This page presents the various features and implementations in the models submitted to the Scenario Hub. To learn more about each model individually, please refer to the latest set of models abstracts.

Model | Authors | License |
---|---|---|

ECDC-CM_ONE | Rok Grah, Rene Niehus, Bastian Prasse | cc-by-nc-4.0 |

ECDC-CM_TWO | Rok Grah, Rene Niehus, Bastian Prasse | cc-by-nc-4.0 |

ICM-agentModel | Rafał P. Bartczuk, Filip Dreger, Łukasz Górski, Magdalena Gruziel-Słomka, Artur Kaczorek, Jan Kisielewski, Bartosz Krupa, Antoni Moszyński, Karol Niedzielewski, Jędrzej M. Nowosielski, Maciej Radwan, Franciszek Rakowski, Marcin Semeniuk, Jakub Zieliński | cc-by-4.0 |

JBUD-HMXK | Jozef Budzinski | cc-by-nc-nd-4.0 |

MOCOS-agent1 | Tyll Krueger, Marcin Bodych, Tomasz Ożański, Radosław Idzikowski, Piotr Szymański, Agata Migalska, Barbara Pabjan, Maciej Filiński, Marek Bawiec | mit |

MODUS_Covid-Episim | Jakob Rehmann, Sebastian Müller, Billy Charlton, Ricardo Ewert, Sydney Paltra, Christian Rakow, Tim Conrad, Christof Schütte, Kai Nagel | agpl-3.0 |

RIVM-vacamole | Kylie Ainslie | cc-by-4.0 |

SIMID-SCM | Nicolas Franco, Lander Willem, Steven Abrams, The SIMID COVID-19 team, Christel Faes, Philippe Beutels, Niel Hens | gpl-3.0 |

SwissTPH-OpenCOVID | Andrew J. Shattock, Cassandra Alvarado, Melissa A. Penny | gpl-3.0 |

TUWien-AustrianCoVABM | Martin Bicher, Claire Rippinger, Dominik Brunmeir, Christoph Urach, Melanie Zechmeister, Niki Popper | cc-by-4.0 |

UC3M-EpiGraph | David E. Singh, Aymar Cublier, Miguel Guzman Merino, Maria Cristina Marinescu, Jesus Carretero, Alberto Cascajo Garcia | cc-by-4.0 |

USC-SIkJalpha | Ajitesh Srivastava | mit |

UVA-EpiHiper | Jiangzhuo Chen, Stefan Hoops, Parantapa Bhattacharya, Dustin Machi, Madhav Marathe | cc-by-4.0 |

**ECDC-CM_ONE**: Considered that behaviour converges to a “new normal”, which may be close, but not equal to, pre-pandemic behaviour**ECDC-CM_TWO**: Considered that behaviour converges to a “new normal”, which may be close, but not equal to, pre-pandemic behaviour**ICM-agentModel**: NPIs are represented in the model by variation (in time and space) of agent contacting rates in appropriate contexts such as schools, workplaces etc.**JBUD-HMXK**: Not applicable**MOCOS-agent1**: NPI as represented through the input parameter to the model controlling intensity of infection kernels**MODUS_Covid-Episim**: No NPIs are implemented.**RIVM-vacamole**: The model uses different contact matrices from the Pienter Corona Study (https://www.rivm.nl/pienter-corona-studie) estimated throughout 2020 and 2021 to approximate contact patterns under different levels of non-pharmaceutical interventions within and between age groups.**SIMID-SCM**: NPI are modelled using CoMix social contact data collected every 2 weeks and additional age-specific proportionality factors detecting the transmission**SwissTPH-OpenCOVID**: Isolation utlized for submission, but model is capable of executing other NPIs.**TUWien-AustrianCoVABM**: NPIs are partially modelled directly (e.g. closure of schools, tracing, …) and partially modelled via reduction of contact numbers or transmissibility (e.g. face mask wearing)**UC3M-EpiGraph**: In this simulation we consider use of facemasks. With no changing conditions in their use for the period between March 13th 2022 and March 7th 2022**USC-SIkJalpha**: No NPI included, contact changes over time derived from Cuebiq data.**UVA-EpiHiper**: (i) A fraction of the population chooses to reduce non-essential (shopping, religion, and other) activities. (ii) All K-12 schools are closed from mid-June 2022 to mid-August 2022 and again from mid-December 2022 to the beginning of 2023. (iii) A fraction of symptomatic people choose to self-isolate themselves at home.

