key: cord-0803783-ul6u62ml
authors: Sjodin, H.; Rocklov, J.; Britton, T.
title: Evaluating and optimizing COVID-19 vaccination policies: a case study of Sweden
date: 2021-04-09
journal: nan
DOI: 10.1101/2021.04.07.21255026
sha: 4e686e06606eb1d2e0db732e2225353bd4d5633e
doc_id: 803783
cord_uid: ul6u62ml

We evaluate the efficiency of vaccination scenarios for COVID-19 by analysing a data-driven mathematical model. Healthcare demand and incidence are investigated for different scenarios of transmission and vaccination schemes. Our results suggest that reducing the transmission rate affected by invading virus strains, seasonality and the level of prevention, is most important. Second to this is timely vaccine deliveries and expeditious vaccination management. Postponing vaccination of antibody-positive individuals reduces also the disease burden, and once risk groups have been vaccinated, it is best to continue vaccinating in a descending age order.

The COVID-19 pandemic has resulted in millions of deaths and seriously ill people, 21 affecting societies all over the world. Fortunately, effective vaccines (1,2) have now 22

become available and are being distributed in many countries. At the same time new 23 and more transmissible and virulent strains invade. Perhaps most notable, the new 24 lineage B.1.1.7 originating from the UK, which has become increasingly dominant in 25

Europe and Sweden and estimated to have a 39-130% higher reproduction number 26

(3), and has been suggested to cause more severe illness (4). As a consequence, it is 27 now of paramount importance to roll out the vaccines as quickly and effectively as 28 possible in order to minimize the damage from new epidemic waves caused by the 29 circulation of new virus strains to which current control measures are inadequate. 30

Considering the time required for implementation of vaccination to sufficient 31 coverage of the adult population, the vaccine strategy can make a direct difference 32

for the number of people becoming severely ill or dying in COVID-19 in the coming 33 months. 34

The focus of the present paper is to describe and evaluate scenarios and 35 vaccination strategies with respect to incidence and health care load from in countries similar to Sweden up to 1 We assume that 90% of all individuals older than 19 years will take the vaccine 57 when offered, and a that 95% out of this group will be 100% protected by the vaccine 58

and that 5% will have no protection (i.e., an all-or-nothing model with respect to a 59 95% vaccine efficacy). We do not consider vaccination of individuals younger than 20 60 years. People who recover from natural infection are assumed to be 100% protected. 61 We assume no waning immunity. We do not consider changes in disease severity or 62 reductions in vaccine efficacy associated with new strains. 63 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 9, 2021. ; https://doi.org/10.1101/2021.04.07.21255026 doi: medRxiv preprint 

The model is used to simulate a number of different plausible scenarios varying four 76 different factors: 77

I. Transmission scenario: The scenarios follow respectively three different 78 trajectories (see Figure 1 ). These trajectories can be thought of as being 79 affected by the new virus strain taking over, the level of prevention, or 80 seasonal effects They are not affected by additional immunity from vaccination 81 and should hence not be interpreted as " . However, the realized " which 82 arises from the scenario, and which is represented in our computations, takes . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 9, 2021. as Ascending order vaccinates 20-29, followed by 30-39 and so on. A 97 motivation for the Descending order strategy is to prioritize with respect to 98 risk of serious illness, whereas a motivation for the Ascending order strategy is 99 to indirectly protect those at a higher risk by reducing transmission as quickly 100 as possible (10). 101 102

The first factor (Transmission scenarios) has three levels whereas the other three 103 factors all have two levels, resulting in a total of 24 different scenario combinations 104 (see Table S6 in section 2.3 in S.M.). 105

The focus of the analysis lies on learning how the different scenarios affect disease 106 outcomes and how they can inform the development of an optimal strategy. For each 107 scenario we analyse the progress of the epidemic up until October 1 with respect to 108 four different outcomes: infections, regular health care, critical care, and fatalities. 109

The parameters for in the analysis are informed by Swedish data, but the aim is not to 110

give the best possible fit or forecast of the Swedish COVID-19 development. Vaccinating also antibody positive, and Descending age order vaccination. 122

The factor with the largest impact on the number of individuals in critical care is 123

the Transmission scenario (I). More specifically, if T1 happens, or either of T2 or T3. As 124 a consequence, it is of utmost importance to reduce transmission in March and April. 125

