Modeling the Dynamics of COVID-19 Epidemic using Q-SEIR Model with Age-stratified and Non-linear Incidence Rates

School Reopening Simulations with COVID-19 Agent-based Model for the Philippine Regions

We simulated 100% school reopening scenarios with varying vaccination coverages (0%, 25%, 50%, 75%, 100%) to the 17 regions of the Philippines and computed the change in deaths and infections with respect to the control scenario (no school opening). The School Reopening Viability (SRV) metric was then computed from the difference between the required vaccination coverage to get a downtick in deaths/infections and the current vaccination coverage. The following maps show if the regions may already reopen its schools fully from the results of our simulations.

We tested using 2 different dates, 17 January 2022 and 15 September 2021, which are the peaks of Omicron and Delta variant transmissions, respectively. The peaks were chosen to assume the worst case scenario representation

SRV shows if a region has already attained the required vaccination coverage for downticks in deaths and infections to occur. A positive score means that a region may reopen its schools fully; and, negative if otherwise. Hover over the regions in the maps below to see their corresponding scores.

[UPDATED] School Reopening Viability as of 17 January 2022 (Peak of the COVID-19 Omicron Variant Cases)

The SRV values are updated with the vaccination coverage as of April 19. In terms of Deaths, CARAGA still have a negative SRV but is considerably small (-0.93). In terms of infections, only Central Visayas and BARMM are among the regions of concern.

School Reopening Viability as of 17 January 2022 (Peak of the COVID-19 Omicron Variant Cases)

School Reopening Viability as of 15 September 2021 (Peak of the COVID-19 Delta Variant Cases)

Read the Executive Summary of the Interns' Findings on School Reopening Simulations

executive_summary_smsl_internship_2021_ver2.pdf

This research page is owned and operated by the System Modelng and Simulation Laboratory of the Department of Computer Science, UP Diliman. Initial fundng came from the College of Engineering, under the "COVID-19 Engineering Projects", as a public service to the Filipino people.

We are offering LGUs free analyses using tools developed at the System Modeling and Simulation Laboratory, in collaboration with Dr. Salvador E. Caoili, Balik-Scientist Romulo de Castro, Ph.D., Prof. Roselle K. Rivera and Dr. Jesus Emmanuel Sevilleja. To refer to data, analyses or results herein, please cite the following:

COVID-19 Agent-Based Model: An Epidemiological Simulator Applied in Vaccination Scenarios for Quezon City, Philippines

Finalist at the Undergraduate Research Competition, College of Engineering, June 2021

Published November 2021

School Re-opening Simulations with COVID-19 Agent-Based Model for Quezon City, Philippines

Awarded a "Best Oral Presenter Award" at PhilGeos 2021, November 2021

Published November 2021

Age-stratified Infection Probabilities Combined with Quarantine-modified SEIR Model in the Needs Assessments for COVID-19

Published May 2021 (Submitted April 2020)

COVID-19 Agent-Based Model with Multi-objective Optimization for Vaccine Distribution

Protection after Quarantine: Insights from a Q-SEIR Model with Nonlinear Incidence Rates Applied to COVID-19

Modeling the Dynamics of COVID-19 Using Q-SEIR model with Age-stratified Infection Probability

We have already been asked if our tools can be used at the barangay level. Definitely yes! We just need to ask them a few questions on Covid-19 incidence and quarantine conditions. Our colleagues at the Center for Informatics, University of San Agustin use FluSurge and the Hill model to predict the number of hospital beds, testing kits, PPEs needed, etc. They need a bit more data, and their analysis is better suited for the town and municipality level. You may email bongolan@up.edu.ph, or any of the modelers, for specific requests.

The authors claim proprietary rights to: Age-Strat, Q-SEIR, Age-stratified Q-SEIR and Q-SEIR with Nonlinear Incidence Rates and other methods developed by them, either independently or from sources cited in the paper(s). Please contact vabongolan@up.edu.ph to request a demonstration.

LATEST UPDATES

The Covid-19 effort continues, and we are currently exploring vaccine allocation methods, even as we wait for a reliable vaccine to be developed. We are preparing for the situation that the vaccine will be available in limited numbers, and at a cost to our country and people.

