Hospital-based surveillance for severe diarrhoea has been recommended to assess the burden of disease due to rotavirus. However, information on healthcare-seeking patterns of residents in the hospital catchment area is needed first to obtain the burden of disease in the community using the hospital data. A community-based cluster survey was conducted in two districts of Ghana, each served by a single district hospital, to determine the prevalence of severe diarrhoea among and treatment preferences for children aged less than five years. Caretakers of 619 children in Tema, an urban district, and caretakers of 611 children in Akwapim South, a rural district, were interviewed. During the month preceding the survey, the prevalence of severe diarrhoea in children aged less than five years was similar in the two districts (13.6% urban and 12.9% rural), as was the proportion of mothers who sought care outside the home (69.0% urban and 70.9% rural). 48.8% of urban mothers of children with severe diarrhoea visited public/private clinics, 9.5% pharmacies, and 3.6% the district hospital. Whereas, 22.8% of rural mothers visited public/private clinics, 19.0% pharmacies, and 13.9% the district hospital. Results of the study suggest that rotavirus surveillance should be guided by community studies on healthcare-use patterns. Where hospital use is low for severe diarrhoea, rotavirus surveillance should include other health facilities.