It is important to realize how our own behavior influences the models’ output. Grant Anderson from the YouTube channel 3Blue1Brown created a fine movie to demonstrate what the effect is when, for example, 10% of the population does not observe the lockdown measures; or what the effect would be if infected people would immediately be quarantined, etc. Even though the models in the movie are simplifications of reality to a large extent, and make strong assumptions (e.g., everyone’s behavior is similar), the movie gives a clear feel for what is happening. Worth the watch! https://www.youtube.com/watch?v=gxAaO2rsdIs
What comes out of epidemiological model depends largely on the value or distribution of key epidemiological quantities. For example, in the individual-based model: How often does one go to the grocery store and to what extent does a person talk to other people? In the meta-population model: What fraction of the people in a given town can be considered immune because they already got the disease and recovered? Which groups move between different towns?
Such parameters are entered into the models to investigate the effect of potential measures. For a number of parameters, a range of values can be assumed, so that a good picture emerges regarding the impact of measures, to variable degrees. But how do we get reliable information about the parameters? Based on the survey in which you participate on Tuesdays, we gather a good picture regarding how many people develop symptoms, what age category they belong to, their gender, their contact patterns, etc. The more people participate in the survey, the more accurate the results are. A link to the survey at the University of Antwerp web site: https://www.uantwerpen.be/nl/projecten/corona-studie/ (red button). At this website you can find interactive graphics on this particular data: https://corona-studie.shinyapps.io/corona-studie/.
In the context of the international EpiPose project (referring to efforts to pause the COVID-19 epidemic), an online survey was started in the United Kingdom, the Netherlands, and Belgium, to monitor the frequency and type of social contacts at various time points and for a range of ages. The first Belgian results are expected on April 23, 2020. On http://www.socialcontactdata.org, social contact data from various countries are shared. Evidently, internationally sharing data can help counteract the COVID-19 pandemic (see here for paper).
Further, we use hospital-reported data, such as the number of beds taken, the number of intensive care units in use, etc. For the compartmental transmission models, it is important to know how many people have acquired immunity against COVID-19 because they already went through a disease episode, perhaps unbeknownst to them. The number of confirmed cases is unquestionably an underestimate of the number of people with acquired immunity. Serological studies are being undertaken in Belgium to quantify the population’s immunity; they are based on blood samples collected as part of routine care (i.e., not specifically for COVID-19, but rather in the context of, for example, routine pregnancy monitoring). The SIMID team provides advice regarding the design and analysis of a serological study. To be precise, the study team calculates sample sizes per age category so that age-specific infection rates can be estimated with sufficient precision. Our ‘in house’ immunologist Joris Vanderlocht provides the required immunity know-how; the team also counts on the expertise of Johan Neyts (KU Leuven) and his research group. This is important to examine how the epidemic would evolve in case antiviral products would be available for administration to patients. Antiviral medicinal products slow the virus down in an infected person, so that infectiousness would decrease. The team tested the effect of antiviral means based on simulation models that simulate the viral spread (similar in spirit to the transmission model used in the meta-population context) when also contact tracing is applied. In other words, in this scenario, it is assumed that we know with whom the infected person was in contact, and that each one of these people in turn is followed for symptoms; isolation, quarantine, and/or whether a COVID-19 test are also parts of this strategy. Recent results indicate that powerful antiviral means combined with intensive contact tracing can be effective to get the epidemic under control and to keep it under control (see here for paper). This follows from the fact that, in such a scenario, there are less infections and less local resurgences of the virus than what would be seen if there were no antiviral means.
This approach is based on a number of assumptions regarding the actual behavior of the virus (e.g., how long lasts the incubation period, what fraction of infected people becomes symptomatic) and regarding the activity and availability of antiviral means. Clinical trials are ongoing to this end.
Figure from Torneri and Libin et al. (2020, MedRXiv preprint). On the Y-axis we see the total number of infections in a simulated population of 500 people. The X-axis displays various levels at which contact tracing is possible (the higher the value, the more contact tracing data are available). The colors refer to various scenarios for which the number of infections is calculated: yellow for the scenario without antiviral products and without testing when symptoms manifest themselves; green for the scenario without antiviral means but with testing when there are symptoms; blue for the scenario where antiviral medication is administered upon a positive COVID-19 test.
Calculation of disease burden and economic impact
Researchers from the University of Antwerp take the lead in the analysis of COVID-19’s impact on the burden on the health care system as well as on the economy (see paper and opinion piece). At the site www.covid-hcpressure.org, the burden for various countries is updated every hour, based on a compound capacity measure that takes into account, among others, the number of physicians, nurses, and intensive care units, relative to the number of COVID-19 cases and deaths. Various scenarios are considered when calculating the disease burden in an effort to quantify the impact of measures taken.
The economic impact in a number of European countries is being followed using media reports and market indicators across various sectors, intervention of the central banks, and data regarding employment, public finance, sales data, etc. At European level, also data on interventions by the European Central Bank are taken into account.
A hotspot for clinical trials
The statisticians at the Data Science Institute at Hasselt University are grouped in the research group CenStat. This group, in turn, is the UHasselt entity of the Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat; www.ibiostat.be), a joint venture with KU Leuven.The Leuven entity, termed L-BioStat, is active in the clinical studies side of the fight against COVID-19. This comes to no surprise, given the proximity of the UZ Gasthuisberg
At KU Leuven, the DAWN (Direct Antivirals Working Against nCoV) consortium was founded that coordinates all studies towards COVID-19 therapy. Within the consortium, various therapies are investigated in a parallel fashion. First and foremost, it is explored whether known therapies, such as Itraconazole or Azithromycine, can suppress viral replication. It is further studied how hyper-inflammation, frequently observed in COVID-19 patients, can be avoided. Finally, certain studies are targeted at investigating whether reconvalescent plasma obtained from recovered COVID-19 patients can help new patients towards faster recovery. Most studies are conducted in collaboration with other partners, such as the Red Cross, the KCE, but the coordination always rests with a Leuven-based expert, such as Geert Meyfroidt.
Geert Verbeke, head of L-BioStat, represents the institute in the consortium’s steering committee. His team supports the DAWN consortium in the preparation of study protocols, the calculation of sample sizes, the development of randomization schemes, the analysis of interim data to support the Data Monitoring Committee (DMC), as well as negotiations with external partners such as the Federaal Agentschap voor Geneesmiddelen en Gezondheidsproducten (FAGG). IN addition, the team provides advice to clinical experts regarding the choice of study design and primary endpoint.