Old prediction model – 2020-03-30

COVID-19 in Iceland

A prediction model for the number of individuals diagnosed with COVID-19 and the corresponding burden on the health care system.

2020-03-30

Summary

The method of estimating the prediction model has been improved from March 25th, but now we assume that the daily number of detected infections is accompanied by a probability distribution that allows a more realistic estimate of uncertainty (variance) from day to day. As a result, the pessimistic forecast has risen.

The prediction of the total number of infections has changed little since March 25th, but forecasts of the burden on the health system have risen, as the age distribution of diagnosed infections in Iceland has changed. In order to prepare for the possibility of further change, there is now another forecast which assumes that the age distribution will be more unfavorable.

The main results of the prediction model, with data through March 29th, are as follows:

  • It is expected that throughout the epidemic, about 1700 individuals in Iceland will have been diagnosed with COVID-19, but the number could reach as high as 2800 individuals, according to the pessimistic forecast.
  • It is expected that the number of diagnosed individuals with an active disease will peak in the first week of April and will at that time probably be about 1200, but could be as many as 1800 individuals according to the pessimistic forecast.
  • It is expected that while the epidemic is ongoing, about 120 individuals will need hospitalization but this number might reach 200 according to the pessimistic forecast.
  • The greatest burden on health care services due to hospital admissions will be around mid-April. At that time it is expected that around 60 individuals will be hospitalized but the pessimistic forecast predicts 100 individuals.
  • It is expected that during the epidemic, approximately 25 individuals will become seriously ill (i.e. hospitalized in intensive care units), but this number might turn out to be as high as 44, according to the pessimistic forecast.
  • The heaviest burden on intensive care units is likely to be in the second week of April, when 10 COVID-19 patients are anticipated to  need intensive care at the same time, or even 18 according to the pessimistic forecast.

The analytical work continues and the prediction model will be updated regularly with new data. It is necessary to keep in mind that as Iceland‘s population is small (about 360.000) the number of confirmed cases can vary greatly from day to day, which will affect the results of the prediction model. The model will, though, become more stable as time passes.

Methods and premises of the prediction model

  • We used a logistic growth model with Poisson distribution of the number of already diagnosed COVID-19 infections in Iceland to predict the median (likeliest prediction) and 97.5% upper limit (pessimistic prediction) of the cumulative number of confirmed COVID-19 cases in Iceland and active diagnosed cases (assuming 21 days of illness) in the upcoming weeks.
  •  The epidemic prediction model assumes exponential growth of diagnosed infections to slow down at some point stop, as the epidemic reaches its peak and the number of active infections subsequently starts to decline.
  • The calculation method used to assess the shape of the growth curve in Iceland takes into account information on COVID-19 epidemic processes in other countries (see annex) to  estimate the possible shape of the process in Iceland. Countries that have progressed further in the epidemic, e.g. South Korea, weigh more than the ones not as far into it.
  • Since all infected individuals in Iceland are clients of the Icelandic health care system, the forecast is based on the total number of infected individuals in Iceland, regardless of the source of infection, whether individuals are diagnosed in quarantine or not, whether they are diagnosed through the screening through the healthcare services or deCODE genetics. It should be kept in mind that infected individuals in quarantine may possibly add less to the exponential growth than other individuals.
  • It should be kept in mind that the age dstribution of infected individuals in Iceland is favorable so far. If the number of infections among the elderly increases, it will significantly impact the prediction model towards an increased burden on the health care system.
  • All code can be found here: website In addition, techincal report on methods behind the development of the prediction model can be found here. Finally, a dashboard measuring the development of COVID-19 in Iceland and elsewhere can be found here.

 

Results

Diagnosed COVID-19 cases

Cumulative diagnosed infections

Cumulative confirmed infections in total

Active diagnosed infections each day

Hospitalization

Cumulative hospitalizations

Cumulative total of hospitalizations

Hospitalizations each day

Number of individuals in hospitals in total

Intensive care

Cumulative intensive care admissions

Cumulative total of individuals in intensive care

Number of individuals in intensive care each day

Total number of individuals in intensive care

Distribution by age

Diagnosed infections

Cumulative

Cumulative diagnosed infections by age

Active

Active infections by age

Hospitalizations

Cumulative

Cumulative total of hospital admissions by age

Active

Active hospital admissions by age

 

Intensive care

Cumulative

Cumulative total of individuals in intensive care units by age

Active

Total of individuals with active infections by age

Results with a different age distribution

Different age distribution

The following is a simulation of development based on the fact that transmission is proportionally more prevalent in older age groups than it currently does:

Hospitalizations

Cumulative

 

Active

 

