Department of Mathematical and Computer Sciences, Indiana University of Pennsylvania, Indiana, PA, USA.
Department of Wastewater Treatment, Borough of Indiana, Indiana, PA, USA.
Sci Total Environ. 2021 Sep 10;786:147451. doi: 10.1016/j.scitotenv.2021.147451. Epub 2021 Apr 30.
The building of an effective wastewater-based epidemiological model that can translate SARS-CoV-2 concentrations in wastewater to the prevalence of virus shedders within a community is a significant challenge for wastewater surveillance. The objectives of this study were to investigate the association between SARS-CoV-2 wastewater concentrations and the COVID-19 cases at the community-level and to assess how SARS-CoV-2 wastewater concentrations should be integrated into a wastewater-based epidemiological statistical model that can provide reliable forecasts for the number of COVID-19 infections and the evolution over time as well. Weekly variations on the SARS-CoV-2 wastewater concentrations and COVID-19 cases from April 29, 2020 through February 17, 2021 were obtained in Borough of Indiana, PA. Vector autoregression (VAR) model with different data forms were fitted on this data from April 29, 2020 through January 27, 2021, and the performance in three weeks ahead forecasting (February 3, 10, and 17) were compared with measures of Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). A stationary block bootstrapping VAR method was also presented to reduce the variability in the forecasting values. Our results demonstrate that VAR(1) estimated with the logged data has the best interpretation of the data, but a VAR(1) estimated with the original data has a stronger forecasting ability. The forecast accuracy, measured by MAPE, for 1 week, 2 weeks, and 3 weeks in the future can be as low as 11.85%, 8.97% and 21.57%. The forecasting performance of the model on a short time span is unfortunately not very impressive. Also, a single increase in the SARS-CoV-2 concentration can impact the COVID-19 cases in an inverted-U shape pattern with the maximum impact occur in the third week after. The flexibility of this approach and easy-to-follow explanations are suitable for many different locations where the wastewater surveillance system has been implemented.
建立一个有效的基于污水的流行病学模型，能够将污水中的 SARS-CoV-2 浓度转化为社区内病毒排放者的流行率，这是污水监测的一个重大挑战。本研究的目的是调查 SARS-CoV-2 污水浓度与社区层面 COVID-19 病例之间的关系，并评估应如何将 SARS-CoV-2 污水浓度纳入基于污水的流行病学统计模型，以提供对 COVID-19 感染数量及其随时间演变的可靠预测。从 2020 年 4 月 29 日到 2021 年 2 月 17 日，在宾夕法尼亚州印第安纳县获得了 SARS-CoV-2 污水浓度和 COVID-19 病例的每周变化情况。对 2020 年 4 月 29 日至 2021 年 1 月 27 日期间的数据，采用不同数据形式的向量自回归（VAR）模型进行拟合，并比较了均方误差（MAE）和平均绝对百分比误差（MAPE）的三周前预测（2 月 3 日、10 日和 17 日）的性能。还提出了一种固定块 bootstrap VAR 方法，以减少预测值的可变性。我们的结果表明，用对数数据估计的 VAR(1) 对数据的解释最好，但用原始数据估计的 VAR(1) 具有更强的预测能力。1 周、2 周和 3 周未来的预测精度（以 MAPE 衡量）可低至 11.85%、8.97%和 21.57%。该模型在短时间跨度内的预测性能并不令人印象深刻。此外，SARS-CoV-2 浓度的单次增加会以倒 U 形模式对 COVID-19 病例产生影响，最大影响发生在第三周之后。这种方法的灵活性和易于理解的解释适用于许多已经实施污水监测系统的不同地点。
