Impact of vaccine coverage and disruption to health services on COVID-19 in Ukraine

COVID-19 surveillance in Ukraine ceased after the Russian invasion of the country in 2022, on a background of low vaccination rates of 34.5% for two doses at this time. We conducted a modelling study to estimate the epidemic trajectory of SARS-COV-2 in Ukraine after the start of the war. We use a COVID-19 deterministic Susceptible-Exposed-Infected-Recovered (SEIR) model for Ukraine to estimate the impact of increased vaccination coverage and masking as public health interventions. We fit the model output to case notification data between 6 January and 25 February 2022, then we forecast the COVID-19 epidemic trajectory in different scenarios of mask use and vaccine coverage. In the best-case scenario, 69% of the Ukrainian population would have been infected in the first half of 2022. Increasing mask use from 50 to 80% reduces cases and deaths by 17% and 30% respectively, while increasing vaccination rates to 60% and 9.6% for two and three doses respectively results in a 3% reduction in cases and 28% in deaths. However, if vaccination is increased to a higher coverage of 80% with two doses and 12.8% with three, or mask effectiveness is reduced to 40%, increasing vaccination coverage is more effective. The loss of health services, displacement, and destruction of infrastructure will amplify the risk of COVID-19 in Ukraine and make vaccine programs less feasible. Masks do not need the health infrastructure or cold-chain logistics required for vaccines and are more feasible for rapid epidemic control during war. However, increasing vaccine coverage will save more lives. Vaccination of refugees who have fled to other countries can be more feasibly achieved.