In contrast, the difference between T2 and T3 is very minor indicating that reducing 126 transmission later is less significant for the critical care need since by then all 127

individuals above 50 will have been vaccinated (given a descending age order 128 vaccination . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 9, 2021. ; https://doi.org/10.1101/2021.04.07.21255026 doi: medRxiv preprint and elderly starting already in January, but see also Discussion). Second, when 145

comparing Ascending and Descending age order of vaccination, the outcomes for 146 regular healthcare, c ritical care and fatalities are lower (i.e., better) for a Descending 147 age order of vaccination, whereas incidence is lower (i.e., better) for an Ascending age 148 order of vaccination. As a consequence, vaccination in ascending order does reduce 149 transmission more, but not enough to compensate for the large healthcare load as an 150 effect of that individuals in ages 50+ are being vaccinated only later. 151 Figure 4 shows the relative reduction between March 

In Figure 4 we also see that all ranges of relative effects exclude the null effect; 171 meaning that the value zero is not contained in any interval. This result suggest that 172 CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 9, 2021. above the age of 80+, for example in care homes. One explanation could be that the 198 contact rates among this demographic group is larger than contacts otherwise in the 199 same age group, and that also contact rates with some lower age-groups is increased 200 due to frequent home service. This is not well captured in our study. As a 201 consequence, the model underestimated deaths in Sweden, but only until this age 202 group is fully protected by vaccines. Since our study objective was to evaluate the 203 impact of vaccination strategies in ages below 70 years of age, after everyone at age 204 70 years and above had been vaccinated, we consider this mismatch to be of reduced 205

importance to the study findings. We assume a mean of ten days from vaccination to 206 full immunity and do not explicitly model two vaccine doses (1,2) -relating to the fact 207 that the time spent in different states are expressed by constant exit rates leading to 208 exponential durations that may not be realistic. With respect to transmission 209 dynamics we do not consider age-dependent susceptibility or infectiousness (yet age-210 dependent contact rates). Further, we do not consider partial or waning of immunity. 211

Naturally infected individuals are thus assumed 100% protected (11). We note that 212 the latter has implications for the size of the effect of the severity-reduction (i.e., 213 relative reduction) from delaying vaccinations for antibody-positive individuals as 214 such natural protection may last for a more limited time. We do not capture higher 215 severity and case fatality associated with new strains, as reported for B1.1.7 (4). Our 216 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 9, 2021. ; https://doi.org/10.1101/2021.04.07.21255026 doi: medRxiv preprint modelling could thus be improved in several aspects to correspond even better to the 217 real situation. 218

The main purpose of our analysis, however, was not to fit our model to data in the 219 best possible way, but instead to study and learn about what we think are the most 220 central qualitative properties of the future progress of the epidemic under a set of 221 different relevant scenarios. Whereas we believe that our qualitative conclusions are 222 primarily applicable in Sweden, they are likely applicable also to other countries with 223 similar characteristics for COVID-19. We think that future studies could benefit from 224 extending mainly in the directions of: spatial heterogeneity; age-structured degrees 225 of social distancing; and, waning immunity, where the latter would be relevant when 226 consider slightly larger time-scales. Hopefully, this current study can still contribute to 227

better understanding how to plan and act during the vaccination phase in order to 228 best reduce the potential disease-burden in this high-pace race against the pandemic. 229

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 9, 2021. ; https://doi.org/10.1101/2021.04.07.21255026 doi: medRxiv preprint

Oxford-AstraZeneca COVID-19 vaccine efficacy. The 231 Lancet

Safety and efficacy of the BNT162b2 mRNA Covid-19 233 vaccine

Estimated transmissibility and impact of SARS-CoV-2 235 lineage B.1.1.7 in England

Increased risk of hospitalisation associated with infection with 237 SARS-CoV-2 lineage B.1.1.7 in Denmark

COVID-19 healthcare demand and mortality in Sweden in 239 response to non-pharmaceutical mitigation and suppression 240 scenarios

Projecting social contact matrices in 152 countries using 242 contact surveys and demographic data

Svenska intensivvårdsregistret, SIR

Transmission heterogeneities, kinetics, and controllability of SARS-250 CoV-2

Vaccine optimization for COVID-19: who 252 to vaccinate first?

Model-informed COVID-19 vaccine prioritization strategies by 255 age and serostatus