Use of data mining techniques in clinical care and management of patients with Methicillin-Resistant Staphylococcus aureus (MRSA) infections

• Main Authors: Lin, I-hsuan, 林怡萱
• Other Authors: Dr. Lin, Yu-sen
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/80865587739232043697

碩士 === 國立高雄師範大學 === 人力與知識管理研究所 === 103 === Methicillin-resistant Staphylococcus aureus (MRSA) is a multi-drug-resistant pathogenic bacteria that receive the most attention. Different strains of S. aureus can produce toxins and cause skin, wounds, osteomyelitis, pneumonia, bacteremia, and other infections. Infection is mainly spread through direct physical contact, skin wounds, crowded environment, and poor personal hygiene may also cause infection. Patients infected with MRSA bacteremia within 30 days mortality was 34%. Selection of antibiotics mainly depends on microbial culture and antibiotic sensitivity test. The purpose of this study is predict outcome of patients with MRSA bacteremia using data mining techniques to provide better management of patient therapy. We used MRSA persistence in bacteremia after 7 days, 30 days, and death as endpoints in patients receiving a traditional vancomycin therapy. In this study, we used 29 risk factors associated with MRSA bacteremia. Our results showed that we are able to predict the 7-day persistence of MRSA in blood cultures at accuracy ranged 82.0% ~ 86.6%; death of patient within 30 days at accuracy ranged 53.4% ~ 69.2%. Such a prediction method can be applied in hospitals by use of a prospective study to collect patient data in order to establish their predictive models.