key: cord-0996404-2p67ek49 authors: Mendoza, V. M. P.; Mendoza, R.; Ko, Y.; Lee, J.; Jung, E. title: Modeling behavior change and underreporting in the early phase of COVID-19 pandemic in Metro Manila, Philippines date: 2022-04-05 journal: nan DOI: 10.1101/2022.03.29.22273148 sha: 7777ef02b9b1feff4769f97af2ef3cdf4a792dda doc_id: 996404 cord_uid: 2p67ek49 When the Philippine government eased the community quarantine restrictions on June 2020, the healthcare system was overwhelmed by the surge in coronavirus disease 2019 (COVID-19) cases. In this study, we developed an SEIQR model considering behavior change and unreported cases to examine their impact on the COVID-19 case reports in Metro Manila during the early phase of the pandemic. We found that if behavior was changed one to four weeks earlier, then the cumulative number of cases can be reduced by up to 74% and the peak delayed by up to four weeks. Moreover, a two- or threefold increase in the reporting ratio can decrease the cumulative number of cases by 29% or 47%, respectively, at the end of September 2020. Results of our finding are expected to guide healthcare professionals to mitigate disease spread and minimize socioeconomic burden of strict lockdown policies during the start of an epidemic. In the Philippines, the first confirmed local transmission of the coronavirus dis-4 ease 2019 (COVID-19) was reported on March 7, 2020. To curb the infection, the 5 Philippine government implemented a four-level community quarantine protocol [1] . 6 The Philippine government's policy on community quarantines is one of the world's 7 strictest and longest lockdown policy. The policies focused on mobility restrictions 8 implemented by the military, while contact tracing and testing remained low com-9 pared to other countries [2] . The Philippine Department of Health reported on May 10 2020 that the country lacked 94,000 contact tracers to reach the ideal ratio of one 11 contact tracer per 800 people [3] . Further, daily testing output for the whole country Statistics Authority [36] . 75 Timeline of Community Quarantines in Metro Manila 76 We divide the timeline from March 8 to November 30, 2020 into three periods. We 77 use the data from the first two periods to estimate the rates of the transmission, 78 behavior changes, and reporting. On the third period, we compare model results 79 using the obtained parameter estimates to the observed data. Figure 1 summarizes 80 the community quarantines observed on the three periods. The first period is from 81 March 8 to May 30, 2020, which covers ECQ and a two-week MECQ. The second 82 period is from June 1 to September 30, 2020, wherein Metro Manila was mostly 83 under GCQ except for another two-week MECQ. It was during this period that a 84 major epidemic wave that overwhelmed the healthcare system occurred. The third 85 period is from October 1, 2020 to November 30, 2020 and was entirely under GCQ. Vaccines and antiviral therapy were not yet available during these periods and 87 hence, NPIs such as community quarantines, social distancing, and mandates on 88 wearing face masks were the only measures implemented by the government. susceptible with behavior changes (S F ), exposed (E), reported infectious (I), unre-93 ported infectious (I u ), isolated (Q), and recovered (R). The flow of the COVID-19 94 transmission is illustrated in Figure 2 . 95 We assume that the change in behavior of susceptible classes is influenced by the 96 S, Q, and R classes. As the number of confirmed and isolated individuals increases, 97 more individuals change their behavior and transfer from S to S F . On the other 98 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.03.29.22273148 doi: medRxiv preprint Figure 2 The flow diagram of a COVID-19 transmission model that incorporates unreported cases. Susceptible S may change behavior S F at rates β F or µ. These classes can be exposed E to the virus at a rate β and become confirmed infectious I or unreported infectious Iu in 1/κ days on average. The reduction of infections caused by behavior change is denoted by δ and the reporting ratio is denoted by ρ. The confirmed cases are isolated Q in 1/α days and recover 1/γ days on average. The average fatality rate is denoted by f . The unreported infectious class recovers 1/ν days on average. on average. The model is described by the following differential equations: The average number of days from symptom onset to case confirmation is set to Table 1 . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2022. We set the initial number of infectious individuals I 0 , exposed E 0 , and unreported We estimate the transmission rates in the first (β 1 , β F,1 ) and second (β 2 , β F,2 ) pe-121 riods, and the reporting ratio. Estimation was done by minimizing the squared 122 difference between the cumulative confirmed cases and the model at corresponding is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.03.29.22273148 doi: medRxiv preprint probability distribution structure. In our simulations, we sample 10,000 synthetic 138 data sets using Poisson distribution. The parameters are re-estimated from the 139 generated data sets. The mean, standard deviation, and confidence intervals of the 140 re-estimated parameters are calculated. According to this formula, the reproductive number of the susceptible group without We use the cumulative number of infected individuals κE as the model output 168 to consider every infection. In the implementation of LHS, we sampled 10,000 com-169 binations of parameters, all following a uniform distribution. We can see that the 170 parameter β is the most influential parameter, followed by the reporting ratio (ρ), . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 5, 2022. The results of the parameter bootstrapping are shown in Figure 6 . After 10,000 186 simulations, the re-estimated values of the parameters β 1 , β 2 , β F,1 , β F,2 , and ρ follow 187 a normal distribution. The obtained standard deviations are several degrees lower 188 than the mean estimates. Furthermore, the estimated parameters shown in Table 1 189 fall inside the 95% confidence interval of the estimates, and are close to the mean 190 estimates, suggesting good reliability of the estimated parameters. 