We plan to have an agent-based version of our current age-stratified, quarantine-modified SEIR model with non-linear infection rates. This model attempts to take into account our young population (an advantage in Covid-19), the different quarantine regimes, as well as human behavior! Per our models, our behavior (susceptibles and infectives) could protect us post-quarantine!

With a graduate student, we will be exploring other models aside from compartmental, and may include stochastics.

Latest results sent to the QC Mayor's office are below.

The obvious limitation of the classic SEIR model is its assumption of a homogenous population. This is where age stratification comes in, which, as a probability game, is clearly in the Filipino's advantage! Early this year, initial reports from Wuhan identified elderly males with pre-existing conditions to be most vulnerable, with an average age of 73 for fatalities. With our median age at 25.7 years (even younger than the world's median of 29), using China's data on infections and fatalities as of February 11, 2020, we see that at least half of our population has only a little over 10% chance of getting infected! Fatalities are similarly expected to be lower. We modified the infection term using proxy values for infection probabilities per age stratus. Proxies came from two data sets: China, as of Feb. 11, 2020, and Quezon City, as of April 20, 2020, which turned out to give very close multipliers for the infection term!

As expected, the projection on peak infectives in Quezon City is now considerably lower (only 3.19% or 103,300). An unwelcome, but totally expected effect, however, is a delay in the peaking of infectives (to Oct. 4, 2020).

Non-linear incidence rate actually answers the question: can we retain the benefits of community qurantine even after it is lifted? And the answer is yes! It will require susceptibles to maintain social distancing, and awareness of their fellows; infectives have to be disciplined about self-isolation and wearing masks. This model also has implications on tolerance to the sensitivity of tests for Covid! Intriguingly, this model predicts it is safe to lift the quarantine in Iloilo (which has few cases) by April 30, but fears a surge of infections in Quezon City if the quarantine is lifted April 30.

BACKGROUND

The SEIR model[1] is a commonly used model to estimate the number of individuals susceptible (S), exposed (E), infectious (I) and removed (R) during an epidemic. Note that removed individuals consist of recovered patients which are now immune and not contagious, as well as expired individuals who did not make it. A modification to this model includes a quarantine factor Q that can vary in time; we call this method Q-SEIR. Q(t) = 1.0 means no quarantine and Q(t) = 0.4 means that the quarantine is 60% effective in reducing rate of growth of exposed individuals. However, SEIR assumes a homogeneous population: i.e., everyone has similar chances of contracting the disease and recovering/expiring from it.

A probabilistic game, which we call Age-Strat, was created to consider the age of an individual subject to exposure. This uses data of COVID patients in China[2], specifically the proportions infected and removed, stratified by age groups. The stratification were treated as proxies for infection probability of an individual given his age.

SCOPE

This is intended to be used nationwide, with standing requests from Iloilo, and we plan to have a dry run for Metro Manila. Requesting LGUs shall fill-out an online form with their total population and how it stratifies by age; the tools will do the analysis and generate a report when will the epidemic reach its peak.

Age-Strat and Q-SEIR are only half of the 'toolkit' being used. FluSurge and the Hill Model (now called as Q-Hill) are maintaned at the University of San Agustin Center for Informatics, and data collection is at RedCap.

The curators will then prepare one report to the LGUs.

REFERENCES

[1] SEIR Model. http://www.public.asu.edu/~hnesse/classes/seir.html

[2] Chinese Center for Disease Control and Prevention (China CDC) - The Epidemiological Characteristics of an Outbreak of 2019 Novel Coronavirus Diseases (COVID-19) http://weekly.chinacdc.cn/en/article/id/e53946e2-c6c4-41e9-9a9b-fea8db1a8f51

CURATORS and CONSULTANTS

Roselle Leah K. Rivera, M.A., Sociologist

College of Social Work and Community Development

UP Diliman

Romulo de Castro, Ph.D., Medical Bioinformatician

Center for Informatics, University of San Agustin

DOST-PCHRD, Balik-Scientist Program

Jesus Emmanuel Sevilleja, M.D., MPhil, Epidemiologist

National Center for Mental Health

Salvador E. Caoili, MD, Ph.D., Immunoinformatician

UP College of Medicine

MODELING TEAM