CoSim

Online COVID-19 Simulator


Model version: 5200G
Data from: 15.06.2021
ICU capacity data from: 16.06.2021
Appliction version: 4.2

Contact

Prof. Dr. Thorsten Lehr
thorsten.lehr[at]mx.uni-saarland.de

Simulator activity

The connection to the simulator will be closed after 5 minutes of inactivity. Please reload the page to continue.

What it does

This simulator is an interactive tool to explore the current situation of the COVID-19 pandemic in the Germany as well as repercussions of potential changes for the future.
The simulations are based on a modified SEIRD model (Susceptible - Exposed - Infectious - Recovered– Death) developed by this working group. The model utilizes nonlinear mixed-effects modeling to describe and predict a variety of stages relevant to the prevalence of the SARS-CoV-2 virus and its impact on the health care system in Germany. The model relies on several data sources for its descriptive and predictive performance including Johns Hopkins University, WHO, CDC, ECDC, Robert-Koch-Institute, the Berlin Morgenpost and others.
Both data and model are updated and refined on a weekly basis.

Further Information

Additional information can be found in the weekly reports on the website www.covid-simulator.com.

How to Use

  1. Press Simulator on the left-hand sidebar
  2. Choose either Germany, a German federal state or a German district of your interest
  3. Changing # Future vaccination rate gives you the option to simulate with a custom daily vaccination rate in % of population who get fully vaccinated.
  4. Changing # Future R(t)s gives you the option to simulate with up to three custom R(t) values from the time of your choosing to see how each state would fare with more or less new infections.
  5. Select a start and endpoint for the time frame of your interest
  6. Press the simulate button to start the simulation
  7. Important! After each change to the selection the simulate button has to be pressed again to apply the changes
  8. Beneath the plots you can choose further graphing options and/or include additional information
    1. Add or remove some of the observed and predicted data by pressing the respective button
    2. The R(t) button adds a line showing the R(t) value over time
    3. With Rolling Avg you can switch between daily data or a 7-day rolling average
    4. With Log you can switch between linear and logarithmic scale
    5. The Variability button shows or hides the prediction and/or 95% confidence interval of the model
    6. The Observed button shows or hides observations on which the model is based. In the plots they are shown as points
    7. The ICU Capacity button adds a dotted line visualizing the maximal number of ICU beds available in each state

The Authors

Christiane Dings1, Katharina Götz1, Katharina Och1, Iryna Sihinevich1, Dr. Dominik Selzer1, Quirin Werthner1, Lukas Kovar1, Fatima Marok1, Christina Schräpel1, Laura Fuhr1, Denise Türk1, Hannah Britz1, Prof. Dr. Sigrun Smola2, Prof. Dr. Thomas Volk3, Prof. Dr. Sascha Kreuer3, Dr. Jürgen Rissland2, Prof. Dr. Thorsten Lehr1

1 Clinical Pharmacy, Saarland University, Germany
2 Institute of Virology, University Hospital of the Saarland, Germany
3 Department of Anaesthesiology, University Hospital of the Saarland, Germany


The online simulator was developed by Quirin Werthner, Iryna Sihinevich, Katharina Och, Katharina Götz, Christiane Dings, Dr. Dominik Selzer and Prof. Dr. Thorsten Lehr.

Software

CoSim is built with the R programming language and the packages Shiny, Tidyverse and mrgsolve. Model development is accomplished by NONMEM®.

Agreement

This website and its contents herein, including all data and analysis are provided to the public strictly for educational and academic research purposes. Reliance on the Website for medical guidance or use of the Website in commerce is strictly prohibited.

Changelog


Model version: 5200G
Data from: 15.06.2021
ICU capacity data from: 16.06.2021
Appliction version: 4.2

Jun 16th 2021 - Data Update

- Updated observed data from Jun 15th 2021.
- Updated DIVI data from Jun 16th 2021.


Jun 09th 2021 - Data Update

- Updated observed data from Jun 08th 2021.
- Updated DIVI data from Jun 08th 2021.


