An increased incidence of pork-related human salmonellosis in Denmark led to the development of a national control programme for Salmonella in Danish swine herds in 1993. The aim of the programme has been met and now the issue of cost-effectiveness is receiving greater attention. An appropriate way to address this is to bring a risk-based focus to the programme. We describe a practical approach to risk-based surveillance through spatial risk assessment using serological and questionnaire data from 2280 herds in 1995. A mixed effects logistic regression model was fitted and both first- and second-order spatial properties of the random effects were investigated. We identified wet-feeding (OR: 0.64; 95% CI: 0.54-0.75) and SPF health status (OR: 0.65; 95% CI: 0.52-0.81) as protective factors for Salmonella sero-positivity. Purchasing feed (OR: 1.81; 95% CI: 1.61-2.04) was a risk factor. The west of the study area generally, and the north of Jutland in particular, experienced the greatest disease risk after controlling for the covariates. There was some evidence for spatial dependency between farms at distances of 6 km (95% CI: 2-35 km) on the Jutland peninsula. We conclude that when farm location details are analysed in conjunction with routinely recorded surveillance information (such as that collected by the Danish swine Salmonella control programme) and targeted industry surveys (such as those conducted by slaughterhouse co-operatives), our knowledge of the behaviour of disease in animal populations is enhanced and this provides a more informed framework for designing efficient, risk-based surveillance strategies.