September Problem 4: 12 month Evaluation. Scenario 3 Q1a-d (Model extraction, grounding, and unit tests for SEIR-V model with wastewater)

[djinnome]

Evaluation Scenario 3: Modeling Approaches to Address Underreporting with Wastewater

Background:

Official COVID-19 case counts have dramatically underestimated the true number of infections since the initial documented cases in the United States in early 2020, due to supply constraints on testing and variations in test-seeking behavior. The underreporting of COVID-19 cases due to these factors also varies over time, meaning that simply scaling the observed cases to estimate actual infections would yield inaccurate results. Wastewater-based surveillance is a promising tool to estimate the actual, rather than recorded cases in a population, because the concentration of SARS-CoV-2 in the water is not affected by testing supply or test-seeking behavior. This approach is emerging, and researchers have had mixed success in predicting cases based on wastewater signals. Integrating wastewater-based surveillance into phenomenological models is an area of active inquiry but is challenging due to the need to map new concepts and data into existing frameworks (e.g., SIR).

Timepoint: October 2020 - January 2021
Location: Greater Boston area; optional extension to NYC context

Model: SEIR-V model (Phan et al.) https://doi.org/10.1016/j.scitotenv.2022.159326
Data: Derived from supplementary materials in paper

Question 1. You want to utilize this SEIR-V model to estimate the true number of underlying infections (as opposed to officially reported cases) based on wastewater data.

a. (TA1 Model Extraction Workflow, Data Workflow) Extract/replicate the system of equations for the SEIR-V model. Extract the appropriate data columns (cRNA2 and F2, which represent the viral load in wastewater and the flow rate, respectively) from the paper’s supplemental materials.
b. (TA2 Domain Knowledge Grounding; ASKEM Workbench Only) We want to ensure that terms from the model are grounded appropriately given that it involves several nontraditional concepts. For example, typically, “V” in the compartmental modeling framework represents a vaccination compartment. This model retains many concepts of the traditional compartmental modeling framework, but “V” here represents a novel concept of cumulative viral load in wastewater. Demonstrate that state variables and parameters are grounded appropriately to their descriptions in the paper, including through manual adjustment in the workbench as necessary.
c. (TA3 Simulation Workflow / Unit Test) Demonstrate that the extracted model from 1a maintains fidelity to the original model by replicating the fitting exercise in the publication’s Section 3.2 (visual output available in Figure 2A of the paper), using wastewater viral concentration data (available in the supplementary materials). You may simplify wherever necessary, such as by fixing the value of β to the fecal viral shedding rate implemented in the paper (4.49 * 10^7 viral RNA/g). Use assumptions from Table 1 in the paper to inform parameter values that are derived from literature (i.e., “exposed duration”, “infectious duration”) and fit λ, α, and E(0) to the wastewater data from October 02, 2020 to December 18, 2020.

<style> </style>

	Problem 1a	Problem 1b (ASKEM Workbench Only)	Problem 1c-f
Inputs	Paper and supplementary materials (code + data)	Extracted model from 1a	Model from 1b
Task	Extract SEIR-V model, metadata, concepts from paper	Ingest, inspect, and update domain knowledge groundings	Run simulation to include calibration and forecasting
Outputs	SEIR-V system of equations, replicated in form that can be implemented in workbench	Demonstration of groundings (state variables, parameters) mapped correctly to associated concepts from paper	Plots that roughly replicate Figure 2 of the paper