Jun 02nd 2021 - Data Update

- Updated observed data from Jun 01st 2021.
- Updated DIVI data from Jun 02nd 2021.


May 25th 2021 - Data Update

- Updated observed data from May 24th 2021.
- Updated DIVI data from May 26th 2021.


May 19th 2021 - Data Update

- Updated observed data from May 18th 2021.
- Updated DIVI data from May 19th 2021.


May 12th 2021 - Data Update

- Updated observed data from May 11th 2021.
- Updated DIVI data from May 12th 2021.


May 06th 2021 - Data Update & Version 4.2

- Updated observed data from May 04th 2021.
- Updated DIVI data from May 05th 2021.
- Removed user defined ICU cap option.
- Added 'Vaccination willingness' (percentage of population willing to get vaccinated).
- Changed 'Future R(t)' option to 'Infectiousness changepoints' for user defined relative change in infectiousness at certain time points.
- Added 'first' and 'final' vaccinated plotting option to display age-stratified modeled vaccinations.
- Underlying age-structure is now unique for each state.


April 30th 2021 - Version 4.1

- Fixed model computation of time until discharge for ventilated recovered patients.


April 28th 2021 - Data Update & Version 4.0

- New underlying model with age-dependent hospitalizations and vaccination schema.
- B.1.1.7 transmissibility advantage is now fixed at 35% (as the default in model version 3)
- For more details see the weekly reports at www.covid-simulator.com.
- Updated observed data from April 27th 2021.
- Updated DIVI data from April 28th 2021.


April 14th 2021 - Data Update

- Updated observed data from April 13th 2021.
- Updated DIVI data from April 14th 2021.


April 09th 2021 - Data Update

- Updated observed data from April 08th 2021.
- Updated DIVI data from April 09th 2021.


April 07th 2021 - Version 3.3

- Added "day" to date format on plot x-axis.


April 1st 2021 - Version 3.2

- Fixed a bug in simulation export for date ranges without observed data.


March 31th 2021 - Data Update

- Updated observed data from March 30th 2021.
- Updated DIVI data from March 31th 2021.


March 24th 2021 - Data Update

- Updated observed data from March 23th 2021.
- Updated DIVI data from March 24th 2021.


March 17th 2021 - Data Update

- Updated observed data from March 16th 2021.
- Updated DIVI data from March 17th 2021.


March 10th 2021 - Data Update & Version 3.1

- Fixes Rt plots when no category was set.
- Updated observed data from March 09th 2021.
- Updated DIVI data from March 10th 2021.


March 03rd 2021 - Data Update

- Updated observed data from March 02nd 2021.
- Updated DIVI data from March 03rd 2021.


February 24th 2021 - Data Update

- Updated observed data from February 23th 2021.
- Updated DIVI data from February 24th 2021.


February 17th 2021 - Data Update

- Updated observed data from February 16th 2021.
- Updated DIVI data from February 17th 2021.


February 10th 2021 - Data Update & Version 3.0

- Updated observed data from February 09th 2021.
- Updated DIVI data from February 10th 2021.
- Added B.1.1.7 Transmission Advantage/Gain option (experimental)

February 03rd 2021 - Data Update

- Updated observed data from February 02nd 2021.
- Updated DIVI data from February 03rd 2021.

January 37th 2021 - Data Update & Version 2.6

- Updated observed data from January 26th 2021.
- Updated DIVI data from January 25th 2021.
- Updated the About page.
- Simulation end date will be set to the 2 weeks ahead of the latest Future R(t) date if simulation date was set < latest Future R(t) date.
- Rearrangement of plots.

January 21th 2021 - Data Update & Version 2.5

- Updated observed data from January 20th 2021.
- Updated DIVI data from January 21th 2021.
- Up to 3 Future R(t) can know be set.

January 16th 2021 - Interim Data Update

- Updated observed data from January 15th 2021.
- Updated DIVI data from January 16th 2021.

January 13th 2021 - Data Update & Version 2.4

- Updated observed data from January 12th 2021.
- Updated DIVI data from January 13th 2021.
- Backend optimizations
- Future R(t) date can now be set to one day past the last estimated R(t) change point or later.

