Identifying the risk factors of 30 days mortality for ventilator dependent patients

• Main Authors: Loo, Yeong-Dar, 羅永達
• Other Authors: Ma, Tso-Chiang
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/77154987500363986213

碩士 === 中國醫藥學院 === 醫務管理研究所 === 90 === This study was designed to investigate the risk factors of 30 days mortality for ventilator dependent patients on respiratory care ward (RCW), and to establish the predictive model. Logistic regression was utilized to analyze data. A retrospective collection data of 237 ventilator dependent patients gathered from 6 RCWs, located at Northern, Central and Southern Taiwan. The principal findings were as follows. 1. Among 237 patients, total mortality were 81 patients (34.18%). 132 of them (55.70%) were male and the rest of them (103, 44.3%) were female. Age ranged from 26 to 96 years old. Average age was 75 years old, with 107 patients (45.15%) age ranging from 71-80 years. Average hospital stay was 10.68 months. 2.With regards to bivariate analysis, the length of hospital stay, Creatinine, and WBC count were statistically significant associated with mortality at the alpha level of 0.01. In addition, history of DM, duration under ventilator support, artificial airway and dermal infection were statistically significant impaction motality at the alpha level of 0.05. 3.Controlling other variables by using Logistic regression, six variables were identified as the risk factors of 30 days mortality for ventilator dependent patients on RCW at the alpha level of 0.05. There were length of hospital stay, WBC, infection of skin, artificial airway, causative factors of respiratory failure, and creatinine level. 4.A predictive model of 30 days motality for ventilator dependent patients on RCW, consisted of six variables including length of hospital stay, length of time under ventilator support, Creatinine level, different artificial airway, WBC count and dermal infection was established by using stepwise Logistic regression. The accuracy of classification for this model is high enough to validate our result (percent concordant 90.1). Our results could futher assist third-payer to establish the index of case-mix and quality assessment tool. In addition, the manager of RCW also could use the result of this study to redesign care procedure to assure good quality of care. Finally, the predictive model obtained from our study could be utilized as a risk evaluation model for health care providers.