| 摘要: | 近年來，由於許多新興傳染病的出現，加上現今交通運輸進步，使傳染病散播速度更加快速，因此更加顯示迅速了解新興傳染病流行病學的重要性。傳染病的整體防治策略在於如何有效降低該致病原的基礎再生數 ( Basic Reproduction Numbers, R0），其定義為在疾病爆發初期且族群中大部份為易感人口，一個指標病例在其傳染期內平均會傳染的繼發病例數，因此如何準確估計基礎再生數為主要課題之一。流行性感冒統稱為流感( Influenza )，為一種因感染流感病毒所引起的急性病毒性呼吸道疾病。流感疫情控制的重要性除了在疫情爆發時散播速度快且範圍廣泛外可能還會產生嚴重併發症而導致死亡。
本研究主要討論流感疫情之基礎再生數估計。利用每日流感疫情模擬資料，來比較傳統依賴於傳染區間的指數成長率估計法 ( Exponential Growth Rate Estimation, EGRE ) 及可估計疾病基礎再生數與傳染區間的最大概似估計法 ( Maximum Likelihood Estimation, MLE ) ，所得結果進行比較及解釋。此外，實際資料部份，資料來源為行政院衛生署2009年兩百萬人抽樣檔之健保門診及住院明細檔，篩選首次因肺炎或流感( P&I )及首次因流感( Flu )而就診/住院之病患資料，進行基礎再生數估計。
為了符合EGRE對於疫情資料的限制，只分析疫情初期階段資料。模擬資料結果顯示，當傳染區間已知，除了EGRE在R0設定為0.9的模擬資料中有略為高估的情形外，其他參數組合，兩種方法估計準確度皆相當良好。當錯誤假設傳染區間，兩種方法R0估計結果皆會產生偏差。在傳染區間未知時，整體而言，MLE估計R0的準確度皆相當良好;而EGRE則在R0設定為0.9及2的模擬資料中，R0的估計會產生較為嚴重的估計偏差。實際資料估計部份，在第一波及第二波P&I資料，使用MLE所估計的R0分別為1.10及1.05，平均傳染區間分別為1.08天及1.80天;利用EGRE所估計的R0分別為1.15及1.08。在Flu資料，使用MLE所估計之R0為1.06，平均傳染區間為0.94天;EGRE所估計的R0則為1.07。
由於傳染病本身的複雜性，因此對於相同類型的傳染病，可能因收集資料方式或估計方法不同而有不一樣的結果。因此本研究建議在進行傳染病參數估計時，宜採用多種估計模型進行比較，且作一有效的組合運用，以便獲得更多疾病訊息，提高對傳染疾病的了解。
In recent years, infectious diseases spread rapidly due to many emerging infectious diseases and convenient transportations. Therefore, it has become well understand that the epidemiology of infectious disease is important to reduce disease transmissions. One of the primary preventive strategies for infectious diseases is to reduce the basic reproduction numbers of pathogens effectively. The basic reproduction number, R0, is defined as the average number of secondary cases produced by a single infected in a large population of susceptible individuals. The R0 provides a measure to quantify the transmissibility of infectious disease. The value of R0 indicates the strength of preventive interventions necessary for control. Therefore, estimating the basic reproduction number has become one of the primary topics in epidemiology.
In this thesis, we focus on estimations of the basic reproduction number, R0, in the influenza outbreak. Influenza is an acute viral respiratory illness, stemmed from infected with influenza virus. Importance of influenza outbreak control is to prevent the spread of outbreak and the serious complications, which may cause death. We compared traditional exponential growth rate estimation (EGRE) with the maximum likelihood estimation (MLE) through a variety of simulations that are based on different scenarios of daily influenza infection. In addition, we compared the estimations of two methods using an real data from two million people sampling data of the 2009 Taiwan Department of Health which contains outpatients and inpatients files. We selected the patient data from pneumonia or influenza in the first time.
In order to meet the limit of EGRE for epidemic data, we limit our analysis of these data to the initial portion of the epidemic. Simulation studies show that two methods perform well under comparisons when the serial interval is known. However, EGRE yields overestimation when the true R0 is close to 0.9. We also found that two methods may result in inaccurate estimates of R0 if the serial interval is miss-specified. Overall, the estimations of R0 obtained from the MLE method are more accurate than the EGRE method when the serial interval is unknown. In the real-time estimation of P&I outbreaks data, the R0 estimates of MLE method in the first and second phase are 1.10 and 1.05, as well as the estimated mean of serial interval are 1.08 and 1.80 days, respectively. For EGRE method, the R0 estimates in the first and second phase are 1.15 and 1.08, respectively. In the Flu data, the R0 estimate and the mean of serial interval using MLE method are 1.06 and 0.94 day, respectively. The R0 estimate by EGRE method is 1.07.
Due to the complexity of the infectious disease, methods appropriately considering the underlying distribution of infectious diseases data are crucial in obtaining meaningful estimations of R0. Given this reason, we suggest that the estimations of parameters of infectious disease models should use a variety of estimation models to get more diseases information. |