January 06th 2021 - Data Update

- Updated observed data from January 05th 2021.
- Updated DIVI data from January 06th 2021.

January 1st 2021 - Interim Data Update

- Updated observed data from December 31th 2020.
- Updated DIVI data from January 1st 2021.

December 30th 2020 - Data Update

- Updated observed data from December 29th 2020.
- Updated DIVI data from December 30th 2020.

December 23th 2020 - Data Update

- Updated observed data from December 22th 2020.
- Updated DIVI data from December 23th 2020.

December 16th 2020 - Data Update & Version 2.3

- Updated observed data from December 15th 2020.
- Updated DIVI data from December 15th 2020.
- Rolling (7-day window) avarage option for daily data.

December 14th 2020 - Interim Data Update

- Updated observed data from December 13th 2020.
- Updated DIVI data from December 14th 2020.

December 9th 2020 - Data Update

- Updated observed data from December 8th 2020.
- Updated DIVI data from December 9th 2020.

December 8th 2020 - Interim Data Update

- Updated observed data from December 7th 2020.
- Updated DIVI data from December 8th 2020.

December 2nd 2020 - Data Update

- Updated observed data from December 1st 2020.
- Updated DIVI data from December 2nd 2020.

November 25th 2020 - Data Update

- Updated observed data from November 24th 2020.
- Updated DIVI data from November 25th 2020.

November 18th 2020 - Data Update & Version 2.2

- Updated observed data from November 17th 2020.
- Updated DIVI data from November 18th 2020.
- Fixed typo in ICU cap line.

November 14th 2020 - Interim Data Update

- Updated observed data from November 12th 2020.
- Updated DIVI data from November 14th 2020.

November 11th 2020 - Data Update & Version 2.1

- Updated observed data from November 10th 2020.
- Updated DIVI data from November 11th 2020.
- Added observed data to daily cases (infected/recovered/dead; if avalable)
- Added 7-day incidence / 100k plot

November 06th 2020 - Version 2.0

- New layout (more mobile friendly)
- New discrict model
- Export option for simulations to CSV files
- Small UI fixes
- Added DIVI imported ICU capacities (weekly updated)

November 04th 2020 - Data Update

Updated observed data from November 3rd 2020.

October 28th 2020 - Data Update

Updated observed data from October 27th 2020.

October 21th 2020 - Data Update & Version 1.8

- Added 'Show Variability' Option for residual errors and SD for projective R(t)
- Fixed small UI bugs
- Fixed typos in error messages
- Updated observed data from October 20th 2020.

October 19th 2020 - Version 1.7

- Fixed a bug that produced wrong simulations if start time was set to a date later than 01-01-2020

October 17th 2020 - Version 1.6

- Added timer to the simulator to close connection after 5 minutes of inactivity.
- Added link to covid-simulator.com and clinicalpharmacy.me to the navigation pane.
- Custum R(t) default value is now last recent estimated R(t) for the selected state.
- Added FAQs (in German) to the navigation pane.

October 15th 2020 - Version 1.4

Added Software section to About page.

October 14th 2020 - Data Update

Updated observed data from October 13th 2020.

October 07th 2020 - Data Update & Version 1.3

- Updated observed data from October 6th 2020.
- Simulation default end date will now be fetched from current date of the session.

September 23th 2020 - Data Update

Updated observed data from September 22th 2020.

September 10th 2020 - Data Update & Version 1.2

- Updated observed data from September 09th 2020.
- Fixed an UI bug for ICU cap (beds) that was displayed when ICU cap display option was not enabled.

September 1st 2020 - Version 1.1

Fixed a bug where the last R(t) for non-custom R(t) was set to R(t-1).

August 27th 2020 - Data Update

Updated observed data from August 26th 2020.

