| Summary: | 碩士 === 長庚大學 === 臨床醫學研究所 === 105 === Stroke events cause damage to brain tissues. Some of the damage leads to long-term neurological and physical disability or even death. It’s an important topic to develop accurate prediction model to predict mortality hazard of stroke patients. Accurate hazard prediction could benefit clinician not only on optimizing patient management but on administering limited medical resources. Generally speaking, Cox proportional hazard model is a statistical method which was constantly applied to analyze relationship between risk factors and mortality in stroke patients. However, this study method has several limitations. Every patient need to be clearly classified into specific subgroup if the risk factors you are interested in is a categorical factor. There will be bias in study results when there is no gold standard to help you assign patients into specific subgroup. The objective of this study is establishing mixture Cox proportional hazard model to predict short-term and long-term mortality hazard of ischemic stroke patients and compare prediction ability between mixture Cox proportional hazard model, Cox proportional hazard model and stratified Cox proportional hazard model. In order to control the effect of admission blood pressure status on stroke mortality without labeling patients as specific status, baseline proportion of two mixture model components were set according to distribution of stroke patients’ admission blood pressure status. Study results showed different risk factors estimation between mixture Cox proportional hazard model and two other models. Prediction ability of mixture Cox proportional hazard model is better than the other two.
|
|---|