Model parameters and data
The parameters (Table 1) reflect the transmission of the BA.2 Omicron variant which was the circulating strain at the time COVID-19 surveillance ceased, with a latent period reduced to 3 days.The infectious period is divided into 2 pre-symptomatic days and 7 symptomatic days 3,4,19 .We use R0 of 6 [5][6][7] .In Ukraine four different vaccines were used, Pfizer, Sinovac, AstraZeneca (AZ), and Moderna.It has been shown that all vaccines have a reduced effectiveness against symptomatic infection for the Omicron variant 20 .As the combination of Pfizer and AZ represents just over 60% of the total doses used, in this study we used the estimated vaccine effectiveness of Pfizer and AZ against Omicron.Vaccine effectiveness against Omicron infection 21 is assumed to be 5.9% and 34.2% after 6 months from vaccination following two doses of AZ and Pfizer respectively in adults.We used 71.4% and 75.5% protection against symptomatic infection 21 for AZ and Pfizer recipients respectively following the Pfizer booster dose 20 .For hospitalization, ICU, and death rates, available European estimates (for EU/EEA countries, with the data collected from The European Surveillance System (TESSy) or the GISAID EpiCoV database) are 1.14%, 0.16%, and 0.06% respectively overall, due to the Omicron variant 22 .However, this estimation does not account for vaccination status, and Ukraine has very low vaccination coverage 1 compared to the rest of Europe.Therefore, we used estimates of hospitalization, ICU, and death rates from the US 23 , which are available by vaccination status.Due to the lack of age-specific data on hospitalization, ICU, and deaths, we used the age distributions of the Delta variant from a previous study 19,24 , based on the COVID-19-Associated Hospitalization Surveillance Network (COVID-NET) data on 1482 people hospitalized in March 2022 in the US.
The median length of hospital stay was assumed to be 3, 4.5, and 5 days respectively for people < 50, 50-79, and 80 + years old 25 , and ICU stay is assumed to be 4 days.The total number of hospital beds in Ukraine came from a 2018 estimation of 293,446 beds 26 .This is a maximal estimate that does not account for the loss of hospital infrastructure and staffing during the war.
The implementation of mandatory mask use has been estimated to be 60% effective in reducing transmission if a combination of cloth and surgical masks is used 27 .If mask coverage is 50% and mask effectiveness is 60%, then using masks in Ukraine is estimated to be 30% effective 27 .We estimate that there may be 50% of people wearing masks at any one time as a base case scenario, based on 60% effectiveness, when fitting the model output to data.Mask use is implemented in the model with a reduction in the force of infection by the combination of the proportion of the population wearing it and mask effectiveness.We assume masks, if supplied, can be distributed to the population, but there is no data on the supply chain and logistics for medical items in Ukraine.We used the total population and age distribution from 2022 from Ukraine 28 and an age-specific contact matrix from Ukraine 29 , where we used contact rates estimated for "all locations".Then we used case and death notification data from the 6th of January 2022 to the 25th of February when Ukraine stopped reporting them 1 .We chose to start from the 6 of January as it is the start of the fourth wave attributed to the Omicron variant.As of this date, there were 4711 new cases notified, and as testing was only being done for symptomatic people, we considered this to represent 65% of the cases on that day, based on an estimated 35% of cases being asymptomatic 30 .Therefore, the total number of new infections on that day is estimated to be 7247 and we assumed the non-notified (2536) cases were infectious.The case notification data were fitted with the modelled case incidence using the estimation of restrictions from the Blavatnik School of Government restrictions tracker 17,18 .The nine metrics used to calculate the Stringency Index are school closures; workplace closures; cancellation of public events; restrictions on public gatherings; closures of public transport; stay-at-home requirements; public information campaigns; restrictions on internal movements; and international travel controls.The index on any given day is calculated as the mean score of the nine metrics.Movement restrictions are reflected in the model as a reduction of the contact rate.The model assumes that contact tracing, testing, and isolation of symptomatic people are not occurring or occurring at low levels.The restriction index for Ukraine was estimated to be 63% between the 6th and 17th of January (first pre-war period), then from the 18th of January to 25th of February it increased to 75% (second pre-war period), with the addition of the use of face masks in any public space and the closure of public transport.Those percentages were used in the model as the proportion of contacts reduced in those www.nature.com/scientificreports/periods.The change from 63% to 75% restriction was gradually implemented in the model from the 18th of January to the first of February, to allow some adjustment time in the population to the new restrictions.From the 18th of January, the Ukraine policy changed from testing anyone to testing only symptomatic people 17,18 .
We implemented this policy change in the model reducing the proportion of contact tracing from 60% to 20%, and then further reduced it to zero after the 26th of February when the war began.The vaccination rates as of 25 February, when last reported 1 were 34.5% of the total population with 2 doses and only 1.7% with 3 doses.We distributed vaccine coverage with two doses to the 15 + age group (39.3% vaccinated) while three doses coverage is distributed only in the 60 + age group (6.3% vaccinated), following the policy for vaccination distribution.All the parameters used in the model are listed in Table 1.

Model outputs
Firstly, we tested the model fit to notification data of cases, using the estimated restrictions from the 6 th of January to the 25 th of February.The model used overall death rates reported from European and US studies, of 0.3%, 0.08%, and 0.07% in unvaccinated, vaccinated with two and three doses, respectively.In this period, we assumed 50% of mask use and 60% to 20% of contact tracing (as detailed in Table 1) while estimating the reduction of contacts (movement restrictions).Then we forecast the epidemic curve.After the 25th of February, we assumed a small reduction (of 20%) in movements due to the unstable situation and people sheltering in place.This is because the number of contacts between different people would drastically be reduced, however, there is no data available on the impact of war on contact patterns.The model accounted for uncertainty with sensitivity analyses on mask use coverage and vaccination coverage.The base case scenario uses the last notified vaccination coverage and assumes 50% of people wearing masks.Results are shown with varying assumptions about mask use (0%, 50%, 80%) and varying vaccination rates from 39.3% (2 doses) and 6.3% (3 doses) as the base case, to 60% and 9.6%, and 80% and 12.8% respectively vaccinated with two and three doses in 15 + and 60 + age group.We included the scenario of 0% mask use to then estimate the averted cases when masks are used and show results with no mask use.We also tested decreasing mask effectiveness from 60% to 40% in the case of only poor-quality masks being available.

