Modelling data

Summary

A request was made for modelling data regarding hospital admissions and deaths from reduced border controls, to which the Cabinet Office responded that the specific information was not held, though they provided a preliminary report from the University of Warwick outlining modelling assumptions and caveats.

Key Facts

• The Cabinet Office stated the requested information was not held.
• Modelling was conducted by the University of Warwick (Louise Dyson, Edward Hill, Michael Tildesley, and Matt Keeling).
• The model assumes no control measures are in place to simulate the impact of reduced border restrictions.
• The analysis covers scenarios for the B.1.1.7 (Kent) variant and notes potential risks from the B.1.351 (South African) variant.
• The report explicitly states these are scenarios and assumptions, not predictions or forecasts.

Data Disclosed

• 1 January to 30th June
• 17/04/2021
• 60%
• 85%
• 65%
• 80%
• 90%
• 14 days
• 24th to 30th March
• 11 weeks
• 18th February 2021
• B.1.1.7
• B.1.351

Original Request

I would request copies of any modelling undertaken by external agencies, such as universities, for Isle of Man Government, in the period 1 January to 30th June with respect to the possible numbers of hospital admissions and deaths resulting from reduction in border controls

Data Tables (2)

Full Response Text

Modelling analysis of potential future outbreaks on the Isle of Man: Preliminary Report, 17/04/2021 Louise Dyson, Edward Hill, Michael Tildesley and Matt Keeling, University of Warwick

Questions 1. How would outbreaks now and in the future compare to the February outbreak? 2. What resulting hospitalisations and deaths would we expect from such outbreaks? Data description & processing We used:
● age-stratified cases data over time (age-gender-travelhistory-data.csv from https://covid19.gov.im/about-coronavirus/open-data-downloads/),
● vaccination data over time (covid19-vaccination-record-open-data.csv), and ● population data (2016 Census by quintile.xlsx). Modelling assumptions We make the following assumptions regarding vaccine efficacies, with each efficacy taking effect 14 days after the dose is administered.

Pfizer AstraZeneca

Dose 1 Dose 2 Dose 1 Dose 2 Efficacy against infection 60% 85% 60% 65% Efficacy against hospitalisation / death 80% 90% 80% 90%

Note that the majority of the efficacy is gained from the first dose.

We assume that the vaccine rollout continues with the number of vaccine doses administered weekly matching that from 24th to 30th March. Doses are split into Pfizer and AstraZeneca, and second doses are given 11 weeks after the first dose, if possible. When the number of doses available is insufficient to achieve this we assume any second doses are given at the next opportunity. We assume 90% coverage for all 5-year age groups (for adults aged 18 and over), except for age groups where recorded coverage is already higher than 90%. Modifications to the speed of vaccine roll-out may be needed to reflect changes in capacity or to capture the discontinuation of the AstraZeneca vaccine in younger age groups.

We take the POLYMOD age mixing matrices for the UK, and the model structure as given in Keeling et al. (2020). We also use parameters from the UK model fits, but assume that there are no control measures in place (i.e. adherence=0 in the Keeling et al. model). This represents a situation in which there are no restrictions and imported cases will cause an outbreak. We assess the severity of this outbreak by calculating the R number, including the immunity generated up to that time by both prior infections and vaccinations. We scale the susceptibility to give a specified value for R on 18th February 2021, for the most recent outbreak.
Caveats This work presents scenarios, not predictions or forecasts. That is, we give estimates for what might happen in different potential future situations with assumptions for how the government and population might act. These are not predictions of what we expect to happen on any particular date.

This work makes the assumption that the population of the Isle of Man is currently mixing in a similar way to pre-pandemic UK, and that the fitted parameters, age-dependent hospitalisation probabilities and stay times in hospital, taken from the UK, can be used for the Isle of Man.

We assume no immunity as a result of prior infections that do not appear in the case data. That is, we assume no unreported infections and that all infected people took a test that was positive.

The value of R (the reproductive ratio at any point in time) gives information regarding the initial doubling time of infections, but a large outbreak might give rise to a governmental response, and/or a public behavioural response, so that this doubling time would not necessarily continue unrestricted. If focused efforts were made to contain any future outbreaks with local surge testing and/or contact tracing then R could be substantially reduced while these measures are in place. Vaccination will also reduce the hospitalisations and deaths due to these cases.

