Abstract
As COVID-19 vaccines become available, different model-based approach have been developed to evaluate strategic priorities for vaccine allocation to reduce severe illness. One strategy is to directly prioritize groups that are likely to experience medical complications due to COVID-19, such as older adults. A second strategy is to limit community spread by reducing importations, for example by vaccinating members of the mobile labour force, such as rotational workers. This second strategy may be appropriate for regions with low disease prevalence, where importations are a substantial fraction of all cases and reducing the importation rate reduces the risk of community outbreaks, which can provide significant indirect protection for vulnerable individuals. Current studies have focused on comparing vaccination strategies in the absence of importations, and have not considered allocating vaccines to reduce the importation rate. Here, we provide an analytical criteria to compare the reduction in the risk of hospitalization and intensive care unit (ICU) admission over four months when either older adults or rotational workers are prioritized for vaccination. Vaccinating rotational workers (assumed to be 6,000 individuals and about 1% of the Newfoundland and Labrador (NL) population) could reduce the average risk of hospitalization and ICU admission by 42%, if no community spread is observed at the time of vaccination, because epidemic spread is reduced and vulnerable individuals are indirectly protected. In contrast, vaccinating all individuals aged 75 and older (about 43,300 individuals, or 8% of the NL population) would lead to a 24% reduction in the average risk of hospitalization, and to a 45% reduction in the average risk of ICU admission, because a large number of individuals at high risk from COVID-19 are now vaccinated. Therefore, reducing the risk of hospitalization and ICU admission of the susceptible population by reducing case importations would require a significantly lower number of vaccines. Benefits of vaccinating rotational workers decrease with increasing infection prevalence in the community. Prioritizing members of the mobile labour force should be considered as an efficient strategy to indirectly protect vulnerable groups from COVID-19 exposure in regions with low disease prevalence.
Introduction
Vaccines for COVID-19 have become available (Corum et al., 2020), and authorities are facing decisions as to which groups to prioritize to mitigate severe illness due to COVID-19. Model-based approaches have been developed to compare the efficiency of different vaccination strategies in reducing hospitalization, intense care unit (ICU) admission, death, or long COVID (Bubar et al., 2021; Mulberry et al., 2021; Chen et al., 2021). A common strategy has been to prioritize vulnerable individuals, defined as people at high risk from COVID-19 (NHS), such as older adults (Bubar et al., 2021), or people with health conditions (e.g., heart or lung conditions, weakened immune systems, obesity, or diabetes) (Persad et al., 2021). Other effective vaccination strategies have included prioritizing essential workers (Mulberry et al., 2021), or individuals with a large number of social contacts (Chen et al., 2021).
An alternative approach is to vaccinate groups that are likely to introduce the disease into the community, such as rotational workers or other members of the mobile labour force (Lemke, 2021). These are workers that, to perform essential functions, are required to cross provincial or international borders on a regular basis (Neis et al., 2020). Examples include fisheries workers, truck drivers, flight crews, or other individuals that alternate times away at work with time at home. These workers may become infected with COVID-19 while working in another province, and when returning home may initiate a community outbreak. No current studies have considered how reducing the importation rate by vaccinating rotational workers could lower the risk of community infections and severe illness due to COVID-19 by preventing spread to all individuals, including those in vulnerable groups.
The population of Newfoundland and Labrador (NL) is ∼522,000 and more than 90% of the population lives on the island of Newfoundland (Statistica, 2021). The island has very few ports of entries and during the pandemic has been subject to stringent border control (Hurford et al., 2021). In part, due to these strict measures, NL has experienced very minimal community spread for extended periods of time (Berry et al., 2020). In NL rotational workers are a significant part of the work force (Hewitt et al., 2018), where a rotational worker is defined as ‘a resident of Newfoundland and Labrador who travels to another province or territory of Canada to work, on a set schedule of time away at work alternating with time at home in Newfoundland and Labrador’ (GovNL, 2021). The large majority of rotational workers from NL are intraprovincial workers employed in Ontario, Alberta, and Nova Scotia (Hewitt et al., 2018), and since October 2020 infection prevalence in Ontario and Alberta has been substantially higher than in NL (Berry et al., 2020). Vaccinating rotational workers could reduce the rate at which the virus is imported in a region, reducing the chance of a community outbreak and providing indirect protection for those individuals that are likely to experience medical complications due to COVID-19.