Results
Figure 1 shows the modelled epidemic around the time of the start of the war in early 2022 fitted to observed data to estimate the reduction in population contacts.In the base case scenario where masks are 60% effective, the best fit to data shows a reduction in contacts of 33% in the first period and 47% to 70% in the second period (Fig. 1).Otherwise, in the case of masks being effective at 40%, the best fit resulted in a reduction in contacts of 40% in the first period and 53% to 73% in the second period.
The modelled incidence of deaths in the same period, using reported rates from European and US studies, of 0.3%, 0.08%, and 0.07% in unvaccinated, vaccinated with two and three doses respectively (Fig. 2 green line), is much lower than rates reported in Ukraine (Fig. 2 blue line).Multiplying those numbers by 5 times, produces a much better fit (Fig. 2 red line), suggesting an under report of case numbers or higher death rates for Ukraine compared to the ones estimated in EU and the US used.
The results of sensitivity analysis on mask use, based on the last reported vaccination rates (39.3% of the 15 + age group with two doses and 6.3% of the 60 + with 3 doses), are shown in Figs. 3 and 4. The epidemic forecast is shown in Fig. 3, with hospitalization and ICU daily bed requirements in Fig. 4. In each scenario, the epidemic peak was expected to be at the start of April 2022.
The outbreak peaks at about 3.7, 2.3, and 1.4 million cases, with a total of almost 90%, 80%, and 70% of the population being infected at the end of the outbreak, with 0%, 50%, and 80% of the population using masks, 6 January is the start of the Omicron wave; 6-17 January is the first pre-war period with the implemented restrictions policy 17,18 ; 18-25 January is the second pre-war period where the reduction in contacts increases with the additional policy of closure of public transports and enforced mask use in all public spaces 17,18 ; 25 February represent the start of the war, where vaccine rollout and tracing stops.respectively (Fig. 3). Figure 4 shows that the maximum number of daily hospital beds required at the peak is estimated to be about 140, 103, and 69 thousand in scenarios of varying mask use, with a total of almost 300,000 beds available in Ukraine before the war started.The number of daily ICU beds required at the peak is estimated to be about 21, 15, and 10 thousand with 0%, 50%, and 80% mask use.
The results of the sensitivity analysis on vaccination coverage are shown in Figs. 5 and 6. Figure 5 shows the epidemic forecast and Fig. 6, the hospitalization and ICU daily bed requirement with mask use at 50% and vaccination rates increased from 39.3% to 60% and 80%.
The outbreak peaks at about 2.3, 2.1, and 1.8 million cases, with a total of about 80%, 76%, and 74% of the population being infected at the end of the outbreak, at the three different vaccination scenarios (Fig. 5).The maximum number of hospital beds required at the peak is estimated to be about 103, 80, and 56 thousand in each scenario, while requirements for ICU beds have been estimated to be about 15, 11, and 7 thousand (Fig. 6).
From the base case scenario, the model shows increasing mask-wearing from 50% (base-case) to 80% could result in a 17% reduction in cases (from a total of 33,432,800 at 50% to 28,006,300 at 80%) and a 30% reduction in deaths (from 56,028 to 39,241).If vaccine coverage is increased from 39.3% and 6.3% with two and three doses respectively (base-case) to 60% of people aged 15 + with two doses and 9.6% of people aged 60 + with three doses, the reduction in cases and deaths could have been 3% (from 33,432,800 to 32,365,700) and 28% (from 56,028 to 39,867) respectively.However, when comparing the results of increasing mask use at 80% with the scenario     where 80% and 12.8% of the population are vaccinated with two and three doses respectively, we found that high mask use results in a lower cumulative total number of cases (about 28 million) compared with high two doses vaccination coverage (about 31 million), but a higher number of total death (about 38,000 against 23,000).When testing if results were consistent in the case of only poor-quality masks (40% instead of 60% effectiveness) being available, we found that increasing mask use from 50% to 80% could have reduced cases and deaths by 6.1% (from 35,246,400 at 50% to 33,070,800 at 80% mask use) and 11.3% respectively (from 65,758 to 58,335).If vaccine coverage was increased to 60% with two doses and 9.6% of people aged 60 + with three doses, the reduction in cases and deaths could have been respectively 2.3% (from 35,246,400 to 34,443,600) and 25.6% (from 65,758 to 48,781).

