Impact of non-pharmaceutical interventions (NPIs) to reduce COVID19 mortality and healthcare demand

This paper models the spread of COVID-19 in the UK and USA populations and considers 1) the strain the disease will put on healthcare systems, 2) the degree to which various interventions might slow its spread. The paper considers two broad strategies, mitigation and suppression, but notes that the cost of suppression may be extreme and it may not be achievable in all circumstances.

mitigation and suppression descriptions

The report contains details of the model assumptions (information is still very incomplete so the real values of key parameters are not known precisely) and tests a variety of parameters. Some of the more important points are summarized below.

In this graph a “no mitigation” strategy is projected to result in 510,000 deaths in UK and 2.2 million deaths in USA - with the demand for critical medical care being exceeded by over 30 times the availability. The red line at the bottom represents Intensive Care Unit places, where the line on the graph is above the red line, that means people would require intensive care but be unable to obtain it.

The other lines represent predictions when a range of social distancing measures are introduced. Even if they are all introduced, without suppressing the virus the intensive care capacity is expected to be exceeded by 8x. With these mitigation strategies the virus is projected to have a single peak in which up to 80% of the population become infected, which begins tailing off and goes back beneath ICU capacity by the end of August 2020.

When the models parameters were updated with information from Italy, and they started producing this kind of prediction, that seems to be when the UK Government decided to pursue a more aggressive approach of suppression.

This figure shows that if suppression is effective and ICU capacity is not exceeded (by much), then once the distancing measures are stopped we should expect to see infection rates begin to climb again, resulting in a peak which is of similar scale to the mitigation strategies, just later.

This is a model of how phased on-off extreme distancing restrictions might be applied in response to triggers (reliant on good testing coverage), so that measures can be relaxed during times of low infection rates and reintroduced when rates cross a threshold. The end point for this strategy would not be reached until a vaccine or other means can be developed to halt spread of the virus indefinitely. Without a vaccine, most of the population would remain susceptible to the virus.


Social data scientist. Peer Production fan.