Coronavirus dashboard

On 17 october 2020 there are 233 additional cases reported from the municipal health service Noord-Holland Noord due to an administrative delay.

Due to a technical malfunction there were a number of confirmed cases missing in the reports of 1 september 2020. These have been included in the report of 2 september 2020. This means the new number of confirmed cases on 2 september 2020 (734) includes around 300 additional reports from 1 september 2020 (462).

These figures are supplied daily as open data by '[Stichting NICE]'(https://stichting-nice.nl/covid-19/public/new-intake/confirmed/) (the National Intensive Care Evaluation Registry). The source data refers to the previous 24 hours. The calculation of the ICU admissions is based on patients with a confirmed infection of the COVID-19 virus.

From the source file, we calculate the average admissions over the past three days. Thus, this does not include the number for the current day. The average is obtained by adding up each day’s values (the figures for one day, two days and three days ago), dividing the total by three and rounding off the result to one decimal place.

ICU occupancy rates are obtained daily via the LCPS website. The data includes the number of ICU beds occupied by COVID-19 patients and the number of ICU beds occupied by all other patients. The percentage of ICU beds occupied by COVID-19 patients is calculated by dividing the number of ICU beds occupied by COVID-19 patients by the total number of occupied ICU beds. The total ICU occupancy numbers are not available prior to the 1st of June because certain quality checks have not been performed before this date. Therefore that data might be flawed and is not published by LCPS.

These figures are supplied daily as open data by the Stichting NICE (the National Intensive Care Evaluation Registry), which is a partnership of Dutch ICUs. These figures are continually updated and once every afternoon are uploaded to the dashboard. The average number of daily hospital admissions is calculated using the figures for one day ago, two days ago and three days ago. Data for the present day is not included in this calculation, as the present day is not yet over and registration data is therefore incomplete. Although hospitals strive to register COVID-19 patients as fast as possible, there is usually some delay. As a result, the figures for the previous several days tend to be incomplete. That is why data on the dashboard is refreshed daily.

Previously, the figures on the dashboard concerning hospital admissions were provided by open data from RIVM, which comes from the Osiris database. However, this database suffers from significant underreporting. The NICE database is more complete and uses a broader definition of hospital admission. RIVM figures include only those patients admitted to hospital because of COVID-19, whereas NICE figures also include patients who were admitted to the hospital for another reason but also have COVID-19.

NICE figures cannot currently be broken down by safety region or municipality. Data at these levels is therefore still taken from the Osiris database. Work is being done to provide NICE figures for hospital admissions at safety region and municipal levels.

Hospital occupancy rates (non-ICU) are obtained daily via the LCPS website. The total occupancy numbers are not available prior to the 1st of June because certain quality checks have not been performed before this date. Therefore that data might be flawed and is not published by LCPS.

This data is supplied daily as open data by the RIVM. The file is published online every day at 10.00 a.m. and registers the number of confirmed cases (people who tested positive) reported to the RIVM in the past 24 hours. The date used is the date of notification. This is not the same as the day on which people were tested.

The RIVM can retroactively correct data that was previously published. These corrections are then applied to the open data file on their site and will automatically be shown on the dashboard.

The source file contains figures per safety region and municipality. To obtain the national figures, all reports are added. In the sourcefile, not all reports mention a municipality or a safety region because this information is not available. These cannot be shown on that lower level, but will be shown in the national figures.

We use the absolute number of confirmed cases to calculate a relative rate per 100,000 inhabitants. We do this by first dividing the number of confirmed cases by the number of inhabitants in the Netherlands. This produces the number of confirmed cases per 100,000 inhabitants, as shown on the dashboard. The number of inhabitants is determined by this CBS file.

The division of new cases by age group is based on the national figures, because ages are not recorded in the file containing municipal data. The table on the RIVM website in the 'COVID-19 weekly epidemiological update of the Netherlands' may therefore not always match the table on the dashboard.

