Objectives: This paper considers the methodological problems that arise in conducting cost analyses in economic evaluations where only observational (rather than experimental) data are available and where the technology being evaluated is diffuse, such that the unit of analysis has to be the institution rather than the patient.

Methods: A case study is reported that concerns the application of computer technology in radiology: picture archiving and communication systems (PACS). A range of different approaches were used to estimate changes in running costs, including time series analyses of routine data and direct observation of resource use.Results:The analysis illustrates some of the difficulties involved in costing the introduction of a diffuse technology. Nevertheless, it provides a firm indication that, overall, the introduction of PACS was found to be associated with a significant increase in hospital costs, suggesting that the initial expectations of financial savings were unduly optimistic.

Conclusions: The research demonstrates that, using multiple methods, it is possible to estimate cost changes within a single hospital. In addition, the paper discusses the nature of the uncertainties in such analyses and possible ways of representing such uncertainty in terms of confidence intervals.