【正文】 e, transmission, and processing of information would be useful tools in hospital management. In fact, as early as the 196039。s, hospital administrators in the . began to adopt a variety of information technologies (IT) with the goal of improving quality, reducing costs, and improving operations (Harrington 1964).The goal of this work is to measure the impact of information technology use on hospital operating costs during the late 198039。s and early 199039。 Using a new and broad dataset of hospital IT that contains applicationbased (software based) measures of IT use, this study assesses whether the adoption of HIS is associated with reductions in operating costs in subsequent years. In addition, detailed applicationlevel data is used to isolate specific systems associated with reduced costs as well as the mechanisms that underlie these changes.The basis for this study is a proprietary eightyear panel dataset (19871994) that catalogues the degree of automation at the application level for the plete census of nearly 3,000 . hospitals with more than 100 beds. These data provide a solution to a substantial problem in IT measurement as they are the first that measure the number and type of applications that are automated, regardless of their location within or outside the firm. Outsourcing of hospital IT is mon during this period and as a result, applicationbased measures of IT are more appropriate for this industry.To measure IT39。s effects, this study estimates a series of fixedeffects panel regressions, utilizing data from the American Hospital Association, the Medicare Cost Reports, and the IT dataset. Variable costs are regressed separately on three measures of IT, all based on the applicationlevel data, and each utilizing lagged values of the IT variables. The first measure captures the overall IT intensity of the firm in both the financial/administrative area as well as the clinical domain. The second and third measures capture any cost effects from individual applications and from groups of applications, respectively. Taken as a set, these three measures offer not only evidence on whether reduced costs are found at more automated hospitals, but also clues regarding the mechanisms by which cost changes occur.The current results find significant associations between hospital information systems and hospital costs. In particular, the aggregate results indicate that adoption of new systems at the most thoroughly automated hospitals is associated with lower operating costs three and five years after adoption. In contrast, the least IT intensive hospitals show higher costs after adopting new systems. Further, this pattern is evident within both the financial/administrative domain as well as the clinical domain.These results are mirrored in the applicationlevel regressions. Adoption of older systems within either domain is associated with rising costs after adoption while adoption of newer financial management and patient care systems is associated with lower costs. The unifying feature of many of these latter systems is their ability to provide newer and better information to decision makers, regardless of the domain in which they lie. As well, the clinical systems may aid managers in monitoring doctors39。 behavior. Cost savings seem to arise from increasing the productivity of decision makers or from solving the hospital39。s agency problems. In contrast, there is little evidence that systems which simply automate information processing have the same effect.There are two reasonable explanations for the rising costs associated with adoption at the less automated hospitals: unobserved timevarying hospital effects that drive both higher costs and slower adoption, and learning effects. To rule out the first explanation, statelevel trends are included in the regressions. As well, additional regressions are performed which cast doubt on the importance of these unobserved effects. Instead, the more likely explanation is based on learning effect: the organizational costs of HIS adoption are higher at less automated hospitals (as a result of their inexperience) resulting in higher operating expenses. The importance of learning effects is also supported by the evidence showing lags in costs reductions until three or five years after adoption.The next section focuses on hospital costs, laying out the background and motivation for the empirical approach of this study. It outlines the potential savings from the use of information technologies, describes how IT may impact the use of both fixed and variable resources, and details the precise mechanisms by which this may occur. Section 3 builds on this discussion, developing a more formal model of hospital behavior and deriving the regression equations. The various applicationbased measures of IT are also detailed. The results of the estimation are described in Section 4 while the final section offers a discussion of the results and future directions for research.2 Hospital Costs and the Role of ITThis section focuses on hospital costs, laying out the background and motivation for the empirical strategy of this study. Potential savings from the use of information technologies are outlined and IT39。s potential impact on the use of both fixed and variable resources is discussed. The precise mechanisms by which this may occur are also discussed in order to motivate the empirical specification as well as the interpretations of the results.The Harris (1977) model of the hospital serves as the underlying framework for this analysis. This model highlights the structural features of the hospital and allows a precise delineation of hospital costs. In turn, the model offers the opportunity to identify information related problems exacerbating costs as well as the possible savings that IT may engender. An analysis of these costs and the associated savings from IT are the focus of this section. According to Harris (