透過您的圖書館登入
IP:18.232.187.47
  • 學位論文

應用樣式探勘演算法分析大型臨床資料庫之共存疾病

A Novel Pattern-Based Comorbidity Mining Algorithm in Large-Scale Medical Database

指導教授 : 歐陽彥正
共同指導教授 : 黃乾綱

摘要


在醫學研究上,共存疾病為針對單一病人之指標疾病相關的附加疾病。目前有許多研究皆已證實單一疾病控管並非是病人最佳的治療方針,而共存疾病的診斷能提供更完備的病人健康狀態。因此探索指標疾病之可能共存疾病是相當重要的研究議題,亦能落實預防勝於治療之預防醫學理念。 臨床試驗與觀察性研究為主要的兩種共存疾病之研究方法。臨床試驗為一種需經妥善設計的研究流程,普遍上能提供較強制性的研究證據。然而,進行臨床試驗所花費的時間與經費是研究團隊最關注的問題;隨機分配處理至處理組與對照組更易引發道德爭議。觀察性研究為在現存的資料庫中所進行之研究,相較於臨床試驗,能耗費較少的時間與經費。但資料庫之收集過程常無進行觀察性研究的團隊參與,研究結果的信賴度容易因無嚴格的抽樣流程而受質疑。上述討論說明臨床試驗與觀察性研究能互相補足彼此優缺點,因此使用觀察性研究能提供一些線索以進行臨床試驗的設計。 本研究提出一種樣式探勘演算法,透過結合以族群為基數之大型醫療資料庫,偵測指標疾病可能相關之共存疾病。研究亦以老人型失智症為研究對象,討論演算法的共存疾病偵測效能。由研究結果說明使用本研究所提之樣式探勘演算法所偵測的關聯樣式,能提供醫學人士有價值的線索,可作為後續臨床試驗的重要設計依據。

並列摘要


In medical study, comorbidities refer to the additional diseases a patient may suffer other than the primary disease of concern. In recent years, many medical studies have concluded that focusing on one single disease is not the most effective strategy for disease treatment and diagnosis of comobidities can provide a more comprehensive picture of the health condition of the patient. Accordingly, identifying possible comorbidities of the primary disease of concern is an issue of great significance and is essential for development of preventive medicine. Clinical trial and observational study are the two principal methods for analysis of comorbidity. A clinical trial, which is conducted with a well-designed procedure, generally can provide compelling evidences. However, the cost and time involved in conducting a clinical trial may become a major concern for a research team to carry out such a study. Furthermore, a clinical trial typically involves randomly assigning the patients to the treated group or the control group and therefore may lead to an ethical controversy. On the other hand, an observational study is normally conducted with an existing dataset and therefore is substantially less costly and time consuming in comparison with a clinical trial. However, as the dataset employed in an observational study typically has been collected without the involvement of the research teams that use the dataset, the reliability of the results may be questioned, especially when the samples have not been selected with a rigorous procedure. The discussions above show that the clinical trial and the observational study complement each other in terms of their merits and deficiencies. Accordingly, an observational study can be conducted to collect some clues for the design of clinical trial. This thesis proposes a pattern mining algorithm for identifying the possible comobidities of the primary disease of concern in population-based mediccal databases. This thesis also discusses the effects of applying the proposed algorithm to identify the comobidities of senile dementia. Experimental results show that the comorbidity patterns identified by the proposed pattern mining algorithm provide medical personnel with valuable clues for designing follow-up clinical trials.

參考文獻


14. Fuh, J.-L. and S.-J. Wang, Dementia in Taiwan: Past, Present, and Future. Acta Neurologica Taiwanica, 2008. 17(3): p. 1-9.
1. Feinstein, A.R., The pre-therapeutic classification of co-morbidity in chronic disease. Journal of Chronic Diseases, 1970. 23(7): p. 455-468.
2. Valderas, J.M., et al., Defining Comorbidity: Implications for Understanding Health and Health Services. Ann Fam Med, 2009. 7(4): p. 357-363.
3. Starfield, B., et al., Comorbidity: Implications for the Importance of Primary Care in 'Case' Management. Ann Fam Med, 2003. 1(1): p. 8-14.
4. Piccirillo, J.F., Importance of Comorbidity in Head and Neck Cancer. The Laryngoscope, 2000. 110(4): p. 593-602.

延伸閱讀