The aim was to provide evidence of mortality and community care costs of people living in care homes and to investigate its association with mental health based on the Mental Health Clustering Tool (MHCT). In an observational study, 5,782 residents living in 104 care homes were followed from 2014 to 2016. Residents were categorised into four groups using the MCHT: three with mental health conditions, ‘non-psychotic’, ‘psychotic’ or ‘organic’; and one without mental health conditions, ‘non-clustered’. Generalised estimating equations were used to explore associations between mean community care costs over 6 months per patient and the clustering of residents into the four groups. Differences in survival rates of residents were plotted using Kaplan–Meier curves and tested with the log-rank test and Cox regression analysis. Community care costs were similar among residents with dementia (£431) and without mental health conditions (£407), while costs were higher among residents with non-psychotic (£762) and psychotic (£1,724) mental health conditions. After adjusting for patient and care home characteristics, residents with dementia were 30 per cent less likely to die compared with residents without mental health conditions. Similarly, residents with psychotic conditions and residents with non-psychotic conditions were 25 and 20 per cent less likely to die, respectively, than residents without mental health conditions. The MHCT seems to provide an informative stratification of care home residents with regards to survival and community care use.