August 18th 2020 - Version 1.0

Initial release

Frequently Asked Questions (Stand Mai 2021)


Was für eine Art Modell wurde hier implementiert?
Das zugrundeliegende Modell ist ein auf 27 Kompartimente erweitertes SEIR (Susceptible, Exposed, Infectious, Recovered) Modell. Details des Modells sind in den wöchentlichen Berichten (covid-simulator.com) abgebildet. Modellcode und Parameterschätzwerte sind auf GitHub abgelegt (github.com).

In welcher Software wurde das Modell entwickelt?
Es wurde die Software NONMEM, Version 7.4.3 verwendet und die Technik des non-linear mixed effects (NLME) modeling angewendet. Das Modell wird mittels gewöhnlicher Differentialgleichungen (ODEs) modelliert.

Welche Faktoren gehen in das Modell ein?
Die Alters- und Geschlechtsstruktur der Infizierten, Impfungen, Variants of Concern (B.1.1.7), Testanzahl und Positivenrate gehen als Einflussfaktoren auf das Infektionsgeschehen und den Krankheitsverlauf in das Modell ein. Details des Modells sind in den wöchentlichen Berichten (covid-simulator.com) abgebildet.

Können sich im Modell alle Menschen der Population infizieren?
Ja. Im Modell wird keine Grundimmunität in der Bevölkerung zu Beginn der Pandemie angenommen.

Können sich im Modell Menschen nach einer ausgestandenen Infektion wieder infizieren?
Nein, da die Immunitätsdauer momentan noch unklar ist. Sollte die Immunitätsdauer wissenschaftlich belegt sein, dann kann diese problemlos eingebaut werden. In der momentanen Pandemiephase hat dies noch keinen Einfluss auf die Simulationen.

Macht es Sinn, sehr weit in die Zukunft zu simulieren?
Nein. Eine Vorhersage weiter als einige Wochen in die Zukunft ist mit erheblicher Unsicherheit verbunden. Das Modell dient zur Veranschaulichung der momentanen und retrospektiven Entwicklung der Pandemie in Deutschland. Eine Simulation in die ferne Zukunft würde eine Vorhersage der politischen Maßnahmen, Verhaltensänderungen, Saisonalität oder Altersstruktur der Infizierten etc. voraussetzen. Dies kann das Modell nicht leisten.

Das Modell sagt eine [SEHR HOHE ZAHL] an Infizierten für die Zukunft voraus. Kann dies eintreffen?
Die Infektionsdynamik ist von vielen Faktoren abhängig und zukünftige Entwicklungen (vor allem in der fernen Zukunft) können von dem Modell nicht vorhergesagt werden.

Was zeigt die "ICU Capacity" an?
Die ICU Capacity zeigen eine Näherung aller verfügbaren Intensivbetten (auch Betten die momentan belegt sind) in den einzelnen Bundesländern und Deutschland. Die Zahlen werden wöchentlich dem DIVI Intensivregister entnommen.

Was passiert wenn die "ICU Capacity" überschritten wird?
Die Anzahl der Intensivbetten ist nur als Hilfslinie implementiert. Es gibt keine erhöhte Sterblichkeit im Modell wenn diese Grenze überschritten wird. Dies könnte zwar implementiert werden aber es ist unklar, wie sehr die Sterblichkeit bei dem Erreichen der Kapazitätsgrenzen ansteigt. Beim Überschreiten der Kapazitätsgrenzen unterschätzt unser Modell die Todeszahlen.

Werden im Modell auch nicht nachgewiesene Infizierte berücksichtigt (“Dunkelziffer“)?
Das Modell unterscheidet intern zwischen "Exposed" (Infiziert aber (noch) nicht identifiziert) und "Infected" (Infiziert und durch z.B. einen Test bestätigt). Da das Modell zur Beantwortung von Fragen zur Hospitalisierung entwickelt wurde, sind keine Annahmen zur Dunkelziffer implementiert, können aber nach gesicherter wissenschaftlicher Erkenntnis problemlos implementiert werden.

Kann das Modell den Effekt einer potentiellen (durch Infektion bedingten) Herdenimmunität simulieren?
Nein. Hierfür wurde es nicht entwickelt. Das Modell wurde speziell zur Beantwortung von Fragen zur Hospitalisierung entwickelt. Worst-case Szenarien und (noch) unklare immunologische Langzeitkonsequenzen nach Infektion sind nicht abgebildet.

