重新檢視 COVID-19 邊境管制政策：動態數理模式研究

Abstract

背景：2019年底 2019 新型冠狀病毒疾病 (Coronavirus disease 2019, COVID-19) 疫情爆發後，COVID-19疫情迅速蔓延全球，對公共衛生體系及國際經濟貿易往來造成重大衝擊。為減緩病毒輸入風險與境內傳播壓力，各國實施不同程度的邊境管制措施以遏制COVID-19的擴散。尤其在疫情初期，疫苗尚未問世、醫療資源有限的情況下，邊境檢疫政策成為延緩社區流行、爭取防疫準備時間的關鍵措施。然而，入境檢疫日數與社會成本間取捨兩難。在難以無限期管制邊境出入的前提下，如何在疫情初期建立一套兼具防堵疫情效果與經濟可行性的邊境管理策略，成為關鍵政策議題。本研究呼應流行病預防創新聯盟 (CEPI) 提出的「百日疫苗任務」(100 Days Mission) 理念，亦即在新興傳染病出現後的關鍵100天內，透過科學建模輔助決策，在疫苗問世前的空窗期延緩疫情爆發降低死亡與公衛醫療崩潰風險，維持社會與經濟秩序的穩定。因此，本研究以COVID-19疫情初期為背景，運用動態數理模型模擬不同邊境政策對本土疫情之影響，旨在為未來面對新興傳染病時，提供可迅速部署並具防疫效益的邊境應變依據，提供具實證基礎之政策建議。 方法：本研究採用SEIR動態流行病學模型，評估不同邊境檢疫策略組合和國內控制措施對國內COVID-19疫情控制的影響。我們模擬不同檢疫日數 (0、3、5、7及14日) 、入境人數 (每日1,500、3,000、6,000、10,000、20,000、30,000) 及不同傳播力參數 (包含不同的傳染率、住院率及無症狀比例) 對社區病毒傳播的影響。並納入社區防疫措施，包括口罩佩戴 (高配戴率90%；低配戴率77%) 及其防護效力、接觸者追蹤率、有症狀者就醫率等。並模擬三個情境 (base scenario、high carriage scenario 以及 Omicron scenario) ，以真實地模擬在不同政策組合下的疫情發展曲線，再利用4項政策可行性指標 (每日確診個案數、ICU人數、住院人數以及隔離人數) 對情境是否能維持在低度流行狀態，進行量化評估。本研究也以敏感度分析 (sensitivity analysis) 檢視兩項無法直接觀測的關鍵參數–接觸者追蹤率與感染者就醫率–的不確定性對模型模擬結果的影響。 結果：模擬結果顯示：在基本情境下，即便每日入境人數達 30,000 人，若能夠維持5天入境檢疫配合解隔時採檢，仍可將國內疫情4項指標維持於低度流行水準。然而，若入境者帶原率升高至 5% 或是 Omicron 傳播力更強的情境，每日入境人數需降至 3,000 人以下，方能控制境內疫情不爆發；若口罩配戴率下降至 77% ，臨界值進一步降至 1,500 人。敏感度分析顯示若接觸者追蹤率或症狀者就醫率降至 50%，則隔離人數將迅速突破 3,000 人之安全閾值，成為首波超載指標，顯示公衛系統在社區防疫效率下降時將最先承受壓力。為驗證模型真實性，本研究將模擬結果與 2020 年 3 月至 6 月間台灣實際數據進行擬合，結果顯示 14 天、7 天與 5 天檢疫情境皆高度符合實際數據，驗證模型具預測效能。分析 Omicron 期間台灣每日實際入境人數與確診數變化，發現自 2022 年 3 月 7 日起每日入境人數突破本研究分析顯示的臨界值1,500 人後，國內確診人數迅速上升，最終導致 2022 年境內疫情大爆發，與本研究模型之預測一致。 討論：本研究透過數理模型模擬與實證資料分析，首度明確指出每日入境人數為以往邊境政策研究中被忽略的關鍵變數。相較以往僅著重檢疫天數之政策評估模式，本研究以整合檢疫日數及入境量能的分析架構，估計入境人數臨界值，提供防疫政策具體且操作性高之建議。擬合2020年台灣真實數據亦驗證本研究使用模型之準確性。而2022年Omicron流行期間邊境管制鬆綁與後續疫情發展驗證當入境人數突破1,500人後，本土疫情迅速升高，顯示境內防疫量能有限的情境下，邊境政策鬆綁確實就是疫情失控主因。在面對高傳染力變異株時，入境檢疫天數與每日入境人數均必須嚴格管控，才能兼顧防疫與經濟。本研究結論為：在疫苗出現前，臺灣 2020-2021年間防疫成功之關鍵是：以邊境檢疫措施管控入境人數，防止境內防疫量能崩潰，成功地達到防疫需求與經貿需求之間的平衡。此每日入境人數須保持在臨界值以下的原則，可提供未來在類似情境時的重要決策參考。

