應用資料探勘技術分析多重疾病間的共病現象

Applying Data Mining Techniques in Analyzing the Comorbidity between Multiple Diseases

王凱勳(Kai-Hsun Wang)

指導教授 ： 歐陽彥正

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2011年)

https://doi.org/10.6342/NTU.2011.10775

摘要

共病，指除了主要的疾病外還出現了其它的疾病，傳統在分析兩個疾病(一對一)的共病關係時，會依流行病學病例對照研究法(case control study)算勝算比(odds ratio)來代表二者的相關程度，但假如想一次討論某個目標疾病和另外多個疾病(一對多)的共病關係時，一對一的算法便不符效益，本論文提出一種方法可解決上述問題，觀念類似LDA(Linear Discriminant Analysis)，藉著修改勝算比人數分配表左邊的判斷式成一個線性方程，導入勝算比(odds ratio)作為目標函數(objective function)，配合使用RAME搜尋演算法，多種疾病和目標疾病的相關性可被算出，每種疾病所占的比重也能知道。
本論文從全民健康保險研究資料庫中選出三高疾病患者，在這些病人上分析冠狀動脈疾病和7種精神疾病的共病關係，結果顯示，對高血脂的人來說，同時有焦慮症、恐慌症，或同時有恐慌症、強迫症的病人，二者皆和冠狀動脈疾病有很高的相關性；對高血壓的人來說，同時有焦慮症、恐慌症的病人，其和冠狀動脈疾病的相關性也比較高；至於對高血糖的人來說，同時有焦慮症、恐慌症，或同時有恐慌症、精神官能性憂鬱症，或同時有恐慌症、躁鬱症的病人，三者和冠狀動脈疾病的相關性也很高。

關鍵字

多變量分析 ； RAME演算法 ； 勝算比 ； 三高疾病 ； 精神疾病 ； 冠狀動脈疾病

並列摘要

Comorbidity is the presence of one or more diseases in addition to a primary disease. Traditionally when analyzing the comorbidity between two diseases (one-to-one), odds ratio is calculated using case control study to represent the strength of association. But if we want to discuss the association between one target disease and other diseases (one-to-many) at the same time, the one-to-one calculation is insufficient. This paper presents an approach to the above problem. The concept is similar to LDA (Linear Discriminant Analysis). Through the modification of the inequality beside odds ratio’s contingency table into a linear equation, and importing odds ratio to the objective function, plus in coordination with RAME search algorithm, the strength of association between the target disease and other diseases could be calculated; and further, each proportion of diseases to target disease is realized.
This paper chooses patients with the three highs: hyperlipidemia, hypertension, and diabetes from National Health Insurance Research Database (NHIRD). Then, it analyzes over these patients the comorbidity relation between coronary artery disease (CAD) and 7 mental disorders. The result shows, in patients with hyperlipidemia, those who suffer from anxiety state and panic disorder, or from panic disorder and obsessive compulsive disorder, have strong association with CAD. In patients with hypertension, those who suffer from anxiety state and panic disorder have high association with CAD. In patients with diabetes, those who suffer from anxiety state and panic disorder, or from panic disorder and neurotic depression, or from panic disorder and bipolar, have strong association with CAD.

並列關鍵字

multivariate analysis ； RAME algorithm ； odds ratio ； hyperlipidemia ； hypertension ； diabetes ； mental disorder ； coronary artery disease

參考文獻

[6] M. Kenneth S. Kendler, et al. (2009) Major Depression and Coronary Artery Disease in the Swedish Twin Registry: Phenotypic, Genetic, and Environmental Sources of Comorbidity. Arch Gen Psychiatry.

[9] P. Ronald C. Kessler, et al. (2005) Prevalence, Severity, and Comorbidity of 12-Month DSM-IV Disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry.

[15] M. Imran Shuja Khawaja, et al. (2009) Depression and Coronary Artery Disease: The Association, Mechanisms, and Therapeutic Implications. Psychiatry.

[16] M. James L. Levenson. (2006) Psychiatric Issues in Heart Disease. Primary Psychiatry.

[17] P. Laura D. Kubzansky, et al. (2007) Prospective Study of Posttraumatic Stress Disorder Symptoms and Coronary Heart Disease in the Normative Aging Study. Arch Gen Psychiatry.