The alert value for the number of confirmed cases is now added to the dashboard. The value is 7 confirmed cases per 100.000 inhabitants. This means that if this value reached, the number of infections is possibly increasing too fast to keep the novel coronavirus under control. The same value is also used by Germany for example. An alert value functions as a “warning” to examine the situation with urgency. It is possible that this alert value is further refined at another point in time.

The percentage of confirmed cases only concern tests performed by the municipal health services (GGD), and of which a test result is known. However, tests also take place by other parties than the GGD. The data is provided on a weekly basis by the RIVM to the Ministry of Health, Welfare and Sport.

This data is supplied daily as open data by the RIVM. The file is published online every day at 10.00 a.m. and registers the number of confirmed cases (people who tested positive) reported to the RIVM in the past 24 hours. The date used is the date of notification. This is not the same as the day on which people were tested.

The RIVM can retroactively correct data that was previously published. These corrections are then applied to the open data file on their site and will automatically be shown on the dashboard.

The source file contains figures per safety region and municipality. To obtain the national figures, all reports are added. In the sourcefile, not all reports mention a municipality or a safety region because this information is not available. These cannot be shown on that lower level, but will be shown in the national figures.

The increase in the number of daily confirmed cases is calculated by deducting the current day’s cumulative number of from the cumulative number reported the day before. This produces the absolute number of daily confirmed cases reported, as shown on the dashboard.

We use the absolute number of confirmed cases to calculate a relative rate per 100,000 inhabitants. We do this by first dividing the number of confirmed cases by the number of inhabitants in the Netherlands. This produces the number of confirmed cases per 100,000 inhabitants, as shown on the dashboard. The number of inhabitants is determined by this CBS file.

The division of new cases by age group is based on the national figures, because ages are not recorded in the file containing municipal data. The table on the RIVM website in the 'COVID-19 weekly epidemiological update of the Netherlands' may therefore not always match the table on the dashboard.

The alert value for the number of confirmed cases is now added to the dashboard. The value is 7 confirmed cases per 100.000 inhabitants. This means that if this value reached, the number of infections is possibly increasing too fast to keep the novel coronavirus under control. The same value is also used by Germany for example. An alert value functions as a “warning” to examine the situation with urgency. It is possible that this alert value is further refined at another point in time.

These figures are supplied weekly (on Tuesdays) as open data by the RIVM. The values in the file are copied directly onto the dashboard. The reproduction number is not an exact value but a reliable estimate. Reproduction numbers (R) using data from less than two weeks ago are not reliable because the R depends on the incubation period (the time between someone becoming infected and becoming ill) and the reporting delay (the time between someone becoming ill and being reported as a COVID-19 case). That is why the line depicting R stops at a date two weeks in the past.

There is a clear break in the R trend on 11 June. On that date, RIVM began calculating R using the number of daily positive tests and the number of hospital admissions. Before 11 June, the R was based only on the number of new hospital admissions. With a low number of new admissions, the number of daily positive tests is a better indicator of R.

The National Institute for Public Health and the Environment (RIVM) provides the Ministry of Health, Welfare and Sport with information about the COVID-19 situation in nursing homes on a daily basis. This data cannot always be clearly linked to a specific safety region. As a result, the total sum of numbers at the safety-region level may be less than the numbers at the national level. The number of deaths of nursing home residents is based on the date of death. However, when the date of death is unknown that death is not included in this statistic in the dashboard.

The method used to estimate the numbers of nursing homes and nursing home residents infected with coronavirus has been improved. Since 1 July 2020, for every person newly infected with coronavirus, the municipal health services (GGDs) have noted whether that person lives in a nursing home. Until the end of September 2020, no use was made of this data and a different method was used to calculate infection numbers, which led to a significant overestimate of the number of infected nursing homes. As of 29 September 2020, the estimates on the dashboard from 1 July 2020 onwards have been recalculated using the this data and therefore now give a much better picture of the actual number of infected nursing homes and nursing home residents.

