5 Data and Methodology
5.5 Scenarios
Based on our models (hereafter, baseline model and extended baseline model described in subsections 5.3.1 and 5.3.2, respectively), we simulate and analyse the effect of three main scenarios under different levels of intervention and the impact of additional control efforts in reducing the burden of disease in Portugal. In practice, those interventions are usually modelled by assuming that they decrease the contact rate or reduce the relative infectiousness of infectious cases. In our study, we simulate these effects by changing a
35 component of the contact matrix (𝑎𝑆,𝑡,𝑎𝑊,𝑡, 𝑎𝑂,𝑡) and/or the transmission probability (𝛽0−19,𝑡𝑃 , 𝛽20−39,𝑡𝑃 , 𝛽40−59,𝑡𝑃 and 𝛽+60,𝑡𝑃 ), as it has been done in other COVID-19 modelling studies [13,60,61]. For each scenario, we analyse the fraction of infections averted, the incidence and cumulative infections and deaths, and the impact on the healthcare system in relation to the baseline model and the baseline capacity provided by the Portuguese National Health System for the COVID-19 patients.
5.5.1 First Scenario: Unmitigated scenario
We consider a first hypothetical scenario in which no measures are implemented over the course of the epidemic. To simulate an unmitigated scenario, we use the estimates of the values of 𝛽0−19,𝑡𝑃 , 𝛽20−39,𝑡𝑃 , 𝛽40−59,𝑡𝑃 , and 𝛽+60,𝑡𝑃 for the eight periods (1 ≤ t ≤ 8) from the baseline model (see Table 6.1), whereas the coefficients of the contact matrix are fixed as 𝑎𝑆,𝑡= 1, 𝑎𝑊,𝑡= 1, and 𝑎𝑂,𝑡 = 1 throughout the entire simulation (1 ≤ t ≤ 8). Setting the contact matrix as mentioned enables us to capture a mixing contact pattern similar to the pre-pandemic period in which no physical distancing restrictions, such as school closure, remote working, banning of social events and mass gatherings, are adopted. Nevertheless, note that using the baseline values of the transmission probabilities might reflect some changes due to an increased effort of active protective measures and public health sanitation over time.
5.5.2 Second Scenario: Physical distancing interventions
Between 3rd March 2020 and 10th February 2021, the Portuguese Government implemented two states of emergency (announced on 18th March 2020 and 6th November 2020 [6]), which included social- and physical- distancing policies to reduce mixing contacts, to slow the transmission of SARS-CoV-2. The lifting of such restrictions, even partially and gradually over time, was usually accompanied, weeks later, by an increase in the infection incidence, leading to a resurgence of COVID-19 waves, as seen between November 2020 and January 2021. Such lifting of social interventions has prompted multiple questions regarding how effective they were and what would happen if they were not relaxed. Among many likely questions, in this scenario’s subsection, we explore the effect of three of them, which are described in A, B, and C, as follows:
A. What would happen if the government had not lifted the first state of emergency before the 2020 summer break?
The end of the first state of emergency was a phase-out process, starting by the beginning of May and ending by June 2020 [75]. Along with lifting extreme restrictions policies, the Portuguese government
36 introduced soft distancing public health measures, such as face mask usage in indoor settings and restrictions of mass-gatherings, which were maintained during summer break and forward. Although the number of cases of COVID-19 remained stable during summer, as well as the value of Rt at a national level, some Portuguese regions, mainly Lisboa e Vale do Tejo (LVT), remained under a contingency state due to a significant daily incidence in the number of cases and hospitalisations [76] which were also extended to the rest of the country in mid-September 2020. Following the summer break, school reopening, and resumption at workplaces, infection incidence increased. For this reason, we explore the consequences of maintaining the first state of emergency during the summer. To simulate this scenario, we maintain all parameters values (i.e., coefficients of contact matrix and transmission probabilities for all age groups) estimated for 27th March to 27th April 2020 until 25th August 2020. Note that the period between 27th March and 27th April 2020 corresponds to the first period of extreme restriction policies in our baseline model.
B. What would happen if schools remained under virtual learning for the 2020/2021 academic year?
School closure was one of the primary measures implemented early in the SARS-CoV-2 pandemic, and they remained so until the end of the academic year by July 2020
,
except for children under the age of 10 who returned by the beginning of June 2020. The adoption of such intervention was based on previous knowledge regarding the leading role of children in the spread of influenza and other viruses. The early adoption of school closure has hampered an accurate estimation of children's contribution in spreading SARS-CoV-2, leading to many uncertainties regarding their susceptibility and infectiousness. With schools’ reopening on 18th September 2020 in Portugal, even under the restriction of mask usage for young people over 10 years old in classrooms, several outbreaks in those settings were occurring weeks later.In this scenario, we explore the effect of maintaining virtual learning with schools remaining closed for all children and adolescents under the age of 19 years. To simulate this scenario, we consider the summer break as a proxy for school closure, as it has been done in other studies for other infectious diseases [77].