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: NPIs are represented in the model by variation (in time and space) of agent contacting rates in appropriate contexts such as schools, workplaces etc.**JBUD-HMXK**: Not applicable**MOCOS-agent1**: NPI as represented through the input parameter to the model controlling intensity of infection kernels**MODUS_Covid-Episim**: Not applicable.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Compliance to the intervention measures taken is assumed to be gradual over 5 days**SwissTPH-OpenCOVID**: Not explictly modelled**TUWien-AustrianCoVABM**: Adherence values are not modelled directly but calibrated**UC3M-EpiGraph**: We assume that the NPI do not change during the period from March 2022 to March 2023**USC-SIkJalpha**: Not modeled.**UVA-EpiHiper**: Compliances are 15% on average for (i), 100% for (ii), and 75% for (iii).

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: Not applicable**JBUD-HMXK**: Not applicable**MOCOS-agent1**: input parameter to the model understood as a probability that any of secondary infected persons for a positively tested individual will be detected**MODUS_Covid-Episim**: Contact tracing with minimal capacity and 50% success rate.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Not applicable**SwissTPH-OpenCOVID**: Not explictly modelled**TUWien-AustrianCoVABM**: A certain percentage of agent-agent contacts within the last days are traced and isolated for a certain timespan. Parameter values vary with time and are in accordance with the official guidelines. K1-contact testing is implemented but was never activated in the calibrated parametrization.**UC3M-EpiGraph**: We consider that a certain percentage of the infected individuals are self-quarantined at home.**USC-SIkJalpha**: Not applicable.**UVA-EpiHiper**: Not applicable.

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: simple scaling of the real infections to theconfirmed cases**JBUD-HMXK**: Not applicable**MOCOS-agent1**: input parameter to the model understood as a probability that given infected person will be tested positively**MODUS_Covid-Episim**: No mandatory testing considered.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Not applicable**SwissTPH-OpenCOVID**: Individuals can be tested, diagnosed, and can enter isolation for a given period with a given delay. For this application we assumed testing rates of 1% without symtptoms and 10% with symptoms.**TUWien-AustrianCoVABM**: dependent on an age-dependent detection rate, infections become confirmed with a certain time-delay**UC3M-EpiGraph**: Not considered for this version of the code (we will include it in future versions)**USC-SIkJalpha**: Not applicable.**UVA-EpiHiper**: Not applicable.

**ECDC-CM_ONE**: See vaccine_immunity_duration**ECDC-CM_TWO**: See vaccine_immunity_duration**ICM-agentModel**: VE is expressed as the reduction of daily probability of infection. In simple terms, the reduction constitutes partial immunity which is expressed as the product of some maximum immunity level and the profile of immunity variation in time.**JBUD-HMXK**: Initial vaccine effectiveness (full dose): 90 %; Initial vaccine effectiveness (booster): 50 %**MOCOS-agent1**: vaccine can protects from infection and from severe progression / death**MODUS_Covid-Episim**: Model vaccine efficacy is via our antibody model https://doi.org/10.1016/j.isci.2023.107554, which is built upon http://DOI.org/10.1056/NEJMc2119236 and http://DOI.org/10.1056/NEJMc2201607.**RIVM-vacamole**: Depends on age, vaccine product, and dose**SIMID-SCM**: We use a “leaky” vaccination approach. For example, vaccination with 50% effectiveness, implies that for a vaccinated individual the likelihood to acquire infection is 50% less compared to a non-vaccinated individual of the same age. The levels of protection against infection and hospital admissions after infection for different VOCs are presented in Page 5 Table 1 of https://covid-en-wetenschap.github.io/assets/20220415_technical_note_SIMID.pdf**SwissTPH-OpenCOVID**: Assumed the proportion of efficacy (0.85) that blocks infection is 0.8**TUWien-AustrianCoVABM**: Immunity is modelled as vector of boolean states. Each vector entry represents, whether the agent is immune against infection with the regarded virus strain. After an immunization event (i.e. vaccination and recovery), two random numbers decide about, whether (a) immunity is gained at all and (b) how long immunity lasts. Hereby, a parameter matrix is used containing assumptions for each immunization cause (vaccine, recovery from certain variant) against infection from each virus variant.**UC3M-EpiGraph**: The vaccine efficacy depends on the number of doses, the virus variant and its related immunity is also subjected to waning.**USC-SIkJalpha**: Not explicitly modeled**UVA-EpiHiper**: Not applicable.

