Use the Information Technologies to Classify the Outcome of Spontaneous Intracerebral Hemorrhage Patients: the Prolong Intensive Care Unit Length of Stay, Mortality and Medical Expenditure Analysis

• Main Authors: Hsien-Wei Ting, 丁賢偉
• Other Authors: Chien-Lung Chan
• Format: Others
• Language: en_US
• Online Access: http://ndltd.ncl.edu.tw/handle/wqrun3

博士 === 元智大學 === 資訊管理學系 === 103 === Purpose: Spontaneous intracerebral hemorrhage (sICH) patients have a high mortality and morbidity, and use a large amount of medical care resources. This study used the information technologies to classify the outcome of spontaneous intracerebral hemorrhage patients: the prolong intensive care unit length of stay, mortality and medical expenditure Methods: This study analyzed the Longitudinal Health Insurance Database (LHID2005) of the National Health Insurance Research Database from 2005 to 2010. The inclusion criterion was sICH patients suffering a first attack. The change point analysis, Student’s t-test, multiple linear regression, the χ2 test and data mining methods were employed as the statistical methods. Results: The incidence of sICH was 40.1 cases per 100,000 populations per year in Taiwan. The incidence increased by age and was greater in males than in females. The old sICH patients more frequently choose the conservative treatment and the outcome of conservative treatment was better than the surgical intervention. The comorbidities of monthly mortality and annual mortality were shown in this study. This study used SVM for the both features selection and prediction model construction. The area under ROC curve >0.80 were included and 69 attributes (CCS/diseases) were included. Conclusion: The incidence of sICH in Taiwan was higher than that in whites and blacks American and lower than that in the populations of Japanese and Chinese. The features of male and female sICH patients were different. The surgical interventions were both higher hospital expenditure and mortality rate. The old age sICH patient was chosen for conservative treatment but the total hospital expenditure were not significant different from the younger patients. The attributes selections were done and the SVM were better than other methods. The big data studies do not find the relation between cause and effect; they just explored the correlations between factors and discovered the fact. This study found that some CCS/diseases were correlated with mortality indirectly. Although these CCS/diseases were indirect correlated with mortality, they still can be good indicators for the mortality prediction found with SVM.