will be important to consider the long-term sustainability of the WHELD programme and cost-effective means of long-term implementation. Trial registration: Current Controlled Trials ISRCTN40313497 and ISRCTN62237498.