- 學位論文
應用空間統計探討臺灣COVID-19 確診數和環境因素之關係
Spatio-temporal analysis of Environmental Factors and their relationship with COVID-19 in Taiwan
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
隨著疫情升溫，病毒在傳播力方面的重要性日益顯現，也引起各 界關注，然而臺灣因為和他國的文化差異以及公衛政策的不同，使得 疫情爆發時間軸截然不同，因此，研究疫情潛在風險因子和區域才能 有效進行防疫工作。本文研究目的是評估環境因子對臺灣COVID-19 感染風險的影響，使用克利金法對缺乏測站區域之空氣汙染物進行 補值，運用貝氏階層模型分析，為避免共線性問題將PM2.5 和PM10 分開建模，在確診數階層分別採用卜瓦松分布、負二項分布以及零 膨脹卜瓦松分布，空間效應則以Besag York-and-Molliè 模型(BYM) 和 Leroux 模型作為比較，WAIC 作為選模標準，在此六種模型中，以 負二項分布結合BYM 模型的效果最佳。結果顯示，不論是以PM2.5 或PM10 建模的模型，SO2、O3 和人口密度皆為顯著正相關，另外在 PM10 的模型中，PM10 亦為顯著正相關，而在PM2.5 模型中，當分別提 高1ppb SO2、1ppb O3 和1 單位標準化後人口密度，將會提升22.6%、 2.05% 和7.14% 的感染率，在PM10 模型中則為提升27.07%、2.24% 和 6.53% 的感染率。本文表明部分空氣汙染因子與新冠肺炎確診數具顯 著相關，以及不同時間點的熱點區域，在未來的空氣監管和醫療資源 分配政策具有參考價值。
並列摘要
As the COVID-19 spreads rapidly, studying virus transmission plays a crucial role in epidemiology. However, there is a very different COVID-19 pandemic timeline from other countries because of Taiwan＇s unique cultural differences and public health policies. Consequently, there is a need for understanding potential risk factors and the spatio-temporal trend of COVID-19 incidence. This study aims to investigate the impact of environmental factors on the risk of COVID-19 infection in Taiwan. The missing values of air pollutants in areas without monitoring stations are imputed using the kriging method. Hierarchical Bayesian spatio-temporal models are implemented to investigate the association between covariates and the number of cases by assuming different distributions for the confirmed case data, such as Poisson distribution, negative binomial distribution, and zero-inflated Poisson distribution. Spatial effects are considered using the BYM and Leroux models. The model selection criterion based on WAIC reveals that the combination of negative binomial distribution and BYM model best explains COVID-19 incidence rate. In addition, the results show that SO2, O3, and population density have significantly positive association with COVID-19 incidence in both PM2.5 and PM10 models. Additionally, the PM10 model shows a significant positive correlation between PM10 and the incidence rate. In the PM2.5 model, an increase of 1 ppb in SO2, 1 ppb in O3, and the standardized population density increases by 1 unit corresponds to a 22.6%, 2.05%, and 7.14% increase in incidence rates, respectively, while in the PM10 model, the corresponding increases are 27.07%, 2.24%, and 6.53%. This study reveals the significant associations between certain air pollutants and COVID-19 incidence, as well as the identification of hotspots at different time periods. These findings provide valuable insights for future air pollution regulations and resource allocation in healthcare.
參考文獻
政府資料開放平台 (2021). 鄉鎮市區界線 (TWD97 經緯度) 各鄉 (鎮、市、區) 行政區域界線圖資. https://data.gov.tw/dataset/7441.
政 府 資 料 開 放 平 台 (2022). 村 里 戶 數、 單 一 年 齡 人 口 （新 增 區 域 代 碼）.https://data.gov.tw/dataset/77132.
行 政 院 環 境 環 保 署 (2022). 空 氣 品 質 監 測 網 污 染 物 測 值 月 報 表.https://airtw.epa.gov.tw/CHT/Query/Month_Value.aspx.
中央氣象局 (2022). 觀測資料查詢 (CWB Observation Data Inquire System). https://eservice.cwb.gov.tw/HistoryDataQuery/index.jsp.
Adhikari, A. and Yin, J. (2020). Short-term effects of ambient ozone, pm2. 5, and meteorological factors on covid-19 confirmed cases and deaths in queens, new york. International journal of environmental research and public health, 17(11):4047.