The aim of the study reported here was to investigate the relationship between health and household composition among older people. The 1 per cent and 2 per cent SARs (Samples of Anonymised Records) drawn from 1991 British Census data were used to examine the frequency of a limiting long-term illness among older people according to different types of living arrangements. These data include the population in institutions and our results show that previous studies based only on the private household population have underestimated the prevalence of illness among older people. Long-term illness rates vary across family and household types, with higher frequencies found for those individuals not living in families (either alone or with others) or in lone parent families, compared with those living as part of a couple. Importantly, our results show a previously unreported clustering of long-term illness in households. Those over 45 suffering from a limiting long-term illness were more likely than those without such an illness, to live in households including others with long-term illness. These results indicate that health should be considered from a household, rather than just an individual, perspective. Our findings support those who have argued that families including an older ill member need more help from formal services. However, it is unlikely that this can be achieved solely by redeploying services from those living alone as long-term illness rates were also high in this group.