Kann das Modell den Effekt einer zukünftigen Durchimpfung der Bevölkerung simulieren?
Ja. Die Impfwilligkeit kann zwischen 50% und 100% eingestellt werden. Allerdings wird diese für alle Altersklassen gleich angenommen.

Kann das Modell politische Maßnahmen/Verhaltensänderungen/Saisonalität/Alter der Infizierten etc. für die Zukunft vorhersagen?
Nein. Das Modell simuliert mit der zuletzt abgeschätzten Kinetik in die Zukunft. Der R(t) Wert nimmt aktuell durch die zunehmende Impfung ab. Eine Änderung des Reproduktionswerts kann über die Option "Infectiousness changepoints" gewählt werden, um beispielsweise Maßnahmen oder Verhaltensänderungen zu simulieren. Die Infektosität kann um bis zu 120% gesteigert werden (z.B. Lockerungen) oder um bis zu 80% gesenkt werden (z.B. Beschränkungen). Geeignete Werte zur Änderung der Infektiosität müssen vom Nutzer selbst gewählt werden.

Kann man Änderungen der Infektionsdynamik simulieren ("Infectiousness changepoints")?
Ja. Im Simulator kann man eine Änderung der Infektosität ab einem bestimmten Datum in der Zukunft bestimmen (Option "Infectiousness changepoints"). Die Infektosität kann um bis zu 120% gesteigert werden (z.B. Lockerungen) oder um bis zu 80% gesenkt werden (z.B. Beschränkungen). Die Änderung ist in % angegeben, da sich der R(t) Wert durch die Impfungen weiterhin dynamisch ändert.

Gibt es auch Simulation für einzelne Land- und Stadtkreise?
Ja, Stadt- und Landkreise können nach Auswahl des entsprechenden Bundesland unter "District" gewählt werden. In den Land- und Stadtkreisen werden als Observationen nur Fälle gezeigt, da auch nur diese im Modell angepasst wurden und weitere Daten oft nicht pro Kreis verfügbar sind. Die Vorhersage von Krankenhausbelegung, Todeszahlen und Genesenen wird basierend auf den Fällen und dem jeweiligen Bundeslandmodell vorhergesagt. Die ICU Kapazität wird nur auf Ebene des Bundeslandes nicht pro Kreis angepasst.

Gibt es auch Simulationen für andere Länder?
In der Online-Version gibt es diese Option noch nicht. Intern wird das Modell auch zur Untersuchung der Infektionsdynamik in anderen Ländern (Frankreich, Italien, Spanien, USA, Schweiz) genutzt. Da die Datenbasis und auch Teststrategien meist recht heterogen sind, werden hier oft Spezialversionen des Modells implementiert. Eine Onlineversion von anderen Ländern kann derzeit aufgrund einer fehlenden Ressourcen nicht realisiert werden.

Was bedeuten die Punkte in den Grafiken?
Punkte sind observierte (gemessene) Daten aus verschiedenen Quellen. Durchgezogene Linien entsprechen den Simulationen.

Warum haben manche Bundesländer vollständige Daten zu Hospitalisierung und andere nicht?
Aus machen Bundesländern werden/wurden keine oder lückenhafte Daten zur Hospitalisierung geliefert.

Werden im Simulator auch Unsicherheiten berücksichtigt?
Die Online-Version ist ein sogenannter Punkt-Schätzer, der die größte Plausibilität (Maximum-Likelihood) der unbekannten oder sich über die Zeit ändernden Modellparameter anstrebt. Es werden nur diese Parameter zu Simulation benutzt. Es werden nun Residualfehler und prospektivische Variabilität für den letzten abgeschätzen R(t) dargestellt.

Wird das Modell in Zukunft mit neuen Daten aktualisiert?
Das Modell wird im Moment jeden Mittwoch mit neuen Daten trainiert/aktualisiert. Bei einer veränderten Infektionslage kann sich auch die Aktualisierungsfrequenz ändern.