Intensive care

Cumulative

 

Active each day

Distribution by age

Diagnosed infections
Cumulative

 

Active

Hospitalizations

Cumulative

 

Active

 

Intensive care

Cumulative

Active

Annex

Information on data for prediction model:

Rate
Country First observation Total of days Beginning Now
Albania 2020-03-19 11 0.0204796 0.0683810
Andorra 2020-03-14 15 0.0259262 3.9926370
Armenia 2020-03-18 12 0.0263716 0.1433531
Australia 2020-03-19 11 0.0224178 0.1511316
Austria 2020-03-11 19 0.0203236 0.9258409
Bahrain 2020-02-27 31 0.0201076 0.2882087
Barbados 2020-03-21 9 0.0209041 0.0905844
Belgium 2020-03-10 20 0.0207118 0.7915539
Bermuda 2020-03-20 10 0.0319969 0.3519662
Bosnia And Herzegovina 2020-03-22 8 0.0278703 0.0778552
Bulgaria 2020-03-22 8 0.0232853 0.0472849
Canada 2020-03-20 10 0.0226136 0.1439682
Cayman Islands 2020-03-21 9 0.0461908 0.1231755
Chile 2020-03-21 9 0.0228999 0.1007280
Costa Rica 2020-03-21 9 0.0223870 0.0584441
Croatia 2020-03-20 10 0.0251797 0.1590682
Cyprus 2020-03-16 14 0.0254334 0.1517527
Czech Republic 2020-03-15 15 0.0200202 0.2491298
Denmark 2020-03-11 19 0.0457390 0.3813318
Ecuador 2020-03-21 9 0.0245199 0.1056196
Estonia 2020-03-13 17 0.0203674 0.4827828
Finland 2020-03-13 16 0.0280180 0.2201673
France 2020-03-10 20 0.0216798 0.5769255
French Polynesia 2020-03-21 9 0.0537082 0.1217386
Germany 2020-03-13 17 0.0283655 0.6291770
Greece 2020-03-15 15 0.0217693 0.1013037
Greenland 2020-03-20 10 0.0352908 0.1764540
Iceland 2020-03-02 28 0.0294958 3.2032469
Iran 2020-03-04 26 0.0281738 0.4270454
Ireland 2020-03-15 15 0.0264209 0.4946242
Isle Of Man 2020-03-22 8 0.0236451 0.3783221
Israel 2020-03-15 15 0.0208935 0.4247963
Italy 2020-03-02 28 0.0278943 1.5271988
Kuwait 2020-03-14 16 0.0237694 0.0558582
Latvia 2020-03-18 12 0.0319917 0.1599586
Lebanon 2020-03-20 10 0.0217337 0.0600959
Liechtenstein 2020-03-05 19 0.0263026 1.6044609
Lithuania 2020-03-21 9 0.0250034 0.1427729
Luxembourg 2020-03-13 16 0.0422264 2.9737108
Malaysia 2020-03-18 12 0.0210643 0.0726140
Maldives 2020-03-16 14 0.0244843 0.0301345
Monaco 2020-02-29 18 0.0256647 1.1035828
Montenegro 2020-03-20 10 0.0207011 0.1337607
Netherlands 2020-03-11 19 0.0223429 0.5709730
North Macedonia 2020-03-19 11 0.0201588 0.1156730
Norway 2020-03-07 23 0.0210082 0.7148359
Panama 2020-03-18 12 0.0202523 0.2121778
Portugal 2020-03-16 14 0.0239581 0.5055648
San Marino 2020-02-28 30 0.0295334 6.6154755
Seychelles 2020-03-15 15 0.0204627 0.0716193
Singapore 2020-03-06 24 0.0201573 0.1383448
Slovakia 2020-03-20 10 0.0225398 0.0540589
Slovenia 2020-03-12 18 0.0274216 0.3324267
South Korea 2020-02-26 33 0.0223718 0.1870755
Spain 2020-03-10 20 0.0257613 1.5458490
Sweden 2020-03-09 21 0.0202264 0.3434506
Switzerland 2020-03-07 23 0.0243268 1.5308394
Trinidad And Tobago 2020-03-22 8 0.0351261 0.0530476
United Kingdom 2020-03-16 14 0.0205982 0.2530573
United States 2020-03-19 11 0.0286114 0.3788462
Uruguay 2020-03-19 11 0.0228209 0.0878173

 

The head of the prediction model on behalf of University of Iceland’s Health Sciences Institute is Dr. Thor Aspelund. The prediction model is conducted by scientists from the University of Iceland, the Directorate of Health, and the National Hospital. Contact us: covid@hi.is