191 Figure 6 Results of the parameter bootstrapping for β 1 , β 2 , β F,1 , β F,2 , and ρ with the mean, standard deviation, and 95% confidence interval. 197 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.03.29.22273148 doi: medRxiv preprint We investigate what happens if people changed their behavior sooner, that is, one, During this time, the number of infections peaked and the proportion of S F , accord-214 ing to our model simulations, ranged from 80% to 94%. Aside from the reduction 215 in cases, the peak was also delayed. Another parameter related to behavior change is µ, which quantifies the easing 217 of behavior from S F to S. In Figure 8 , we illustrate the effects of reducing µ and 218 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.03.29.22273148 doi: medRxiv preprint increasing β F on the cumulative number of cases and timing of the peak of infections. Higher values of β F and lower reduction factors of µ result to a significant decrease 220 in the number of COVID-19 cases and delay in the arrival of the peak. The blue 221 area on the right panel in Figure 8 indicates that there is no peak of infections 222 on Period 2. In other words, if the value of β F was increased and µ was reduced Figure 8 The effect of reducing the behavior change eased rate and increasing the transmission rate of the awareness of the disease on the cumulative cases (left, in log scale) and the peak of the epidemic (right). In Figure 9 , we illustrate the scenario if community quarantine and behavior in Table 1 ). The results, indicated by the blue curves, show that 230 there is a significant decrease in the number of cases and the second peak was not 231 present. This simulation aligns with the early prediction results presented in [45] , 232 where they showed that the cases were expected to decrease during Period 2, based 233 on the data from Period 1. Note that the simulation in Figure 9 did not consider CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.03.29.22273148 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 5, 2022. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.03.29.22273148 doi: medRxiv preprint Policy responses and government science advice for the COVID 19 315 pandemic in the Philippines Misplaced priorities, unnecessary effects: Collective suffering and 318 survival in pandemic Philippines PH needs 94,000 contact tracers-DOH Health: Beat COVID-19 Today: A COVID-19 Philippine Situationer Full Weekly Report Issue Department of Health: COVID-19 Tracker Philippines COVID-19 Testing: Understanding the "Percent Positive Inter-Agency Task Force for the Management of Emerging Infectious Diseases: Omnibus guidelines on the 332 implementation of community quarantine in the Philippines Department of Health: Beat COVID-19 Today: A COVID-19 Philippine Situationer Issue 106 Central Disaster and Safety Countermeasure Headquarters 337 (CDSCH)): Korea's Response to COVID-19 and Future Direction Government of the Republic of Korea: All About Korea's Response to COVID-19 342 srchTp=&multi_itm_seq=0&itm_seq_1=0&itm_seq_2=0&company_cd=& COVID-19 among income-poor households in the Philippines: A cross-sectional study An SIR epidemic model for COVID-19 spread with fuzzy 348 parameter: the case of Indonesia A case study of 2019-nCOV cases in 350 using classical and 351 fractional derivatives A mathematical model for the spread of COVID-19 and control 353 mechanisms in Saudi Arabia Modeling and forecasting the spread of COVID-19 with stochastic and deterministic 355 approaches: Africa and Europe A mathematical model of COVID-19 using fractional derivative in India with dynamics of transmission and control A fractional complex network model for novel corona virus in China A dynamic optimal control model for COVID-19 and cholera co-infection 361 in Yemen Mathematical model of COVID-19 spread in Turkey and South Africa: theory, 363 methods, and applications Effects of masks on the transmission of infectious diseases An SEIR model with infected immigrants and recovered emigrants Modeling the effects of contact tracing on COVID-19 transmission On an SE (Is)(Ih) AR epidemic model with combined vaccination and antiviral 25 Prediction of COVID-19 transmission dynamics using a mathematical model 375 considering behavior changes in Korea How Important Is Behavioral Change during the Early Stages of the COVID-19 Mathematical Modeling Study School opening delay effect on transmission dynamics of coronavirus 380 disease 2019 in Korea: based on mathematical modeling and simulation study The impact of social distancing and public behavior changes on 383 COVID-19 transmission dynamics in the Republic of Korea Understanding unreported cases in the COVID-19 epidemic China, and the importance of major public health interventions Predicting the number of reported and unreported cases for the COVID-19 388 epidemics in China Modeling the impact of unreported cases of the COVID-19 in 391 the North African countries Unreported cases for age dependent COVID-19 outbreak in Japan A new extension of state-space SIR model to account for Underreporting-An application 395 to the COVID-19 transmission in California and Florida Ein mathematisches Modell zur Schätzung der Dunkelziffer von SARS-CoV-2-Infektionen in der Frühphase der Pandemie am Beispiel 398 Highlights of the National Capital Region (NCR) 2020 Census of Population and Housing COVID-19 in the Philippines The COVID-19 pandemic and transport policy implications for a developing megacity. Sustainable Cities and 406 Society Center for Disease Control and Prevention: Ending Isolation and Precautions for People with COVID-19: 408 Interim Guidance Transmission, viral kinetics and clinical characteristics of the emergent SARS-CoV-2 Delta VOC in Guangzhou COVID-19 case fatality rates across Southeast Asian countries 414 (SEA): a preliminary estimate using a simple linear regression model A methodology for performing global uncertainty and 417 sensitivity analysis in systems biology Fitting dynamic models to epidemic outbreaks with quantified uncertainty: A primer for parameter 420 uncertainty, identifiability, and forecasts The construction of next-generation matrices for 423 compartmental epidemic models Mobility-guided modeling of the COVID-19 pandemic in Metro Manila