Kann man R(t) des Modells direkt mit anderen publizierten Reproduktionswerten (z.B. des RKI) vergleichen?
Ja und Nein. Das hier gezeigte Modell berechnet/schätzt "robuste" R(t)-Werte. Änderungen in der Infektionsdynamik werden in diesem Modell nur manifestiert, wenn eine "signifikante" Änderung über den neuen Observationszeitraum zu erkennen ist. Sieh hierzu Abbildung im PDF Bericht (covid-simulator.com), welche RKI und unsere R(t)-Werte vergleicht.

Wie finanziert sich das Projekt?
Der COVID-19 Simulator wird ohne finanzielle Unterstützung vom Arbeitskreis der Klinischen Pharmazie der Universität des Saarlandes entwickelt. Es gibt keine Verbindung zu Bill Gates, es werden allerdings Microsoft Betriebssysteme und Software Produkte eingesetzt, welche regulär lizensiert wurden.

Wieso wird das Modell von der Klinischen Pharmazie erstellt? Gibt es einen Bezug zur Pharmaindustrie?
Der Arbeitskreis der Klinischen Pharmazie der Universität des Saarlandes arbeitet rein computerbasiert. Im Arbeitskreis werden mathematische Modelle von biologischen Systemen und Arzneimitteln erstellt. Der Leiter des Arbeitskreises, Thorsten Lehr, arbeitet seit über 20 Jahren im Bereich der mathematischen Modellierung. Es werden auch verschiedene Infektionsmodelle auf molekularer und epidemiologischer (z.B. zur HPV Vakzinierung) Ebene erstellt. Es gibt keine Unterstützung zur Pharmazeutischen Industrie für dieses Projekt. Weitere Projekte und Publikationen können auf der Homepage des Arbeitskreises (www.clinicalpharmacy.me) eingesehen werden.

Frequently Asked Questions (Status May 2021)


What kind of model has been implemented here?
The underlying model is a SEIR (Susceptible, Exposed, Infectious, Recovered) model expanded to 27 compartments. Details of the model are shown in the weekly reports (covid-simulator.com). Model code and parameter estimates are deposited on GitHub (github.com).

Which Software was used?
The NONMEM software, version 7.4.3 was used and the non-linear mixed effects (NLME) modeling technique was applied. Ordinary differential equations (ODEs) were used.

Which factors enter the model?
The age and sex structure of infected persons, vaccinations, variances of concern (B.1.1.7), number of tests, and positive rate enter the model as factors influencing the incidence of infection and the course of disease. Details of the model are shown in the weekly reports (covid-simulator.com).

Can all people in the population become infected in the model?
Yes. The model assumes no baseline immunity in the population at the start of the pandemic.

Can people become re-infected in the model once they have recovered from an infection?
No, because the duration of immunity is currently unclear. If the duration of immunity is scientifically proven, it can be incorporated without any problems. In the current pandemic phase, this does not yet affect the simulations.

Does it make sense to simulate very far into the future?
No. A prediction further than a few weeks into the future is associated with considerable uncertainty. The model is used to illustrate the current and retrospective development of the pandemic in Germany. A simulation into the distant future would require a prediction of political measures, behavioral changes, seasonality or age structure of infected persons etc. The model cannot do this.

The model predicts a [VERY HIGH NUMBER] of infected people in the future. Can this come true?
Infection dynamics depend on many factors and future developments (especially in the distant future) cannot be predicted by the model.

What does the "ICU Capacity" indicate?
ICU Capacity shows an approximation of all available ICU beds (including beds that are currently occupied) in each state and Germany. The numbers are taken weekly from the DIVI Intensive Care Register.

What happens if the "ICU Capacity" is exceeded?
The number of ICU beds is only implemented as a guideline. There is no increased mortality in the model if this limit is exceeded. This could be implemented but it is unclear how much mortality increases when the capacity limits are reached. When capacity limits are exceeded, our model underestimates death rates.

