大流行期間台北環境職業衛生研究：COVID-19之環境監測與預測模式

Abstract

研究目的：環境監測已成為補充傳統疫情個案通報系統的關鍵工具，特別是在面對無症狀與氣膠傳播風險的情境下。本論文旨在發展環境監測方法與預測模型，以評估並預測COVID-19在臺北市的傳播風險。研究包含三個相互關聯的子題，具體目標如下：（1）透過持續性的室內空氣品質監測，評估幼兒園中SARS-CoV-2的空氣傳播風險；（2）於大學附設醫院建立機構層級的污水監測系統，用以早期偵測並預測院內及社區的疫情趨勢；（3）建立臺北市12個行政區的全市性污水病毒監測與疫情預測模型。

研究方法：第一項研究於2021年8月至11月期間，在一所幼兒園內持續監測教室與辦公室的二氧化碳（CO₂）濃度，並利用Wells-Riley模型，結合CO₂濃度、使用人數與停留時間等參數推估基本再生數（R₀）。第二項研究於2022年4月至10月，在一所醫院的七個污水井進行每週兩次的污水採樣，採用病毒RNA直接捕捉法進行預處理與濃縮，並以RT-qPCR定量分析。透過迴歸模型預測醫院、周邊社區及全市的COVID-19病例數移動平均。第三項研究將監測擴展至整個臺北市，自2022年5月至8月共90天，於信義與內湖兩行政區進行每日採樣，其餘10區則每週採樣兩次。迴歸模型以病毒相對訊號預測疫情，並透過行政區通報數據進行模型驗證。

研究結果：第一項研究顯示，教室在上課期間的室內CO₂濃度較戶外高出超過400 ppm，且30人教室的R₀估值介於3.01至3.12之間。隨著一週中與每日進程的推移，感染風險逐漸上升，顯示氣膠累積現象。情境分析指出，結合減少室內人數、停留時間與CO₂濃度方能降低感染風險。第二項研究中，即便當地確診數極低，仍能檢出SARS-CoV-2病毒，顯示污水檢測具有高度敏感性。相對病毒訊號與未來疫情指標呈現高度相關，可有效預測醫院、社區與城市層級的趨勢。第三項研究中，最佳對數迴歸模型可解釋未來五日移動平均新增病例變異量的78%。病毒相對訊號每增加1%，新發病例數增加約0.27%。模型驗證顯示10個行政區中預測值與實際通報病例無顯著差異。

研究結論：本研究證實，透過室內CO₂監測與污水病毒量化的環境監測方式，可提供敏感、非侵入性且具可擴展性的工具，有效評估感染風險並預測疫情趨勢，特別有助於發現無症狀個案及低度通報的疫情。政策建議包括：強制於教育場所設置CO₂監測系統、投資改善通風系統，以及推動全國機構與社區層級的污水病毒監測計畫。綜合三項研究成果，本研究呈現一套整合性架構，將環境監測納入防疫規劃與應變策略核心。

Objectives: Environmental surveillance has emerged as a vital tool to supplement traditional case-based surveillance of epidemic, particularly in the context of asymptomatic aerosol transmission. This dissertation aims to develop environmental monitoring approaches and predictive models to assess and forecast COVID-19 transmission risks in Taipei, Taiwan. The research consists of three interrelated studies with the following objectives: (1) to estimate the airborne infection risk of SARS-CoV-2 in kindergartens using continuous indoor air quality monitoring; (2) to establish an institutional wastewater surveillance system for early detection and forecasting of COVID-19 trends within and beyond a university hospital; and (3) to develop a citywide wastewater-based epidemic prediction model across 12 districts of Taipei.

Methods: The first study employed on-site continuous monitoring of carbon dioxide (CO₂) in kindergarten classrooms and staff offices between August and November 2021. The Wells-Riley model was applied to estimate the basic reproduction number (R₀) using data on CO₂ concentrations, occupant numbers, and duration of stay. The second study involved twice-weekly wastewater sampling at seven hospital manholes from April to October 2022. SARS-CoV-2 RNA was extracted using a direct viral RNA capture method and quantified by RT-qPCR. Regression models were built to forecast moving averages of COVID-19 cases at the hospital, community, and city levels. The third study extended this approach to the entire Taipei City, with daily wastewater sampling in two core districts and biweekly sampling in ten others over 90 days. Regression-based epidemic prediction models were trained using relative viral signals and validated using district-level epidemiological data.

Results: In the first study, the indoor CO₂ concentrations during school hours exceeded outdoor levels by over 400 ppm, with corresponding R₀ estimates ranging from 3.01 to 3.12 in classrooms with 30 occupants. Scenario analyses showed that reducing occupancy, stay duration, and CO₂ levels together could decrease R₀. In the second study, SARS-CoV-2 RNA was detected even during low-incidence periods, indicating high sensitivity of the wastewater testing method. Relative viral signals showed strong associations with future COVID-19 trends across the hospital, community, and city levels. In the third study, the best-fitting log-log model explained 78% of the variance in the future 5-day moving average cases. A 1% increase in wastewater viral signal was associated with a 0.27% increase in new COVID-19 cases. Validation showed no significant difference between forecasted and reported epidemic indicators in 10 districts.

Conclusions: The findings demonstrate that environmental surveillance using indoor CO₂ monitoring and wastewater viral quantification can provide sensitive, non-invasive, and scalable methods for assessing infection risk and forecasting epidemic trends. These methods are especially valuable for detecting underreported or asymptomatic transmission. Policy implications include the need for mandatory CO₂ monitoring in educational facilities, investments in ventilation improvements, and nationwide institutional and community wastewater surveillance programs. Together, the three studies offer a cohesive framework for integrating environmental monitoring into pandemic preparedness and response planning.