RIVMNational Institute for Public Health and the Environment uses models to calculate how quickly an infectious disease spreads among a group of people. A model offers a simplified representation of the actual reality and may be used to investigate the potential impact of measures. It is important to be cautious about drawing conclusions based on models since there are always limiting conditions and relevant factors to consider. This is especially true for new diseases such as the novel coronavirus, since many aspects related to the virus are still unknown. The more we know about the virus and how it spreads, the better the models will correspond to reality. To account for the unknowns, we base our calculations on many different options. As we gain more knowledge, we can eliminate options that are no longer applicable. The remaining options that we calculate are then more probable.
The most important starting point of the model is that the disease is contagious. For our calculations, we rely on factors such as how contagious the virus is and how it spreads. For example: How long does it take for someone to become contagious? How long is a person actually contagious? How many other people does one person infect?
Open and transparent research
RIVM is conducting research on the novel coronavirus that includes the use of models. When the studies are completed, the results are published. RIVM publishes research in international peer-reviewed journals. This leads to open access publications: publications that are accessible to everyone. The codes and data we use will also be publicly available. RIVM shares data and research results with all kinds of organisations in the Netherlands and abroad. Internationally, RIVM exchanges data with the World Health Organisation (WHO) and the European Centre for Disease Prevention and Control (ECDC), and with health institutes in other European countries.
Crisis management and policy recommendations based on models
RIVM identifies outbreaks and threats of infectious diseases and advises on how to control them. On the basis of calculations made by researchers using the models, RIVM provides recommendations on the control measures. Results and insights are shared during technical briefings in the House of Representatives. On those occasions, RIVM also answers parliamentary questions about the models used.
What does RIVM calculate and how?
The reproduction number (R) measures how quickly the number of infected people increases or decreases. The number represents the average number of people infected with COVID-19. R is an important number to see how fast the virus is spreading in the Netherlands, and to determine the intensity of the measures needed to prevent its further spread.
To calculate the reproduction number (R), we need to know how the number of cases of COVID-19 is increasing over time. That can be determined by looking at the numbers of patients with COVID-19 who are admitted to hospital, classified according to their first day of illness. We also need to know the time between the COVID-19 patient’s first day of illness and the first day of illness for the person who infected them. Based on data from the Netherlands, we know that this is 4 days on average. RIVMNational Institute for Public Health and the Environment uses an established method to estimate the reproduction number.
As a metric for defining disease spread, hospital admission data are always delayed, because it is not yet clear on a patient’s first day of illness if that patient will be admitted to hospital. Furthermore, reports of hospitalisations are submitted or updated with a delay. We take both forms of delay into account in our models. The estimates for R from before 14 days ago are reliable; the R estimate for the past 14 days involves more uncertainties. That is why we do not look at “today’s R”: it is too uncertain to be reliable.
The data used to calculate R are:
- Data from the National Intensive Care Evaluation (NICE) Foundation. These are figures on the number of hospital admissions in the Netherlands and the number of admissions to intensive care.
- Data from the municipal public health authorities (GGDs). The GGDs enter information about new patients with COVID-19 in the OSIRIS system. These details include the number of new patients, their first day of illness, the possible source of infection, and the number of people admitted to hospital. Via OSIRIS, all the data provided can be accessed by RIVM. RIVM manages the system and works to provide access to the data. The published data must comply with privacy laws as set out in the General Data Protection Regulation (GDPR).
RIVMNational Institute for Public Health and the Environment looked at what the effect would be if the Netherlands opened primary and secondary schools simultaneously or consecutively. The risk of outbreaks in schools was lowest when primary schools opened first, followed by secondary schools. We did this research in collaboration with the London School of Hygiene and Tropical Medicine.
From 20 March on, RIVMNational Institute for Public Health and the Environment has been providing weekly forecasts for IC occupancy. These projections are based on a distribution model in which the population is divided into groups of the same age, and within these groups further subdivided on the basis of infection status:
- The people who are susceptible to infection
- The people who are infected with the novel coronavirus
- The people who were previously infected with the novel coronavirus and are no longer susceptible to infection
In the model used to calculate the possible spread of the virus (the transmission model), the likelihood that two people will have potentially contagious contact depends on their age. To determine this probability, we sent surveys to a randomly selected, representative group of residents in the Netherlands . These people reported their age, and stated the number of different people they had contact with every day. Similar studies were conducted in the Netherlands in 2007, 2017 and 2020. This makes it possible to estimate the number of contacts made within and between age groups. The method used for this is accessible to everyone.
In the model, a person can become infected after contact with another person who is infected with the novel coronavirus. The probability of becoming infected is defined such that the number of estimated IC admissions corresponds as closely as possible to the actual number of IC admissions. The risk of becoming infected may change over time. If a person is infected, their age determines their risk of developing symptoms, their risk of ending up in hospital, and their risk of being admitted to the ICU or dying. If a person is hospitalized or admitted to the intensive care, then the duration of their hospitalisation or ICU admission is defined such that it corresponds to the data provided by the National Intensive Care Evaluation (NICE) Foundation. For this purpose, we fit a distribution to the observed admission figures . The NICE admission figures for the last few days are often incomplete. We correct for the reporting delay.
By now, we know quite a lot about containing the spread and flattening the curve of the COVID-19 epidemic. The chart above shows calculations for the effect of the measures on ICU capacity in the Netherlands. The red line represents a scenario in which the Netherlands does not take any measures. The thick blue line on the chart represents the most likely scenario. The blue shaded region indicates the margins of uncertainty. The margins of uncertainty are smaller than they were at the beginning of the outbreak because we know more about how disease spreads now. The actual line is somewhere within that blue shaded region.
What safeguards are in place to ensure the quality of the models?
RIVMNational Institute for Public Health and the Environment offers its models and research results for publication in international peer-reviewed journals. International experts then assess the manuscript, the models, and the results. These experts often provide comments and feedback anonymously; the model and the results are only published once the experts have approved them. When the results are published, the codes and data that we used are also released to the public.
Speed is important for purposes of crisis management and policy advice. In order to guarantee the same quality, we build on models and algorithms that have already passed the peer review process. RIVM shares and compares this information with all sorts of organisations in the Netherlands and abroad, including the World Health Organisation (WHO) and the European Centre for Disease Prevention and Control (ECDC), and with health institutes in other European countries. Many of the policy recommendations also have added value for research purposes. Also, after the models and results are used for crisis management and policy advice, they are then used for further research, which is then subjected to peer review before publication.