Does the model also take into account infected persons who have not been detected ("dark figure")?
The model distinguishes internally between "Exposed" (infected but not (yet) identified) and "Infected" (infected and confirmed by e.g. a test). Since the model was developed to answer questions about hospitalization, no assumptions about the dark rate are implemented, but can be easily implemented according to established scientific knowledge.

Can the model simulate the effect of potential herd immunity (due to infection)?
No. It was not developed for this purpose. The model was developed specifically to answer questions about hospitalization. Worst-case scenarios and (still) unclear long-term immunological consequences after infection are not represented.

Can the model simulate the effect of future population vaccination coverage?
Yes. Vaccination coverage can be set between 50% and 100%. However, this is assumed to be the same for all age groups.

Can the model predict policies/behavioral changes/seasonality/age of infected, etc. into the future?
No. The model simulates into the future using the last estimated kinetics. The R(t) value is currently decreasing due to increasing vaccination. A change in the reproductive value can be selected using the "Infectiousness changepoints" option to simulate actions or behavioral changes, for example. Infectiousness can be increased by up to 120% (e.g., relaxations) or decreased by up to 80% (e.g., restrictions). Appropriate values for changing infectivity must be chosen by the user.

Is it possible to simulate changes in infectiousness dynamics ("Infectiousness changepoints")?
Yes, in the simulator it is possible to determine a change in infectiousness from a certain date in the future ("Infectiousness changepoints" option). Infectiousness can be increased by up to 120% (e.g. relaxations) or decreased by up to 80% (e.g. restrictions). The change is in % because the R(t) value continues to change dynamically due to inoculations.

Is there also simulation for individual rural and urban counties?
Yes, urban and rural districts can be selected after selecting the appropriate state under "District". In rural and urban counties, only cases are shown as observations, since only these were also fitted in the model and further data are often not available per county. The prediction of hospital occupancy, death rates, and recoveries is predicted based on the cases and the respective state model. ICU capacity is only adjusted at the state level not per county.

Are there simulations for other counties?
The online version does not yet have this option. Internally, the model is also used to study infection dynamics in other countries (France, Italy, Spain, USA, Switzerland). Since the data basis and also testing strategies are usually quite heterogeneous, special versions of the model are often implemented here. An online version of other countries cannot be realized at the moment due to a lack of resources.

What do the points in the graphs mean?
Dots are observed (measured) data from different sources. Solid lines correspond to simulations.

Why do some states have complete hospitalization data and others do not?
No or incomplete hospitalization data are/were provided from some states.

Does the simulator take uncertainties into account?
The online version is a so-called point estimator, which aims for the highest plausibility (maximum likelihood) of the unknown or changing model parameters over time. Only these parameters are used for simulation. Residual error and prospective variability for the last estimated R(t) are now presented.

Will the model be updated with new data in the future?
The model is currently trained/updated with new data every Wednesday. If the infection situation changes, the update frequency may also change.

Is it possible to compare R(t) of the model directly with other published reproduction values (e.g. of the RKI)?
Yes and No. The model calculates/estimates "robust" R(t) values. Changes in infection dynamics are only manifested in this model if a "significant" change is seen over the new observation period. See figure in PDF report (covid-simulator.com) comparing RKI and our R(t) values.

How is the project financed?
The COVID-19 simulator is developed without external financial support and solely financed by the Working Group of Clinical Pharmacy at Saarland University. There is no connection to Bill Gates, but Microsoft operating systems and software products are used, which have been licensed regularly.

Why is the model created by Clinical Pharmacy? Is there a connection to the pharmaceutical industry?
The working group of the Clinical Pharmacy of the Saarland University works purely computer-based. Mathematical models of biological systems and drugs are created in the working group. The head of the working group, Thorsten Lehr, has been working in the field of mathematical modeling for more than 20 years. Various infection models on molecular and epidemiological (e.g. for HPV vaccination) level are also created. There is no support by the pharmaceutical industry for this project. Further projects and publications can be found on the homepage of the working group (www.clinicalpharmacy.me).