Results in context
This is the first study to quantify the effect of war on infectious disease transmission, and can inform public health response and COVID-19 outbreak control programs in Ukraine and neighboring countries accepting Ukrainian refugees.
With low vaccination rates and poor prospects of resuming vaccination campaigns during the war, the COVID-19 epidemic curve in Ukraine was expected to peak in April 2022 based on our estimates, with almost half of all available hospital beds required on the day of the peak.The availability of beds may be much less than the pre-war estimates due to the destruction of hospitals and the lack of healthcare staff who have either fled, are sheltering in place, or have suffered injury or illness.The case fatality rate from COVID-19 will rise if people who require hospital care, even the simplest of care such as oxygen therapy, are unable to receive it.In fact, our results suggest that even pre-war, Ukraine was seeing a higher than expected death rate from COVID-19, and this may reflect low health system capacity.Studies from Europe showed that when hospitals are overwhelmed and ICU occupancy is high, case fatality rates rise 31 .
We showed that increasing vaccination rates to 60% is less effective than increasing mask use to 80% in reducing transmissions, largely due to the low efficacy of two doses against transmission.However, if vaccination coverage is increased to 80% or mask effectiveness is reduced because of low-quality masks being available, increasing vaccination coverage is more effective in reducing the number of deaths.

Study limitations
There are several limitations in this study.One is the reliability of data.We showed a good fit of our model to the case notifications data, but the deaths in Ukraine are much higher than estimated in the rest of Europe 22 and the US 23 for the Omicron variant.The modelled death rates were lower than the observed data, and a good fit is seen only when multiplying those rates by 5 times, suggesting a high proportion of unnotified COVID-19 cases, or higher death rates due to a combination of low vaccination rates and lack of hospital care for the critically ill in Ukraine compared to the rates used estimated for EU and the US.
We assumed the mask effectiveness to be 60% 27 , however, the effectiveness varies depending on the type of masks used, therefore we show results for poor-quality masks.One of the strengths of this study is the use of age-specific parameters and contacts.However, another limitation, highlighted already from a previous similar modelling study of COVID-19 in Ukraine using the same contacts matrix 32 , is that we did not account for variation between different Ukrainian regions in terms of their population structure, local transportation modes, patterns of work and other factors which affect disease spread in those regions 32 .Furthermore, there is no data on the impact of war on contact patterns, hospitalization, and death rates.The hospitalization and death rates previously estimated for the Delta variant in Ukraine 32 are much higher than the ones used here, adapted for the Omicron variant.However, health system disruption, which could result in a large increase in death rates, is not considered in this study and this could show an underestimation of the disease burden.The relative difference between Omicron and Alpha or D614G in an unvaccinated population is less clear.Studies suggest Delta is twice as severe as Alpha, so Omicron may be similar to Alpha in the unvaccinated 33 .Furthermore, we assumed vaccine effectiveness against Omicron infection to be the one estimated for Astra Zeneca and Pfizer.However, 30% to 40% of the Ukraine population received a different vaccine for the first two doses which may had lower effectiveness.The impact of vaccination, therefore, may be less than the model suggests.Following three doses, we used a vaccine effectiveness of over 70%, however, it has been shown that after two months from vaccination, this effectiveness may wanes to 47%, therefore we could have overestimated the benefit of having increased vaccine coverage on the case incidence 7 .Vaccine effectiveness against hospitalization and ICU admission is uncertain, as some studies show consistent effectiveness in time 7 and others show rapid waning 6 .We did not consider reinfection, as the model only runs for 150 days, and data at that time on re-infection was limited.Many other, more immune-evasive variants of Omicron have emerged in the two years since the war began, which may further reduce the impact of vaccination.