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: 14 days since first dose or booster dose**JBUD-HMXK**: 7 days**MOCOS-agent1**: no delay**MODUS_Covid-Episim**: Not applicable**RIVM-vacamole**: Depends on vaccine product and dose but between 7-28 days.**SIMID-SCM**: Vaccine-induced immunity against infection is implemented as a step function with a switch from, e.g., 0% to 50% protection against infection 21 days after the first vaccine dose. Vaccine-induced protection against hospital admission is implemented incrementally on top of the protection against infection. Protection induced by the second and booster vaccine doses is assumed to be achieved fully (i.e., depending on the maximal vaccine effectiveness) 7 days after administration.**SwissTPH-OpenCOVID**: We assume peak efficacy is reached immideatly after the final dose of primary vaccination.**TUWien-AustrianCoVABM**: 14 days since first dose**UC3M-EpiGraph**: For Pfizer we consider, for the second doses and boosters, that after vaccination the maximum immunity is reached after 7 days. Then, it remains steady for 30 days and following this, it decreases following a gamma distribution.**USC-SIkJalpha**: 14 days**UVA-EpiHiper**: as specified in the scenario description.

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: vaccinations in model consistent with official data for powiats (counties) and age groups**JBUD-HMXK**: Not applicable**MOCOS-agent1**: vaccinations in model distributed among agents consistently with official data for age groups**MODUS_Covid-Episim**: No hesitancy integrated, but vaccination rates for round 5 are based on this round’s guidelines.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Not applicable, uptake based on reported data**SwissTPH-OpenCOVID**: Not explictly modelled**TUWien-AustrianCoVABM**: Regarded in a meta-model which generates the vaccine-forecast.**UC3M-EpiGraph**: First and second doses = Children 47%, Adults 16%, Elderly people 4%; Booster = Children 100%, Adults 60%, Elderly people 10%**USC-SIkJalpha**: A “suceptible-infected”model. Boosters are modeled such that the “susceptible” population is the eligible population who are yet to get a booster. To achieve the desired saturation level to create sub-scenarios, we increase the adoption rate over time.**UVA-EpiHiper**: as specified in the scenario description.

**ECDC-CM_ONE**: We assume exponential waning of neutralisation levels that is related to vaccine efficacy (VE) against infection via the following paper: https://doi.org/10.1038/s41591-021-01377-8. The starting VEs against infections for each vaccine product and age group is obtained from literature. The exponential decay time of neutralisation levels is then defined in such a way to obtain 40% (or 60%) decrease VE against infection after 10 (or 4) months. We assume no plateau.**ECDC-CM_TWO**: We assume exponential waning of neutralisation levels that is related to vaccine efficacy (VE) against infection via the following paper: https://doi.org/10.1038/s41591-021-01377-8. The starting VEs against infections for each vaccine product and age group is obtained from literature. Using this approach, the waning curves are sigmoid shaped and fit data points on waned VE over time well. Assuming an X% decrease of VE against infection after Z months (X and Z are defined either by the scenario round or by literature if waning speed was not specified), we obtain the exponential decay time of neutralisation levels to fit to the two VE points Z months apart. We assume no plateau of VE against infection. A similar approach is taken for VE again severe outcomes.**ICM-agentModel**: The immunity builds up to the maximum level for first 14 days, next it stays at maximum until 3 months after vaccination, finally the immunity decays gradually to predefined level.**JBUD-HMXK**: 105/70 days**MOCOS-agent1**: waning function based on empirical data / scenario constraints**MODUS_Covid-Episim**: We assume an exponential decay rate (of antibody levels) whereas the decay rate depends on the scenario.**RIVM-vacamole**: We assume vaccine effectiveness (VE) against transmission and infection wanes following a logistic curve parameterized such that VE decreases by 25 percentage points within the first 6 months after vaccination (Higdon et al. 2022). We assume VE against hospitalisation also wanes following a logistic curve, but at a slower rate than VE against infection and transmission, specifically, that VE decreases by 10 percentage points in the first 6 months following vaccination (Higdon et al. 2022).**SIMID-SCM**: Waning as in scenarios, using exponential law**SwissTPH-OpenCOVID**: Aassumed to wane following an inverse logistic curve but with an upper bound of 0.85, lower bound of 0.10, slope of 1.8, and a half-life of 70 days.**TUWien-AustrianCoVABM**: Varies and depends on the shot number and the target virus-strain. Usually, a Weibull distribution is used to sample the immunity loss date, which is fitted to published effectiveness data.**UC3M-EpiGraph**: 30 days after the dose, then waning following a gamma distribution**USC-SIkJalpha**: Only waning immunity is considered! The probability of transferring to partially immune state by at time t is modeled as a gamma distribution such that: (i) the median is as per the scenario; (ii) the efficacy (given the partial immune protection against infection) after the first 60 days is the “initial” vaccine efficacy.**UVA-EpiHiper**: vaccinated people go to a partially immune state after a period that is sampled from an exponential distribution with either 10-month or 4-month median.

