IS

The average time required by backtracking is analyzed over four models of random conjunctive normal form formulas. Each model gives a result of exp(0(I'(' n)/(r-")), where o is the number of variables from which literals may be formed. s is the number of literals per clause, and o" (1 < a < s) is the number of clauses in the predicate. This indicates that the analysis methods used are insensitive to small changes in the random model.

Clinical care genemtes an immense amount of patient abta that has been archived and manipulated by computerbased information systems. Such computer-based medical record systems improved the accessibi@ of clinical information and made several studies of such information possible. Unfortunately, the care provider's task of retrieving, integmting, and interpreting only those portions of the patient's reconi that are relevant to a specific clinictdproblem is actually becoming increasingly d~~cult. This d~ficu~can be attributedprimarily to the large variety of minimum alzta sets, the heterogeneous formats used to store the ohta, the heterogeneous aiata access methods andprocedures, the varying gmnuhzrity of access to data, the different rigid views of the &ta, and the lack of inter-operability among the information repositories of such data sets. Recognizing the aforementioned issues, we are engaged in a project to buihl a multi-a%ztabaseenvironment tailored for the interopembility of medical information systems. The main building blocks of such a system area multidisciplinary minimum data set and a catalogue for the suppoti of interopembili~and customization functions. In this paper, we report on the design approach used and describe the general architecture of the system. I.

Clinical care genemtes an immense amount of patient abta that has been archived and manipulated by computerbased information systems. Such computer-based medical record systems improved the accessibi@ of clinical information and made several studies of such information possible. Unfortunately, the care provider's task of retrieving, integmting, and interpreting only those portions of the patient's reconi that are relevant to a specific clinictdproblem is actually becoming increasingly d~~cult. This d~ficu~can be attributedprimarily to the large variety of minimum alzta sets, the heterogeneous formats used to store the ohta, the heterogeneous aiata access methods andprocedures, the varying gmnuhzrity of access to data, the different rigid views of the &ta, and the lack of inter-operability among the information repositories of such data sets. Recognizing the aforementioned issues, we are engaged in a project to buihl a multi-a%ztabaseenvironment tailored for the interopembility of medical information systems. The main building blocks of such a system area multidisciplinary minimum data set and a catalogue for the suppoti of interopembili~and customization functions. In this paper, we report on the design approach used and describe the general architecture of the system. I.