Summary: | 碩士 === 淡江大學 === 統計學系碩士班 === 101 === Burn trauma could deeply affect patients and their family members, as well as the social economic burden. According to patients’ condition of burn trauma, they might need hospitalizations. Once one can predict the length of hospital stay more precise, it could lower the stress of uncertainty in the mental or spiritual health for patients and let patients’ family members be prepared for caring the patients. Besides, physicians could be more confidential to evaluate the treatment plan and allocate the medical resources. This study analyzed the data of acute burn injury patients of a medical system in Taiwan collected between January 13, 1997 and November 1, 2010. The following four predictive models are constructed: prognostic factors, total burn surface area, burn index and prognostic burn index models on the basic of median regressions. Model selection was processed by use of backward elimination procedure with AIC index and the goodness-of-fit test statistic Pseudo-R square. Eventually, prognostic factors model had better predictive ability than other models to predict the length of hospital stay. Variables included in the model were second degree of burned area, third degree of burned area, age, burn area 1(head and neck) and admission to intensive care unit.
|