| Summary: | 碩士 === 嘉南藥理科技大學 === 醫療資訊管理研究所 === 95 === BACKGROUND: Previous drug-drug interactions (DDIs) research has focused on the incidence and related factors of DDIs. However, few studies have examined DDIs at the patterns of medication use and even fewer studies addressed the rules for severity level prediction adopting data mining techniques. OBJECTIVES: To examine the incidence in outpatients visits involving DDIs and explore prescription drugs association rules and DDIs level prediction rules. METHODS: This retrospective research was conducted using data from the 2004 Taiwan National Health Insurance Research Database (NHRID) which contains outpatients’ prescriptions, therapeutic and registration files. DDIs prescriptions were retrieved using the DDIs database of the Department of Health. The Anatomical Therapeutic Chemical Classification System with Defined Daily Doses (ATC/DDD system) was used to identify drugs at the name of chemical substance instead of brand name. A decision tree model was built to predict the DDIs level. Apriori computational method was needed for mining the patterns of medication use. RESULTS: The sex did not differ in the level of DDIs. Outpatient’s age was associated with the level of DDIs. The mean age, number of medications and drug supply days of first level DDIs prescriptions were 64.6±15.9years, 6.5±2.6 medication and 24.3±7.8 days, and with a significantly higher in other level DDIs, respectively (P < 0.0001). Outpatients with internal cardiovascular specialty, essential hypertension and hypertensive heart disease were treated most with DDIs prescriptions. Prevalence of secondary level DDIs was 51.41% in internal cardiovascular specialty, 57.13% in essential hypertension outpatients’ prescriptions. The results of elderly and cardiovascular DDIs levels were also similar, with secondary level outnumbering other levels in DDIs. In terms of the medication association rule, isoniazid and sulfonamide (tuberculosis drugs), together with methotrxate and sulfonamide (rheumatoid arthritis drugs) were listed in top ten item sets of the outpatients’ first level DDIs prescription. The medication combination of digitalis glycosides and loop diuretics was listed as top one in spite of the data from outpatients, elderly or cardiovascular prescriptions. The first hierarchy level of C5.0 decision tree was principal diagnosis. Association rules with support and confidence threshold values of 5% and 40% generated one association rule with DDIs secondary level, carvedilol and aspirin, which was contraindication to heart failure. However, this study found that carvedilol and aspirin was only prescribed 0.21% in heart failure patients.
CONCLUSIONS: Digitalis glycosides and loop diuretics was still the most frequent DDIs medication combination. The principal diagnosis was of pivotal importance for the DDIs level. In the light of prescription diagnosis and association rules, the patterns of medication use were reasonable. The knowledge hidden behind the results uncovered the current computerized DDIs alerts are often over “alerts” because of the DDIs knowledge bases are highly inclusive, placing more emphasis on breadth of coverage than on clinical relevancy or severity of adverse events. Therefore it is essential to design a computerized prescribing decision support with customized knowledge bases.
|
|---|