Here, we quantify the reduction in the risk of hospitalization and ICU admission if: (1) vulnerable groups in the community are vaccinated, thus reducing the average probabilities of hospitalization and ICU admission across the remaining susceptible population, where the average is calculated by considering the age structure and the probabilities of hospitalization and ICU admission for each age group in the susceptible population (see Appendix A); or (2) rotational workers are vaccinated, thus reducing the rate that the virus is imported in the community, and, in turn, reducing the risk of community outbreaks, but leaving probabilities of hospitalization and ICU admission of the susceptible population unchanged. We derive a concise analytical criteria to compare the reduction in risk occurring when strategies (1) and (2) are applied. The criteria is based on easily accessible quantities, such as infection prevalence in the community at the time of vaccination, the number of importations to the province due to non-self-isolating rotational workers, the percent reduction in viral transmission after vaccination, and the probabilities of hospitalization and ICU admission of the vulnerable group vaccinated (Fig. 1). We calculate the risk of hospitalization and ICU admission for NL when rotational workers or older adults are vaccinated. Our analysis provides additional support for public health decisions regarding vaccine prioritization strategies.
Model and Methods
Modelling the impact of vaccination on the infection dynamics
We use a SIR model (Kermack and McKendrick, 1927) with importations to simulate changes in the infectious status of a population as a consequence of community and disease introduction from external sources. The disease can spread from infected individuals I to susceptible individuals S, with a transmission rate β per day. The rate that infected individuals enter the province and fail to self-isolate is m individuals per day, and all infected individuals are assumed to be immediately infectious. Infected individuals recover from infection (or die) at rate γ per day, and are subsequently removed from the description of the infection dynamics. We assume that the rate that individuals enter the province remains low over time, and the total population size can be approximated by the constant N. The system of differential equation representing the dynamics described above is: with initial conditions {S0, I0, R0}. We assume that at time zero a certain number of individuals in the community(V0 ≥ 0) are already vaccinated. Additionally, we assume that at this time a certain number of vaccines are available to vaccinate either vulnerable individuals (Vg), or rotational workers Vw. Therefore we will consider the situations where the same number V of either vulnerable individuals or rotational workers are vaccinated (i.e., either Vg = V and Vw = 0, or Vw = V and Vg = 0), as well as the situation where neither rotational workers nor vulnerable individuals are vaccinated (Vg = Vw = 0). We assume that vaccination reduces the probability that susceptible individuals become infected and transmit the disease by (1 − Z), where Z = 0 indicates that vaccination completely prevents viral transmission, while Z = 1 indicates that vaccination does not prevent viral transmission. We derive the following initial condition for the number of susceptible individuals in the community at time zero: where I0 represents infection prevalence at time zero, and is understood as the number of cases that are not in self-isolation and are spreading infections undetected in the community at the time vaccination begins.
Unvaccinated rotational workers are part of the susceptible class, but can also introduce the disease into the community through importations. The importation rate (m infections per day) is determined by the total number of rotational workers W, by the number of rotational workers vaccinated (Vw), by the proportion of importations due to non-self isolating rotational workers (η), by the importation rate due to other non-self-isolating travelers (ω individuals per day), and by the reduction in viral transmission after vaccination (Z). As such: and vaccinating rotational workers reduces the importation rate m. We define m0 as the importation rate before rotational workers are vaccinated, and mr as the reduced importation rate when rotational workers have been vaccinated. Vaccinating vulnerable groups does not affect the importation rate. Model parameters and default values used for the simulations are summarized in Table 1.
Quantification of risk reduction due to vaccination
To quantify the reduction in the risk of hospitalization and ICU admission of a community when rotational workers (RW) or vulnerable groups (VG) are vaccinated, we define:
An equivalent equation can be written for the % reduction in the risk of ICU admission.
To calculate the expected number of hospitalizations and ICU admission we compute the number of cumulative infections C(T) over a certain period of time T by using Eq. (1). We multiply then C(T) by the average probabilities of hospitalization (Ph) or ICU admission (Pu) of the susceptible population and obtain:
Vaccinating RW reduces the expected number of hospitalizations by reducing the importation rate (see Eq. (3)), and thus the cumulative number of infectious cases. Vaccinating vulnerable groups reduces the average probabilities of hospitalization (from Ph to ) and ICU admission (from Pu to ) for the susceptible population because vaccinated vulnerable individuals are no longer included in the calculation of this average. (see Table 1 and Appendix A).
