運用大型臨床資料庫分析疾病與氣候在二十四節氣下之相關性

Abstract

溫度等氣候因素與疾病間的關係，在某些疾病已經被廣為了解，但是有些氣候與疾病間的關係尚未得到全面性的分析，尤其在二十四節氣方面對每年的疾病與氣候關係探討的研究還不夠充份。二十四節氣是中華文化特有的農業社會的曆法，與一年四季中的氣候變化相配合，使農作收穫最佳化。中醫古籍指出四季天氣的變動與身體協調失衡時會導致疾病發生，但文獻中沒有確切統計數據可具體的探討對每年二十四節氣中疾病與氣候的關係。 本研究針對國家衛生院之健保資料庫中2005至2010年的門診資料，與中央氣象局6年內氣溫氣壓等資料，將公曆中十二月份對應二十四節氣後進行探勘，最後以交叉相關係數呈現氣候-節氣共病之關係，並進一步嘗試以現代理論解釋中醫在二十四節氣時間尺度下氣候對疾病的影響。 除了以年平均趨勢研究初步觀察各疾病的節氣趨勢外，本研究亦使用希爾伯特-黃轉換法中的經驗模態解析(Empirical Mode Decomposition, EMD)與整體經驗模式分析(Ensemble Empirical Mode Decomposition, EEMD)將統計之人數解析成數種內部模態函數(Intrinsic Mode Function, IMF)一一探討各疾病在二十四節氣下的趨勢變化，以及互相比較各疾病來探討其中之共病性，同時也對整體經驗模態分析實驗數種參數來選擇對該疾病的理想解。而與氣候的關係亦為本論文探討方向之一。 經由以上方法解析後除了未分解前的結果以外，也可得到 EMD/EEMD雜訊濾除後較平滑的年週期波動，同時從疾病資料分解出的波動變化可以探勘出多種疾病的潛在關係。從選出疾病中的門診人數統計上心理、消化疾病在春夏交際與秋季中旬達到高峰，泌尿系統疾病在夏秋兩季為高峰期。經過直接統計與EMD/EEMD分解後以交叉相關係數比較的結果也十分豐富，其中疾病方面找出高血壓分別與糖尿病二型和氣喘有高度相關性。氣候方面我們從溫度、氣壓、降水量、日照時數等氣候因子對疾病的影響，分析結果顯示其中以溫度與氣壓與疾病的相關性最大。以上相關性分析結果可作為未來在探討氣候與疾病間因果關係研究的重要參考。

The relationships between temperature and other climatic factors and diseases have been widely understood in some relationships between climatic factors and diseases but not been comprehensive enough in the other certain relationships, especially in the twenty-four solar terms time-scale. “Solar Terms” in Chinese culture is the specific calendar in conjunction with the seasonal climate in a year and can create best harvest. In traditional Chinese medicine the climate changes of four seasons makes human body imbalance and cause disease, but there are not enough literatures to present exact statistics specific in the relationship between disease and climate on the annual twenty-four solar terms. In this study we used the data from outpatient records during years from 2005 to 2010 in National Health Insurance Research Database (NHIRD) and climatic factors including temperature and atmospheric pressure during the same years in Central Weather Bureau. Data in Gregorian calendar time scale would be transformed into twenty-four solar terms and mined. We also used cross-correlation coefficients to present the relationships between climates and diseases, and tried to explain how climatic factors impact diseases in the time scale of the solar terms on the view of traditional Chinese medicine by the modern theory. In addition to the observations to the average annual trend of diseases, this study also used Empirical Mode Decomposition (EMD) and Ensemble Empirical Mode Decomposition (EEMD) in Hilbert-Huang Transform to decompose many Internal Mode Functions (Intrinsic Mode Function), exploring the trend of diseases in the time scale of solar terms, as well as comparing diseases to explore the comorbidity. We also compared many parameters to select the ideal solution to the disease data decomposed by EEMD. The trends of climate data and the relationship with diseases were also explored by the aforementioned methods With the analysis procedures addressed above, several interesting results have been observed. Firstly, the numbers of cases of mental disorders and the diseases of the digestive system peak between spring and summer and in mid-fall. Furthermore, the periodical components output by the EMD/EEMD algorithms point to that hypertension is highly correlated to diabetes type II and asthma. With respect to the impact of temperature, barometric pressure, sunshine time, and precipitation on disease development, analytic results reveal that temperature and barometric pressure play more significant roles. In summary, the analytical results presented in this thesis provide valuable clues for future investigation of the cause-effect relations between climate and diseases.