**ECDC-CM_ONE**: We assume exponential waning of natural immunity; the exponential time is obtained from definition that natural immunity protection lowers by 40% (or 60%) within 10 (or 4) months. We assume no plateau.**ECDC-CM_TWO**: The naturally-acquired protection and its waning over time is obtained from: https://www.thelancet.com/action/showPdf?pii=S1473-3099%2822%2900801-5. Hybrid protection (HE) is assumed to be independently impacted by vaccine-induced (VE) and naturally-acquired (RE) protection by the following formula: HE= 1 - (1-VE)*(1-RE) at ever time t.**ICM-agentModel**: Maximum immunity within 3 months since recovery (total immunity in case of the same variant as in previous infection), after 3 months immunity decays gradually to predefined level.**JBUD-HMXK**: 202/135 days**MOCOS-agent1**: waning function based on empirical data / scenario constraints**MODUS_Covid-Episim**: Waning occurs with the same rate as for vaccine induced immunity.**RIVM-vacamole**: We assume natural immunity wanes by 60% after 8 months and follows an Erlang distribution.**SIMID-SCM**: 100% protection vs infection just after recovery, with waning as in scenarios using exponential law**SwissTPH-OpenCOVID**: Assumed to wane following an inverse logistic curve with an upper bound of 0.95, lower bound of 0.15, slope of 1.8, and a half-life of 105 days.**TUWien-AustrianCoVABM**: Varies and depends on the virus-strain the agent recovered from and the target virus-strain. Usually, a Weibull distribution is used to sample the immunity loss date, which is fitted to published effectiveness data.**UC3M-EpiGraph**: 30 days after infection or (if the individual was already infected) after the vaccination, then waning following a gamma distribution**USC-SIkJalpha**: Same as above**UVA-EpiHiper**: recovered (from infections) people go to a partially immune state after a period that is sampled from an exponential distribution with either 10-month or 4-month median.

**ECDC-CM_ONE**: Depending on age and vaccination status, ranging from 2.71e-9 to 1.3e-2**ECDC-CM_TWO**: Depending on age and vaccination status, obtained by calibrating to case and hospitalisation data.**ICM-agentModel**: Not applicable**JBUD-HMXK**: Not applicable**MOCOS-agent1**: age dependend, empirical data based**MODUS_Covid-Episim**: Not applicable.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Not applicable, the model is not based on confirmed cases**SwissTPH-OpenCOVID**: The probability of severe disease is a function of age, current level of acquired immunity, current level of vaccine-indiced immunity**TUWien-AustrianCoVABM**: Varies with the variant. Mechanism implemented, but not up to date since it is hardly used.**UC3M-EpiGraph**: Age-dependant values and dependent on whether the individual has acquired immunity. Not subjected to waning.**USC-SIkJalpha**: Learned from recent data.**UVA-EpiHiper**: age stratified, specified based on recent COVID-19 literatures.