Therefore, we maintain the values of the coefficient of the contact matrix for schools (𝑎𝑆) and the transmission probability for [0-19] age group (𝛽0−19𝑃 ) estimated for 18th July to 25th August 2020 until 10th February 2021. The parameters for the other matrices (𝑎𝑤, 𝑎𝑜) and transmission probabilities for other age groups (𝛽20−39𝑃 , 𝛽40−59𝑃 , 𝛽60+𝑃 ) were fixed at their baseline values. Note that the period between 18th July and 25th August 2020 corresponds to the summer break in our baseline model.
Additionally, periods of school closure, and thus holidays period, in general, may imply a change in mixing contacts across the community due to likely changes in mixing behaviour (e.g., schoolchildren being cared for by their grandparents or other older people) [77]. A sensitivity analysis was performed to assess how this mixing change may diminish the benefits of virtual learning due to the increased contact between
37 children and elders. To do so, we simulated the effect of an increase of 20%, 50%, and 100% contacts at home and other locations between children under 19 years and individuals over 60 years old. Similar analyses and assumptions were considered in [13]. Note that we do not distinguish weekdays and weekends;
however, children are likely at home with their parents during weekends.
C. What would happen if the second state of emergency had not been relaxed by Christmas?
In mid-December 2020, there was a controversial debate about which measures should be implemented on Christmas Day in order to maintain the festive and traditional period while limiting the spread of COVID-19. Whereas many countries (e.g., Germany, Netherlands, Denmark) chose to maintain extreme restriction measures, including lockdowns with shops and markets closed, others decided to ease their measures for a relatively short window around Christmas Day, as happened in Portugal [78]. In particular, between 23rd -26th December 2020, Portugal relaxed containment measures, allowing, for example, internal travel. In turn, the mandatory curfew during weekdays and weekends was extended, and recommendations to avoid large family gatherings and face mask usage were suggested, despite no specific limit on the number of people at Christmas Day gatherings being imposed [78]. Weeks later, a remarkable increase in the number of infections followed, and as a consequence, more restricted interventions were announced in mid-January 2021 to suppress transmission and reduce the burden of the healthcare systems. In this scenario, we explore the effect of extending the second state of emergency until the end of the simulation without lifting any containment measure by Christmas. To simulate it, we maintain all parameters values (i.e., both coefficients of contact matrix and transmission probabilities for all age groups) estimated for 18th November 2020 to 22nd December 2020 until 10th February 2021.
5.5.3 Third Scenario: Contact tracing
Since the beginning of the epidemic, contact tracing (i.e., tracing and quarantining high-risk contacts) has been implemented in many countries and evidence has shown a remarkable impact in reducing transmission chains. In particular, such impact results from tracing infected contacts that remain in their latent periodor are asymptomatic, which are harder to detect, and otherwise, they could contribute to the spread of the infection.
This subsection aims to explore the effect of additional contact tracing efforts when implemented at three different moments of the epidemic in Portugal: (A) from the time the first state of emergency was lifted on 4th May 2020, (B) since schools’ reopening on 18th September 2020, and (C) during the second COVID-19 peak wave on 17th November 2020. Specifically, these scenarios aim to assess what would happen if (A) the lifting of the first state of emergency; (B) schools’ reopening for the 2020/2021 academic year, and (C)
38 the second state of emergency were accompanied by an additional effort in tracing and isolating infected contacts. Note that these three different moments allow us to evaluate the effect of additional contact tracing efforts when implemented at different phases of the epidemic, including a decreasing epidemic activity phase (A), an exponential growing phase (B), and during the peak of the second wave (C).
For all situations (A, B, C), we assess the effect of tracing and isolating more infected contacts that might be in the exposed, presymptomatic, asymptomatic, and symptomatic states with different levels of tracing (0%, 5%, and 10% additional tracing). Note that 0% of additional tracing depicts the baseline scenario. In turn, the choice of low levels (5% and 10%) results from several reasons. First, the baseline model already implicitly incorporates a fraction of contacts that were traced and quarantined through the practice of contact tracing strategy during the epidemic in Portugal. Additionally, since we are tracing only infected contacts (people who are already infected), tracing 5% or 10% more infected contacts implies, in practice, the need for a much larger number of people being contacted. Therefore, the choice of 5% and 10%
additional tracing simulates a reasonable and affordable intervention. Also, for each scenario (A, B, C), we assume that all traced infected contacts are instantly isolated, preventing onwards transmission.
To estimate the number of contact tracers needed to trace the additional infected contacts projected for each scenario, we assume that (i) each infected person has on average 3-8 contacts, (ii) each call between the tracer and the contact has a duration of approximately 20 minutes, and (iii) each contact tracer can contact
≈ 20 people per day. These assumptions were based on information gathered from press releases [79] and on similar estimations made by others [80,81].
To simulate these scenarios, we use the extended baseline model described in subsection 5.3.2 and fix the baseline parameters according to the values of Table 6.1. We assume that the fraction of contacts traced in the exposed, presymptomatic, asymptomatic, and symptomatic states (fQ_e, fQ_p, fQ_a, fQ_s, respectively) are fixed according to the different levels of tracing mentioned above. As an example, for 5% of tracing since May 2020, we set the parameters of tracing as: fQ_e = 5%, fQ_p = 5%, fQ_a = 5%, and fQ_s = 5%
from that date onwards.