OBJECTIVE: Although economic evaluations are increasingly conducted on a multinational basis, the issue of how to analyse and interpret multinational economic data has received little attention. This research illustrates the use of cluster analysis to investigate patterns in resource utilisation between different countries or centres, and examines the use of multi-level modelling to explicitly detect a hierarchical structure to the data if one is present and to explore the relationship between site and outcome. METHODS: The cluster analysis and multi-level modelling using SPSS for Windows and MLn, were based upon resource utilisation data captured in a multinational phase III randomised controlled trial of two alternative drug treatments for rheumatoid arthritis. In this trial, data (e.g. in-patient episodes, out-patient visits, therapeutic procedures, diagnostic procedures, GP visits, physiotherapy sessions) were collected for 374 patients over a 108 week period, enrolled at 56 centres in nine European countries. Clustering was investigated employing various linkage and distance measures, and parameters were estimated for the multi-level model specified with countries as level 3 units, centres as level 2 units and patients as level 1 units. RESULTS: A variety of clustering solutions were generated according to the resource category examined and visually presented in dendrogramatic form and icicle plots. Across all categories, the countries resolved into 4 clusters (Cluster 1=UK, Denmark, Norway, Netherlands; Cluster 2=Germany, Finland; Cluster 3=Belgium; Cluster 4= France). In the multi-level model, variances were observed operating at the different levels which could be explained in terms of the general characteristics of the sites. CONCLUSIONS: The results of this analysis suggest that cluster analysis and multi-level modelling techniques can be valuable tools in the exploration of multinational economic data, although they require considerable care in practice. The presence of clustering may have important implications for costing and the pooling of multinational economic evaluation data.