Comparison with a country not affected by war with a good working health system
Starting from a low baseline of 34.5% and 1.7% of the total population with two and three doses of coverage of COVID-19 vaccination, Ukraine was highly vulnerable to COVID-19 at the time of the Russian invasion of the country.We estimated that in the first half of 2022, in the best-case scenario with 80% of the population wearing masks, about 69% of the Ukrainian population (27.5 million) could have been infected with COVID-19 with almost 70,000 beds needed per day at the peak, and a total of about 28,000 cumulative deaths.Contrasting the situation in Ukraine with a country where there is peace and high vaccination coverage is informative.Australia is such a country, which rapidly achieved over 95% two-dose coverage and was using masks and other mitigations such as antivirals early 2022.In Australia it is estimated that 42% of the population (10.9 million) have been infected 34 with less than 9000 deaths registered in 2022 35,36 , with hospital bed requirements of 1300 per day at the peak in January, and not exceeding 700 per day after January 2022 37 .The population of Ukraine is just under double that of Australia.The estimated death rate for 2022 in Australia was about 0.08%, while in Ukraine, with lower vaccination rates, is estimated to be 0.16%.Hospitalization beds required at the peak in Ukraine were orders of magnitude higher, with 0.24% of the population being hospitalized at the peak compared with 0.005% in Australia.

Conclusions and recommendations
In the context of conflict, COVID-19 could result in a much higher number of deaths, due to poor public health infrastructure, disruption of health systems, and interruption to vaccination programs.Reduced access to hospital or ICU beds, likely resulted in higher case fatality [38][39][40][41] .Our study shows that masks can reduce the impact of COVID-19 in Ukraine.If the cold chain cannot be maintained, health systems are unable to function and vaccines cannot be distributed, masks may be a more feasible and effective way to mitigate the epidemic during war.Indeed, masks can be distributed without functioning health infrastructure.
Vaccination campaigns and the provision of high-quality masks, together with the provision of health clinics and workforce, should be part of any humanitarian response to the situation in Ukraine [38][39][40][41] .Furthermore, since the start of the war on 24 February 2022, several millions of people have fled the country, and vaccination of refugees should be a high priority and may be more feasible than delivering vaccine programs in-country 42 .

Figure 1 .
Figure 1.Model fit to clinical case notification data (COVID-19 incidence notification data) between 6 January 2022 and 25 February 2022, with masks 60% effective.The vertical lines represent the dates of policy changes.6January is the start of the Omicron wave; 6-17 January is the first pre-war period with the implemented restrictions policy17,18 ; 18-25 January is the second pre-war period where the reduction in contacts increases with the additional policy of closure of public transports and enforced mask use in all public spaces17,18 ; 25 February represent the start of the war, where vaccine rollout and tracing stops.

Figure 3 .
Figure3.Number of daily new cases (incidence), cumulative cases, and cumulative deaths in the scenario with 39.3% of the 15 + age group with two doses and 6.3% of the 60 + with 3 doses, varying mask use coverage (0%, 50%, 80%) with 60% effectiveness, from 6 January to 14 April 2022.

Figure 4 .
Figure 4. Hospitalization (H) and ICU beds used over time, keeping the last vaccination coverage notified (39.3% of the 15 + age group with two doses and 6.3% of the 60 + with 3 doses) and varying mask use coverage (0%, 50%, 80%), from 6 January to 5 June 2022.

Figure 5 .
Figure 5. Case incidence, cumulative cases, and deaths in the scenario with 50% of the population using masks and varying the vaccination coverage, for 2 doses (v2) and three doses (v3), from 6 January to 16 April 2022.

Figure 6 .
Figure 6.Hospitalization (H) and ICU beds used over time, with 50% of the population using masks and varying the vaccination coverage for two (v2) and three doses (v3), from 6 January to 5 June 2022.