Our analysis uses our best current understanding of the vaccines. However, new information is accumulating quickly regarding the proportion of onward infection blocked by vaccines and results may therefore be subject to change given new data. These results will also be sensitive to the actual vaccine rollout and coverage achieved. We assume no waning of immunity over time, and no seasonality in virus transmission or population contact patterns.

We assume that the February outbreak and future outbreaks are of the B.1.1.7 (Kent) variant. Other variants may be more transmissible, more severe and/or give rise to reduced protection from previous infections and/or vaccinations. A particular concern is the B.1.351 variant (South African) for which there is some evidence of reduced immune protection (Madhi et al., 2021; Planas et al., 2021; and Kustin et al., 2021).

Further caveats regarding the original Warwick model can be found in the Warwick SPI-M roadmap documents (Warwick Roadmap Scenarios and Sensitivity).
Results Figure 1 (top panel) shows the estimated R (left-hand y-axis) and associated doubling time (right-hand y-axis) for an outbreak on a given date, assuming no further cases before that outbreak, and assuming the vaccinations are given as shown in the bottom plot. Note, given this outbreak was likely to be the B1.1.7 (Kent) variant the estimated R is expected to be significantly larger than observed during the early stages of the outbreak in the rest of the UK during March 2020.

Note that the expected likelihood of importation of cases from mainland UK may be assessed using UK SAGE modelling analyses of the roadmap for easing restrictions. The most recent analyses may be found in the documents for meeting 85 in the March collection of SAGE documents, and these may be expected to be updated for future meetings. These documents also highlight the uncertainties for the future UK epidemic.

Assuming a doubling time of 2 days on 19th February (which scales R for the rest of the epidemic), we do not expect to achieve R<1, even once the vaccination campaign is complete. With our assumed vaccination delivery and uptake, the final first dose is given on 28th April, and the final second dose on 21st July. The protective effect of each dose is assumed to occur 14 days after vaccination, so that the full effect of all the first doses is felt on 12th May and of all the second doses on 4th August.

Each of these key dates is marked with a color on Figure 1 (top plot), and time series of the resultant outbreaks are shown in Figures 2 and 3. The detectable cases time series from 19th February also shows positive test data to assess fit (Figure 2, top right panel, also reproduced at larger scale in Figure S1).

Figure 1: Effects of accumulated immunity due to infection and vaccinations. Top plot shows the value of R over time, assuming no further infections happen before that point, other than those that have already occurred as of 4th April, 2021. Bottom plot shows actual and predicted vaccination rollout as a percentage of the total population (including under 18s).

Figure 2: Time courses for all infections (left column) and detectable (symptomatic) infections (right column) for outbreaks beginning at time points marked in Figure 1, top panel. Time courses are coloured correspondingly to the points marked in Figure 1.

Figure 3: Time courses for cumulative hospital admissions (left column), hospital occupancy over time (middle column) and cumulative deaths (right column) for outbreaks beginning at time points marked in Figure 1, top panel. Time courses are coloured correspondingly to the points marked in Figure 1.

Summary Here we have considered the scale of an uncontained epidemic on the Isle of Man, occurring at different points of the vaccine rollout programme. Our results clearly show that without a range of control measures, the February 2021 outbreak would have generated a large number of hospital admissions (with peak hospital occupancy of over 800 patients) and deaths.

Outbreaks that occur at later times are smaller, due to the protective effects of vaccination. However, even when the vaccination programme is complete (estimated to be 4th August) sizable outbreaks can still occur. We predict that an uncontrolled outbreak happening on or after the 4th August would lead to a peak hospital occupancy of around 40 patients and approximately 80 deaths.

These results are generated with a deterministic model. However, in small populations stochastic (chance) events can play a major role, so precise predictions are impossible - a major superspreader event early in the outbreak could lead to more cases, while behavioural changes in the population, localised surge testing or test-and-trace schemes could reduce the initial wave - although leave the island vulnerable to subsequent outbreaks.

References

Keeling et al. (2020) Fitting to the UK COVID-19 outbreak, short-term forecasts and estimating the reproductive number. medRxiv. doi: 10.1101/2020.08.04.20163782 Madhi, Shabir A., et al. (2021) "Efficacy of the ChAdOx1 nCoV-19 Covid-19 Vaccine against the B. 1.351 Variant." New England Journal of Medicine
Planas, Delphine, et al. (2021) "Sensitivity of infectious SARS-CoV-2 B. 1.1. 7 and B. 1.351 variants to neutralizing antibodies." bioRxiv
Kustin, Talia, et al. (2021) "Evidence for increased breakthrough rates of SARS-CoV-2 variants of concern in BNT162b2 mRNA vaccinated individuals." medRxiv SPI-M document: Warwick Roadmap Scenarios and Sensitivity (2021) SAGE March meeting documents, including SPI-M roadmap analyses in meeting 85 Supplementary figures

Figure S1: Detectable (symptomatic) cases for the February outbreak showing data and model fit.

