近年來,由於環境變遷、人口壽命延長以及生活型態改變,臺灣公共衛生的研究趨勢已從較早期所關注的營養不良或急性傳染病等逐漸轉移心血管、腦血管及慢性下呼吸道等疾病已成為當前相當重要的公共衛生研究課題。造成以上疾病的因素繁多,除個人生活習慣、人口因素(如年齡、性別等)外,目前國內外已累積相當數量的文獻將一部分影響疾病的因素指向氣象因子和空氣污染物。 本研究希望利用統計方法,針對特定疾病族群進行精緻化的風險預測。故本研究目的為針對缺血性心臟病(ICD-9:410-414)以及腦血管心臟病(ICD-9:430-438)建構天氣因子、空氣污染因子與疾病發病風險的預測模型。 本研究利用2000年至2009年的健保資料庫篩選出50-70歲、有住院資料的缺血性心臟病以及腦血管疾病患者做為研究族群。暴露因子包括交通部氣象局提供的天氣因子逐時資料(溫度、相對濕度和大氣壓力),以及環保署空氣品質監測網提供的空氣污染物逐時資料(空氣中懸浮微粒(PM10)、二氧化氮(NO2)、二氧化硫、一氧化碳(CO)以及臭氧(O3))。而地理分析上,採用以鄉鎮為單位劃分。在大台北地區,並非所有的觀察點皆有測站濃度資料,故利用克利金方法(kriging technique)來進行地理空間濃度的推估。 本研究利用stepwise logistic model selection進行模式選擇。在心血管疾病中SO2有達到顯著正相關(OR=1.096)、以及PM10達到顯著負相關(OR=0.986)、男性復發風險也較女性高1.796倍。而在腦血管疾病中,當天最大溫差有達到顯著負相關(OR=0.932),以及年齡達到顯著負相關(OR=0.952),男性復發風險也較女性高1.534倍。
As the public health issues shift from malnutrition and emerging infectious diseases to cardiovascular and cerebrovascular diseases, more and more researchers have focused on the investigation of the causal and risk factors of such diseases. For instance, a considerable amount of literature has indicated that meteorological factors and air pollutants may contribute to the risk of diseases. Funded by the Taiwan Central Weather Bureau, this study aims to develop a predictive model for the risk of disease based on specific populations and meteorological factors. The diseases of interest include Ischemic Heart Disease (IHD with ICD-9 coding 410-414) and Cerebrovascular Disease (CVD with ICD-9 coding 430-438) occurred in Taiwan. Our study population was selected from the National Health Insurance Research Database with an admission record of IHD or CVD between 2000 and 2009 and aged between 50 and 70. Exposure factors included the meteorological hourly data provided by the Taiwan Central Weather Bureau, and the environmental parameters downloaded from the Taiwan EPA air quality monitoring website.For areas with missing covariates, the ordinary Kriging method was applied for imputation. The results of this study provides a probabilistic measure for the risk of disease based on given meteorological and pollutant variables. In addition, such quantification can demonstrate the uncertainty in prediction, and offer risk information for the group of susceptible individuals for health management.