**ECDC-CM_ONE**: Depending on age and vaccination status, ranging from 8.14e-9 to 3.9e-2**ECDC-CM_TWO**: Depending on age and vaccination status, obtained by calibrating to case and hospitalisation data provided an underascertainment factor (infections/cases) that is drawn from a prior distribution.**ICM-agentModel**: derived from probabilities of transitions between agent states**JBUD-HMXK**: Not applicable**MOCOS-agent1**: age dependend, empirical data based**MODUS_Covid-Episim**: Not applicable.**RIVM-vacamole**: Depends on age and vaccination status**SIMID-SCM**: Calibrated**SwissTPH-OpenCOVID**: Age related probability given a person is critical; logistic with upper bound of 2, lower bound of 9, half life of 90, and slope of 14.**TUWien-AustrianCoVABM**: Varies with the variant. Mechanism implemented, but not up to date since it is hardly used.**UC3M-EpiGraph**: Age-dependant values and dependent on whether the individual has acquired immunity. Not subjected to waning.**USC-SIkJalpha**: Learned from recent data.**UVA-EpiHiper**: age stratified, specified based on recent COVID-19 literatures.

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: dependent on the agent age**JBUD-HMXK**: Probability of diagnosis given symptoms: 75.00 %; Symptomatic proportion: 54.60 %**MOCOS-agent1**: age dependend**MODUS_Covid-Episim**: A rate of 0.2.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Calibrated**SwissTPH-OpenCOVID**: Assumed to be 33% of those infected.**TUWien-AustrianCoVABM**: Asymptomatic cases are not specifically regarded, since the terminology is too vague. We distinguish only between detected and undetected cases.**UC3M-EpiGraph**: We assume that 25% of the infected are asymptomatic**USC-SIkJalpha**: Similar to the range obtained for the US states using the CDC sero-prevelance data from the CDC. Note that this covers the currect cumulative fraction of unreported, untested, and asymptomatic.**UVA-EpiHiper**: 60% for naively susceptible people (without immunity), increased for people with immunity (from natural infection and/or vaccination).

**ECDC-CM_ONE**: [0-4, 5-9, 10-14, 15-17, 18-24, 25-49, 50-59, 60-69, 70-79, 80+]**ECDC-CM_TWO**: [0-4, 5-9, 10-14, 15-17, 18-24, 25-49, 50-59, 60-69, 70-79, 80+]**ICM-agentModel**: one-year groups**JBUD-HMXK**: Not applicable**MOCOS-agent1**: one-year**MODUS_Covid-Episim**: Agents can have any age between 0 and 99, but certain parameters and analyses employ different binning systems based on the available data.**RIVM-vacamole**: [0-9, 10-19, 20-29, 30-39, 40-49, 50-59, 60-69, 70-79, 80+]**SIMID-SCM**: 10 years age groups**SwissTPH-OpenCOVID**: [5-17, 18-59, 60+]**TUWien-AustrianCoVABM**: one-year groups**UC3M-EpiGraph**: 0-5, 6-10, 10-15, 15-20, 20-30, 30-40, 40-50, 50-60, 60-70, 80-90, 90-100+**USC-SIkJalpha**: No stratification by age**UVA-EpiHiper**: preschool (0-4 years), students (5-17), adults (18-49), older adults (50-64) and seniors (65+).

**ECDC-CM_ONE**: Not considered relevant for this submission round, but model does provide the functionality to specify importation of cases**ECDC-CM_TWO**: Not considered relevant for this submission round, but model does provide the functionality to specify importation of cases**ICM-agentModel**: seed agents introduced for each virus variant**JBUD-HMXK**: Not applicable**MOCOS-agent1**: Can be defined explicitly as input to the model. Available options include 1) InstantOutsideCases (constant number of imports at a given time), 2) CyclicOutsideCases (constant number of imports during the specified period and with a specified frequency). Each of the importations events include the variant type.**MODUS_Covid-Episim**: Based on disease import for 2021 and 2022. Numbers were provided by the public health authorities in Cologne.**RIVM-vacamole**: Not included in this submission round, but model does allows for the specification of case importation rate**SIMID-SCM**: No importation, except a seed for each new variant**SwissTPH-OpenCOVID**: Assume 1.5 importated infections per day per 100,000 people**TUWien-AustrianCoVABM**: scale with the reported/extrapolated number of overnight-stays per month and region.**UC3M-EpiGraph**: We use data provided the by Spanish National Health service. We scale the detected cases in order to estimate the real ones.**USC-SIkJalpha**: Not applicable**UVA-EpiHiper**: Not applicable.