Government Office Douglas Isle of Man IM1 3PN Telephone: (+44) 01624 686244 Website: www.gov.im/co

Our ref: 1907174 11 August 2021

Dear ###

We write further to your request which was received on 15 July 2021 and which states:

"I would request copies of any modelling undertaken by external agencies, such as universities, for Isle of Man Government, in the period 1 January to 30th June with respect to the possible numbers of hospital admissions and deaths resulting from reduction in border controls"

While our aim is to provide information whenever possible, in this instance the Cabinet Office does not hold the information that you have requested. No modelling was undertaken by external agencies in the period 1 January to 30 June with respect to hospital admission and deaths resulting specifically from reduction in border controls. The University of Warwick undertook modelling to project possible scenarios for the impact of future outbreaks of COVID-19 on the Isle of Man, including hospital admissions and deaths, compared to the February outbreak. In order to provide additional advice and assistance to you, a copy of this report is attached. The report was dated 17 April 2021 and the assumptions/inputs used for it are detailed in the report. The modelling assumed that future outbreaks would, like the February outbreak, be due to alpha variant. The modelling does not, therefore, reflect delta variant which became the dominant variant in the UK (and Isle of Man) after this report. The assumptions for vaccine effectiveness have also changed as a result of delta from those used in the report. It is, of course, now possible to compare the actual progression of spread of COVID over recent weeks with the scenarios generated by the modelling.

Please quote the reference number 1907174 in any future communications.

Your right to request a review

If you are unhappy with this response to your freedom of information request, you may ask us to carry out an internal review of the response, by completing a complaint form and submitting it electronically or by delivery/post.

An electronic version of our complaint form can be found by going to our website at https://services.gov.im/freedom-of-information/Review . If you would like a paper version of our complaint form to be sent to you by post, please contact me and I will be happy to arrange for this. Your review request should explain why you are

dissatisfied with this response, and should be made as soon as practicable. We will respond as soon as the review has been concluded. If you are not satisfied with the result of the review, you then have the right to appeal to the Information Commissioner for a decision on; 1. Whether we have responded to your request for information in accordance with Part 2 of the Freedom of Information Act 2015; or 2. Whether we are justified in refusing to give you the information requested.
In response to an application for review, the Information Commissioner may, at any time, attempt to resolve a matter by negotiation, conciliation, mediation or another form of alternative dispute resolution and will have regard to any outcome of this in making any subsequent decision. More detailed information on your right to a review can be found on the Information Commissioner’s website at www.inforights.im. Should you have any queries concerning this letter, please do not hesitate to contact me. Further information about freedom of information requests can be found at www.gov.im/foi. I will now close your request as of this date.

Yours sincerely

Warwick Modelling – Context

This modelling was undertaken by Warwick University in early 2021 and commissioned by the Isle of Man Government in order to help inform Council of Ministers, the Department for Health and Social Care and Manx Care’s understanding of the impact that Covid-19 could have on the Isle of Man, including projected hospital admissions and deaths, using the February outbreak as a proof point for projections of future outbreaks.

The report, dated 17 April 2021, details assumptions and inputs which were used to formulate the modelling, and the models assumed that future outbreaks would, like the February outbreak be due to the alpha variant, which was dominant at the time.

Warwick modelling demonstrated the severe threat that Covid-19 still posed to the Island at a time where vaccine rollout was in its early phases, and when the alpha variant was dominant, with potential for hundreds of deaths if the virus was left unchecked. These conclusions were a key consideration of the Council of Ministers when responding to the virus outbreak at that time.

As throughout the pandemic, global understanding of Covid-19 has developed over time as we learn more about the virus, how it interacts and as our defences increase. This report is a snapshot in time based on professional best assumptions at this time.

It is therefore important to note that this modelling does not reflect the delta variant which became dominant in the UK and Isle of Man in the weeks following this report. Assumptions for vaccine effectiveness have also changed from those used to develop this mo

[Response truncated — full text is 15,595 characters]