院內感染的問題,在抗生素濫用的台灣,是一個重大的隱憂。細菌的多重抗藥性,讓平常與我們共存的細菌,變成可怕伺機感染的病源。微生物檢驗結果對院內感染的偵測很重要,但是檢驗結果的資料量相當龐大,一般難以在早期偵測出院內感染的發生。本研究的目的是利用資料探勘的方法,從臨床微生物檢驗結果中,挖掘出有效的院內感染關聯法則,作為醫院院內感染的偵測參考。我們以中部某醫學中心的鮑氏不動桿菌檢驗資料為例,運用其與院內感染發生的紀錄,利用卡方檢定與資料探勘的關聯法則,去挖掘與院內感染有關的因素。以卡方檢定來檢測住院病人的性別、年齡層、住院單位與院內感染是否有關。另一方面採用資料探勘的關聯法則,尋找住院病人的性別、年齡層、住院單位、藥物敏感試驗結果與院內感染的關聯法則。卡方檢定的研究結果顯示,以該醫學中心而言,住院病人的性別、年齡層、住院單位等因素與院內感染沒有關聯,另外所擷取出有關院內感染的關聯法則,發現對院內感染的偵測,相當具有參考價值。
The nosocomial infection is an implicitly serious problem for the abusing of antibiotic in Taiwan. The bacteria that could coexist harmlessly with human being became a terrible infective source because of their resistance to multi-drug. The clinical microbiology analysis was very important to detect the nosocomial infections. However, the huge amount of data accumulated from that analysis made it difficult to detect the nosocomial infections in early stages. The purpose of this study is to use data mining to extract effective associative rules about nosocomial infections as references for detecting nosocomial infections in hospital. Taking as examples that the records of the clinical microbiology analysis of Acinetobacter baumannii collected from one of the medical centers in the middle of Taiwan, we use the Chi-Square test and data-mining association rule to identify possible factors regarding nosocomial infection. The Chi-Square test was used to evaluate the relationship between nosocomial infections and the in-patient’s sex, the age levels and the hospital ward. On the other hand, via association rules as the techniques of data mining, we search for relationship, if existed, among in-patient's sexes, the age levels, the hospital ward, antimicrobial susceptibility testing results and the nosocomial infections. According to the results of the Chi-Square test, to that medical center, we found that there was no relationship between in-patient's sexes, the age levels, hospital ward with nosocomial infections. Moreover, we discovered some association rules that might be very valuable for nosocomial infections surveillance.