茶室網絡在2021年四月至七月雙北新冠疫情傳播之角色：代理人模式分析

Abstract

背景與研究目標：2021年四月至七月雙北爆發一波大規模新冠疫情。疫調顯示此波疫情可能與不戴口罩的茶室接觸網絡有關。然而，現有COVID-19代理人建模研究以日常接觸為主，並未探討傳播風險存在顯著差異時，高低風險複合網絡情境下COVID-19的傳播模式，也尚未有針對特殊服務業－如萬華茶室－這類高風險接觸網絡的討論。而這樣假設傳播風險無顯著差異的建模結果，會導向支持在COVID-19疫情期間應減少所有類型的社區中人與人的互動與接觸，進而支持封城。相反地，若口罩確實有降低COVID-19傳播的效果，存在高低風險複合網絡的前提下，則防疫時僅需針對不戴口罩的茶室等高風險網絡進行控管，即可有效控制疫情，根本不需要封城。這顯示COVID-19代理人建模研究中是否需考慮高低風險複合網絡，對公共衛生上的防疫政策有重大影響。因此，本研究進行全球第一個考量高低風險複合網絡COVID-19代理人建模研究，建立一納入茶室網絡結構之代理人模型，模擬2021年四月至七月雙北新冠疫情之流行過程，並期望能以此理解茶室網絡於本土疫情傳播之角色與重要性，同時側面證實外科口罩防護之有效性。

方法：本研究於代理人模型中建立高低風險複合網絡結構，模擬2021年5月在雙北本土爆發的疫情。本研究將臺北市萬華區、新北市板橋區、中和區、三重區等四行政區作為研究區，考量各行政區的實際人口數、年齡性別結構等人口學特徵，使用三種接觸網絡結構：茶室、工作場所與家戶網絡，來表示模型中人與人的接觸型態，作為SARS-CoV-2病毒在代理人之間的傳播方式。傳染病數理模式SEIQRS模型則表示每位代理人於模型中隨時間改變的疾病狀態。模型中假設口罩防護有效，且口罩的配戴行為差異將使得茶室網絡的傳播風險顯著高於其他網絡：茶室網絡為高風險接觸，工作場所與家戶網絡為低風險接觸，以此進行模擬，估計2021年雙北本土疫情下的每日通報病例數、Rt值與血清盛行率，並使用最小平方法對每日通報病例數進行擬合。透過分層分析，本研究進一步說明疫情期間不同階段SARS-CoV-2病毒的主要傳播對象與傳播途徑。而經過敏感度分析，本研究可驗證工作與茶室網絡每日接觸者數的變動是否影響模擬結果，以及代理人模型的最佳運行次數。

結果：根據敏感度分析，本研究將20次定為最佳運行次數進行擬合。首先，代理人模型模擬結果與實際流行曲線趨勢相符。分層分析結果顯示累積至7 月31日為止，確診案例將高度集中於茶室出入者和一般勞工。其中一般勞工佔55.412%，茶室出入者則佔37.371%。儘管過半確診者為一般勞工，然而在茶室出入者中卻有高達42.1% 的人口確診：3,400名茶室出入者中便有1,450人確診，遠高於其他族群的確診率。結果亦顯示，茶室網絡Rt值高峰為4.2，且在三級警戒之前均高於1，也高於工作場所網絡和家戶網絡，因此茶室網絡不僅是疫情初期的主要傳播途徑，也使得本土疫情進入流行階段。工作場所與家戶網絡由於Rt值持續低於1，因此即便為疫情中後期的主要傳播途徑，也並不會造成第二波的本土流行。上述結果皆顯示萬華茶室與2021年5月爆發之雙北本土疫情之間存在緊密關聯，並且高風險的茶室網絡具重要的傳播作用。綜上所述，本研究的代理人模型結果可驗證2021年雙北本土疫情中存在高低風險複合網絡，即茶室網絡的傳播風險應顯著高於工作場所與家戶網絡。

結論：不戴口罩的高風險茶室接觸網絡在2021年雙北本土新冠疫情傳播過程中扮演關鍵角色。本研究結果不僅支持口罩確實有降低COVID-19傳播的效果，亦支持在COVID-19疫情期間不需要去控管所有類型的人際接觸，僅需針對不戴口罩的茶室等高風險網絡進行控管，即可有效控制疫情。

Background and Research Objective: Between April and July 2021, a significant COVID-19 outbreak occurred in Taipei and New Taipei City, linked to mask-free teahouse contact networks. Existing COVID-19 agent-based modeling studies mainly focus on daily contacts and do not consider the varying probabilities of transmission in composite networks of high and low risk. This oversight supports broad interaction reductions and lockdowns. However, if masks effectively reduce transmission, managing only high-risk networks, such as teahouses, could control the outbreak without a full lockdown. This study pioneers incorporating a composite high and low-risk network into COVID-19 agent-based models, simulating the outbreak in Taipei Metropolitan Area, and aims to understand the role of teahouse networks and the effectiveness of surgical masks.

Methods: This study develops a composite network structure of high and low risk within an agent-based model to simulate the May 2021 COVID-19 outbreak in the Taipei Metropolitan Area. Focusing on Wanhua District in Taipei City and Banqiao, Zhonghe, and Sanchong Districts in New Taipei City, it incorporates demographic characteristics such as population size and age-gender structure. The model includes teahouse, workplace, and household contact networks to represent social relationships and SARS-CoV-2 transmission among agents. Using the SEIQRS epidemiological model, it tracks the changing disease states of agents over time. Assuming effective mask protection, it differentiates the probability of transmission, with teahouses as high-risk networks and workplaces and households as low-risk networks. Simulations estimate daily new diagnoses, Rt values, and seroprevalence, fitting daily cases using the least squares method. Stratified analysis identifies primary infected populations and transmission pathways at various outbreak stages. Sensitivity analysis assesses how changes in daily contacts in workplace and teahouse networks impact results and determines the optimal number of runs for the model.

Results: The agent-based model results aligned with the actual epidemic curve with values of the 20-time average. Stratified analyses showed that by July 31, confirmed cases were highly concentrated among teahouse visitors and regular workers, with regular workers accounting for 55.412% and teahouse visitors for 37.371% of the cases. Despite the majority of cases being among workers, a striking 42.1% of teahouse visitors were infected, with 1,450 out of 3,400, a rate much higher than other groups. The results also indicated that the Rt value in teahouse networks peaked at 4.2 and remained above 1 until the Level 3 alert, higher than the Rt values in workplace and household networks. Thus, the teahouse network was not only the primary transmission pathway in the early stages but also contributed to the outbreak entering the epidemic phase. In contrast, workplace and household networks, with Rt values consistently below 1, did not cause a second wave despite being main transmission pathways in the later stages. These findings demonstrate a close association between the Wanhua teahouses and the local outbreak in the Taipei Metropolitan Area in May 2021, highlighting the significant role of the high-risk teahouse network.

Conclusion: During the 2021 COVID-19 outbreak in the Taipei Metropolitan Area, mask-less, high-risk teahouse networks played a crucial role in transmission. This study supports the effectiveness of masks in reducing COVID-19 spread and suggests that controlling high-risk settings—rather than all interpersonal contacts—is sufficient for effective epidemic management.