**ECDC-CM_ONE**: Quantiles followed from MCMC and prior distribution of parameters**ECDC-CM_TWO**: Quantiles followed from prior distribution of parameters**ICM-agentModel**: Not applicable**JBUD-HMXK**: Nonparametric**MOCOS-agent1**: Samples based**MODUS_Covid-Episim**: Not applicable**RIVM-vacamole**: Quantiles from multiple simulations**SIMID-SCM**: Samples coming from independent MCMC chains, plus stochastic realisation on transmissions**SwissTPH-OpenCOVID**: xxx**TUWien-AustrianCoVABM**: Monte-Carlo simulation is used to quantify the uncertainty of the stochastic model mechanisms. Rarely used.**UC3M-EpiGraph**: Based on calibration and statistical analysis of multiple executions.**USC-SIkJalpha**: We generate multiple trajectories by (i) Uniform sampling of the recently seen infection rates, under-reporting/asymptomatic population, vaccine coverage, waning parameters uncertainty, and variant prevelance. (ii) For each of the above case trajectories, we use multiple possibilities of death and hospitalization rates that have been seen in the recent past. The above results in multiple points for each day. Quantiles are generated based on sampling among these points for each day.**UVA-EpiHiper**: estimated from multiple replicates of simulations.

**ECDC-CM_ONE**: Markov chain Monte Carlo**ECDC-CM_TWO**: The vector composed of all compartments and all groups is considered to be proportional (parallel) to the steady-state vector. The proportionality scalar is calibrated to reported cases, hospitalisations and deaths (or a subset if not all three observations are available for a country).**ICM-agentModel**: The hybrid approach is utilised. Many model parameters are based on literature (e.g. duration of latent state) The absolute infectivity of the wild type (parameter alpha ) and the reference (assuming usual behaviour) fraction of symptomatic agents contributing to their non-household contexts (parameter f) is optimised using Bayesian search and historic confirmed cases data. Finally, some parameters, most notably multipliers of context contacting rates, are calibrated using the trial and error method.**JBUD-HMXK**: Not applicable**MOCOS-agent1**: Grid search samples around most likely points fitted to last 4 weeks of observed values**MODUS_Covid-Episim**: Structural calibration against virus concentration in sewage data and shares of different VOCs.**RIVM-vacamole**: The baseline (non-seasonal) transmission rate and initial conditions for forward simulations are calibrated by fitting the model to daily confirmed cases from the national notification database Osiris. The model is fitted to data piecewise to correspond with the correct contact patterns associated with different non-pharmaceutical interventions within each time window.**SIMID-SCM**: MCMC Random-Walk Metropolis algorithm performed on a deterministic version of the model**SwissTPH-OpenCOVID**: Calibrated at the national-level using epidemioloical data to match 4 metrics (daily confirmed cases, deaths, hospital admissions, and ICU admissions) using a Gaussian Process and Latin Hypercube sampling.**TUWien-AustrianCoVABM**: The model uses one free time-dependent factor multiplied to the transmissibility to calibrate the model to past incident cases per age and federal-state. The sequentially used hospitalization model fits the hospitalization/ICU rates and length of stays to the last 100 daily observation points of the occupancy.**UC3M-EpiGraph**: We calibrate the model using the time interval from July 2021 and March 2022**USC-SIkJalpha**: All calibrations using regression. No manual-tuning other than the listed assumption and trivial settings: non-zero lag between infection and death.**UVA-EpiHiper**: For each state, we calibrate the transmissibility parameter in our disease model targeting daily confirmed cases in the recent weeks until the last date of fitting data.

**ECDC-CM_ONE**: Not applicable**ECDC-CM_TWO**: Not applicable**ICM-agentModel**: model agents are geo-referenced**JBUD-HMXK**: Not applicable**MOCOS-agent1**: Not applicable**MODUS_Covid-Episim**: Home and activity locations of agents are simulated using transportation model MATSim (http://matsim.org); in Episim these contact locations are simulated as containers with activity specific parameters.**RIVM-vacamole**: Not applicable**SIMID-SCM**: Not applicable**SwissTPH-OpenCOVID**: Not applicable**TUWien-AustrianCoVABM**: one model agent is used for every inhabitant of the country. Each of them has a static latitude/longitude**UC3M-EpiGraph**: 10 cities, 5,018,260 inhabitants.**USC-SIkJalpha**: Not applicable.**UVA-EpiHiper**: embedded in our synthetic contact network model.