| In order to be reused, the atomic-level data should be formally represented based on shareable schema. This dissertation intended to explore reusability of the atomic-level data stored in Intensive Care Unit (ICU) nursing flowsheets for patient acuity estimation.;Many ambiguous data representations which may hinder reusing the data were clarified by domain modeling. The concept model was built first to disambiguate the meaning of the terms in the flowsheet. Then the information model was built on top of the concept model, by adding more classes and layers of information hierarchies to explicitly represent the logical and semantic relations embedded in the information structure. Also the instances of the information model were directly linked to those in the terminology model.;After clarying the ambiguous representation in the flowsheets, the flowsheet contents and the acuity concepts were paraphrased into short clinical statements and mapped to SNOMED-CT (Systematic Nomenclature of Medicine-Clinical Terms). SNOMED-CT provided complete matches to 78.13% of the clinical statements from the flowsheets and 60.80% of the clinical statements from the acuity indicators. Lack of appropriate terms in SNOMED-CT was the major cause of the incomplete matches with the flowsheet statements, and limitations in the concept models were attributed to the majority of the incomplete matches with the acuity statements.;Based on the inputs from expert reviewers, who took responsibility of educating and auditing the use of the acuity system decision algorithms were developed, then coded with Java Expert Shell System (JESS). This Rule-Base System (RBS) determined the indicator values with the retrospectively collected flowsheet data of 32 patients and the results were compared to the indicator values previously determined by staff nurses who were involved in the clinical care. Also, reference standards were produced for randomly reselected small subset of patients, to evaluate the staff nurses' decisions and the RBS' decisions. In general, the agreement rates between the staff nurses and the reference standards were slightly higher than those between the RBS and the reference standards. Two major factors that compromised the accuracy of the RBS were incompleteness of the source data and the incompleteness of the rules themselves. |