Old definition of a nursing home resident: A person was counted as a nursing home resident if, according to the central database OSIRIS: on the basis of their postcode, they can be linked to a registered nursing home or residential care home for the elderly, a setting (place where infection may have occurred) labelled as a ‘nursing home’ or something similar, or if there is any other term that links them in the database to a nursing home or residential care home for the elderly. In addition, the person:

• must be over 70 years old
• must not be a health worker, and
• must not be employed.

New definition of a nursing home resident: As of 1 July 2020, every time a person tests positive for COVID-19, the GGD asks whether that person is a resident of a nursing home or residential care home for the elderly. This information is used in the new definition. A person is considered to be a nursing home resident if, according to the data in OSIRIS:

• This person is a resident of a nursing home or a residential care home for the elderly.
• If it is not known whether the person lives in a nursing home or residential care home for the elderly, the old definition is used.

The total number of nursing homes per safety region is provided by RIVM. Any nursing homes with the same postcode are counted as one nursing home. This is because the number of infected nursing homes is also determined by postcode. It is necessary to know the total number of nursing homes per safety region in order to calculate what percentage of nursing homes in the safety region is infected. The total number of nursing homes nationwide is the sum of the total number of nursing homes in all safety regions.

Here the number of nursing homes with COVID-19 outbreaks is estimated, by linking the postal codes of residents with COVID-19 with a list of nursing homes. This produces around 15 addresses. The measure is not entirely accurate as precise addresses cannot be used yet. RIVM is working with the municipal health services (GGDs) and Statistics Netherlands (CBS) to develop a more accurate method. The information on the dashboard will be updated as soon as this method is available.

Sometimes there seems to be a discrepancy between the new number of nursing homes with infected inhabitants, and the increase in the total amount of infected nursing homes compared to the day before. This is due to the fact that by newly infected nursing location, only the number is shown from the previous day. However, new infected nursing homes are not only reported from the day before, but sometimes earlier as well. As a consequence "the number of new locations since yesterday" is not always the same as the increase in total number of infected nursing homes. We are currently working to show this more clearly on the dashboard as soon as possible.

These figures are supplied weekly (on Tuesdays) as open data by the RIVM. The values in the file are copied directly onto the dashboard. We also convert these absolute numbers into a relative rate of infectious people per 100,000 inhabitants. We do this by dividing the Netherlands’ total population by 100,000. Then, we divide the absolute number of infectious people by the outcome of that division. The resulting figure is the number of infectious people per 100,000 inhabitants.

The RIVM bases the number of infectious people on the daily reports of the number of people with COVID-19. Until 30 June this value was based on the daily number of patients admitted to an ICU with COVID-19. Because the number of people currently being admitted to ICUs with COVID-19 is low, the number of people reported as having COVID-19 is a better indicator to track the progression of the number of infectious people.

RIVM calculates the number of infectious people based on the number of infected people and the duration of the infectious period. The duration of the infectious period differs from one person to another, but is between 6 and 10 days. The number of infected people is based on the number of COVID-19 cases reported. RIVM uses a conversion factor to take account of people who are infected with coronavirus but who remain unreported. The conversion factor is derived from confirmed COVID-19 hospital admissions and the percentage of people whose blood contains antibodies to the virus that causes COVID-19. RIVM is considerate of the fact that a tiny percentage of the tests could give a false positive result.

There is no alert value for the number of infectious people as this number is an estimate based on a calculation.

A representative sample of around 350 GP practices across the Netherlands provide data once a week to the Netherlands Institute for Health Services Research (Nivel) Primary Care Database. They record reported and observed symptoms and diagnoses of the consulting patients they see. Using this data, Nivel determines how many patients saw their GP in the past week for the first time for symptoms that could indicate COVID-19. Nivel determines this on the basis of the diagnostic codes reported by GPs (‘acute upper respiratory system infection’, ‘other respiratory infection(s)’, ‘influenza’, ‘pneumonia’, ‘other viral disease(s)’, ‘other infectious disease’, ‘fever’, ‘shortness of breath’, ‘coughing’) and any additional information provided that could indicate COVID-19.