Background: The COVID-19 pandemic highlighted the challenge of balancing border controls and socio-economic costs before vaccines. Echoing CEPI's "100 Days Mission," this study uses dynamic mathematical modeling to simulate early COVID-19 border policies. Our goal is to provide evidence-based guidance for effective and rapidly deployable border management strategies during the initial 100 days of future emerging infectious diseases, aiming to mitigate outbreaks, protect healthcare, and ensure stability. Methods: This study employs an SEIR dynamic epidemiological model to evaluate the impact of various border quarantine strategy combinations and domestic control measures on COVID-19 epidemic control. We simulated different quarantine durations (0, 3, 5, 7, and 14 days), inbound traveler numbers (1,500, 3,000, 6,000, 10,000, 20,000, and 30,000 per day), and different transmissibility parameters (including varying infection rates, hospitalization rates, and asymptomatic proportions) on community virus transmission. Concurrently, we incorporated community prevention measures, including mask-wearing rates (high compliance set at 90%; low compliance at 77%) and their protective efficacy, contact tracing ratios, and the rate of symptomatic individuals seeking medical attention and reporting. We also simulated three distinct scenarios: the Base Scenario, a High Carriage Rate scenario, and the Omicron scenario. This allowed us to more realistically model the epidemic's progression under different policy combinations. Subsequently, we conducted a quantitative assessment of these scenarios using four policy feasibility indicators: daily confirmed cases, ICU patient numbers, hospitalized patient numbers, and isolated individuals. To examine the model's stability and evaluate the impact of key policy parameters on the results, this study conducted a sensitivity analysis. We focused on two crucial sources of uncertainty: contact tracing ratio and infected individual isolation ratio. The contact tracing ratio was simulated from a high-efficiency scenario (90%) down to an overloaded epidemic scenario (25%), to reflect the actual availability of public health investigation resources as the epidemic evolves. The infected individual isolation ratio was set at two parameter groups (50% and 90%), to assess the impact of symptomatic individuals actively seeking medical attention and reporting behavior on epidemic control. Results: Simulation results indicated that under the baseline scenario, even with daily inbound traveler numbers reaching 30,000, maintaining a 5-day quarantine period could keep all four key indicators of domestic epidemic burden within a low-transmission threshold. However, if the carriage rate among inbound travelers increased to 5% or in a scenario reflecting Omicron’s higher transmissibility, the number of daily inbound travelers would need to be reduced to below 3,000 to keep the epidemic under control. Furthermore, in a scenario where mask-wearing compliance dropped to 77%, this threshold would need to be lowered further to 1,500 per day. Sensitivity analysis revealed that if either the contact tracing rate or the isolation rate of symptomatic cases dropped to 50%, the number of individuals requiring isolation would rapidly exceed the safety threshold of 3,000, becoming the earliest indicator of system overload. This highlights that the public health system would bear the initial burden when community-based control efficiency declines. To validate the reliability of the model, the simulation results were fitted to Taiwan’s actual confirmed case data between March and June 2020. The model accurately reproduced the observed trends under 14-day, 7-day, and 5-day quarantine scenarios, demonstrating strong predictive performance. Finally, empirical analysis of Taiwan’s inbound traveler data and confirmed case numbers during the Omicron period showed a sharp increase in domestic cases shortly after daily arrivals exceeded 1,500 on March 7, 2022, suggesting a strong temporal correlation between increased inbound volume and the subsequent outbreak. Discussion: This study developed a model identifying daily inbound traveler numbers as a core border policy variable, integrating it with quarantine duration to quantify critical thresholds. Validated by Taiwan's 2020 data, the model reflects epidemic dynamics. Observations from the 2022 Omicron outbreak highlight that exceeding 1,500 daily inbound travelers without simultaneous capacity adjustments causes rapid escalation. We emphasize that for highly infectious variants, quarantine duration and inbound numbers must be controlled concurrently. Our findings offer evidence-based recommendations for future pandemic preparedness and border control.