In a baseline analysis of the faecal microbiota composition of 161 older persons, we previously reported a core microbiota and aggregate composition that was distinct from younger persons. We also identified significant inter-individual variation at phylum level. To investigate further, we analyzed the microbiota composition of 178 elderly subjects, none receiving antibiotics, and for whom we had collected dietary intake information. The data revealed distinct microbiota composition groups. Clustering of subjects was also distinguishable by analysis of faecal metabolites and shot-gun metagenomic data. Major separations in the microbiota correlated with selected clinical measurements. Correlations in the data between diet, microbiota and health status suggested a causative axis. We have extended the analysis to 732 samples across serial time points and a full spectrum of health status, using novel clustering approaches to identify microbiota groupings. Subjects in long-term residential care, previously shown to have a lower Healthy Food Diversity index, had a lower diversity microbiota and their microbiota was also less stable over time (6 months) than people living in the community. Novel clusters defined by microbiota composition and stability showed correlations with food intake and health parameters, including frailty. The data indicate that diet can programme microbiota composition, function and stability, and impact on health.