To improve this indicator’s accuracy, Nivel also recalculates the data for the preceding weeks, so they include information that only became available at a later time. The Nivel Primary Care Database provides this information once a week (Thursdays) and delivers this information in JSON format to the Ministry of Health, Welfare and Sport. The source file gives the numbers per week. On the dashboard, these are shown as the weekly figures indicated by the date of the last day (Sunday) of each week. The weekly values in the file are copied directly onto the dashboard.

It is possible to test samples of sewage water for pathogens like the novel coronavirus, because they enter the sewage system via the faeces. Once a week at each site, researchers from the National Institute for Health and Environment (RIVM) examine samples of sewage water collected over a period of 24 hours. These tests show how many virus particles are in the sewage water. Each week the number of sites may vary, because not every test can be completed successfully. Please note that the values from before 20 July are based on about 28 sites. The figures between 20 July and 31 August are based on about 80 sites. From 31 August sewage water from about 300 sites is examined. In the current research we can see trends in the amount of virus particles, like an increase or decrease. We cannot yet compare the number of particles found at different sites. More research is necessary in order to be able to draw conclusions based on the findings.

Once a week (on Tuesdays), this information is delivered as an open data file by the RIVM. With the availability of more data, the number of virus particles can be determined more accurately. Therefore, the RIVM has adapted their method to calculate the virus particles per milliliter since 18 August. In the open data file you can see if a measurement in representative of not. For the dashboard we only use representative data.

The graph’s y-axis shows the concentration of virus particles in one milliliter of sewage water. The x-axis shows the date on which the samples were collected. The national and regional average are weekly averages. The sum of the reported values in the week is divided by the number of sites that completed the tests that week. The reported weeks are from Monday up and until Sunday. These weekly averages are shown in the graph on the Monday of the week.

The virus particles per milliliter shown on the dashboard can be influenced by a number of other factors. For instance, due to rainfall, the number of virus particles per milliliter can be decreased, because the sewage water is diluted. Furthermore, industrial areas in the catchment areas can bring about this effect, because in the sewage water of industrial areas little to no amount of human faeces can be found. If the concentration of virus particles is too low, it will no longer be found by the test. This will then result in a measurement of zero particles per milliliter.

To determine which location belongs to which safety regions, we look at the catchment area. This is the area for which the sewage treatment plant purifies the water. The RIVM shares per site the percentage that the catchment area is part of the safety region. This is not entirely accurate and will be improved on in the future. Some sites have a catchment area that contains multiple safety regions. If this is the case, the site is attributed completed to the safety region covering the largest part of the catchment area.

To match the sites to the municipalities, we currently use the postal code of the site. Using the Statistics Netherlands (CBS) allocation of municipalities, these postal codes are translated to a municipality. In a few cases, no match was found. For these exceptions the municipalities are established by hand, using coordinates of the sites and a map. Please note: the matching of these sites with municipalities can be partly inaccurate, because the catchment area of purification site is not necessarily similar to the borders of the municipalities. This matching between sites and municipalities is currently being improved on.

The tiles on the dashboard show the term "value of" to indicate the date to which the number refers, while ‘obtained’ is used to refer to the date when the information was received. The purpose of providing this information is to give users a better idea of how up-to-date the data is, which is especially relevant for retrospective indicators. For instance, RIVM publishes the reproduction number (R) once a week, but the most recent R is always for a date two weeks in the past. ‘Value of’ is followed by that date two weeks ago, and ‘obtained’ by the date on which the most recent R was published (the current week).

This dashboard was developed in a short amount of time. So far, we have therefore not been able to publish the code for inputting the data and constructing the .JSON files. However, this will be done soon. The web application code was published on Github on 5 June 2020. The data calculations are available via Github as well.