to bibliographic information when they need it (Association of American Medical Colleges, 1986; Haynes et al., 1989).
CPR systems can provide descriptive, graphical, and statistical analyses of clinical data using standard statistical software packages. In addition, CPR data can be rearranged and analyzed for ongoing quality assessment and to provide physicians and patients with quantitative assessments of the risks of conditions and treatments. (These assessments would promote formal decision analysis of many problems that at present cannot benefit from such analysis because of the lack of required data.) CPR system capabilities can facilitate these tasks in several ways: (1) rapid searching through single or multiple records; (2) sorting information in one record or information aggregated across multiple records; (3) aggregating data across patients by hospital, patient care unit, and department; and (4) allowing easy abstraction of information throughout patients' hospital stays or episodes of care.
Finally, CPR systems should offer practitioners and health care managers a means for implementing quality assurance and cost-management policies at the time and site of care (Barnett et al., 1978; Barton and Schoenbaum, 1990; Tierney et al., 1990). CPR systems can also be a resource for guiding policies and practice by providing analysis of past clinical experience within a provider setting (McDonald and Tierney, 1988).
Managers of health care organizations are faced with ever-increasing demands for data. From outside their organizations, requests come from regulators, payers, and community interest groups, among others; from inside their institutions, inquiries come from researchers and those responsible for utilization management and quality assurance. These managers are well aware that few existing record systems can meet these demands. They also recognize, and are wary of, both the cost and the conflict that may be incurred if they attempt to introduce major changes in record systems. Many institutions lack the capital for a sudden conversion to new computer architectures. Even if the funds were available, however, some managers are concerned that physician-management relationships, already strained in many medical organizations, could be further disrupted by an institution's insistence that physicians accept a new method or pattern of record keeping.
Nevertheless, administrators generally seem to prefer that the patient record not be frozen in its present form. They seek to implement change to a new system in a phased sequence, so that it can be more easily managed. Most user groups, but especially institutional managers, will have trouble implementing a vastly modified record system all at once. Proper design of advanced systems must involve well-considered plans for phasing in the changes that are going to be made. Otherwise, provider institutions in