Application to Newfoundland and Labrador
We compare the reduction in the risk of hospitalization and ICU admission in NL when vaccinating either rotational workers or vulnerable individuals. We estimate that NL has about 6,000 rotational workers. The census data (Population and Demographics) indicates that in NL about 12,000 individuals are between 80 and 84 years old. We consider half of the older adults in the 80-84 age group to be the vulnerable individuals of interest for our analysis, as this group offers a good size comparison with RW. Additionally, we consider that individuals aged 85 or older have already received the vaccine. The same analysis could be performed for a different vulnerable group, whose vaccination leads to a decrease in the average probabilities of hospitalization and ICU admission of the susceptible population.
Using the NL census data (Population and Demographics), the age-stratified probabilities of hospitalization and ICU admission (Ferguson et al., 2020; Kronbichler et al., 2020), and the distribution of asymptomatic cases by age groups (Kronbichler et al., 2020), we calculate the average probabilities of hospitalization and ICU admission in NL, where the average is taken across the age-structure in the susceptible population. The exact data used are given in Appendix A. Our estimates are Ph = 7.89% for the average probability of hospitalization in NL due to COVID-19 infection, and Pu = 2.35% for the average probability of ICU admission (if individuals aged 85 and above are vaccinated). Assuming that vaccination prevents severe illness (Voysey et al., 2021; Knoll and Wonodi, 2021), vaccinating 6,000 older adults aged 80-84 would lower the average probability of hospitalization of the remaining susceptible population to , and the average probability of ICU admission to . We estimate that NL may experience 0.4 to 2 importations of cases of non-self-isolating individuals every month, and we assume rotational workers are responsible for 20% to 80% of those importations.
Results
Quantification of risk reduction due to vaccination
The risk of hospitalization when vulnerable groups are vaccinated can be expressed as where Ph and are the average probabilities of hospitalization before and after vaccination and C(T, β, m) is the number of cumulative cases, which depends on the time interval T considered, on the transmission rate before or after vaccination (βj = βS0/N, for j = 0, v), and on the importation rate m0. Similarly, the reduction in risk of ICU admission when VG are vaccinated is given by where Pu and are the average probabilities of ICU admission before and after vaccination.
The reduction in the risk of hospitalization and ICU admission when RW are vaccinated is: where m0 and mr represent the importation rates before and after vaccination, and depend on the number of RW (W), on the number of RW vaccinated (Vw), on viral transmission after vaccination (Z), on the proportion of successful importations due to non-self-isolating rotational workers (η), and on the importation rate due to other non-self-isolating travelers (ω) (see Eq. (3)). Note that the reduction in the risk of hospitalization when RW are vaccinated is equal to the reduction in the risk of ICU admission, as the probabilities Ph and Pu remain unchanged before and after vaccination of RW, and can be canceled from the numerator and the denominator of the fraction of Eq. (6).
We can determine when a reduction in the risk of hospitalization due to vaccinating VG equals a reduction in the risk of hospitalization obtained by vaccinating RW as follows:
Eq. (7) can be simplified as:
The risk of hospitalization can therefore be equally reduced by a reduction in the average probability of hospitalization, or by a reduction in the expected number of cumulative infections.
When infection prevalence remains low, and changes in the susceptible population over a short period of time remain small, the system of Eq. (1) can be approximated by a single linear differential equation in I, namely: with solution where βj = βS0/N, for j = 0, v. The cumulative number of infections can be computed by solving the integral: with analytical solution
Note that if infection prevalence is initially zero (i.e., I0 = 0), the number of cumulative cases is linearly proportional to the importation rate m (cfr. Eq. (12); to see this note that C(m, βj, T) = mC(βj, T) for I0 = 0), and Eq. (8) can be rewritten as
Thus, for I0 = 0 (i.e., in the absence of community spread at vaccination time), an X% reduction in the importation rate has the same impact on reducing the risk of hospitalization as an X% reduction in the average probability of hospitalization. Note that Eq. (13) does not depend on the transmission rate β in the community or on the time interval T considered.
In contrast, for I0 > 0, an X% reduction in the probability of hospitalization has a much larger impact than the same X% reduction in the importation rate. For example, if I0 = 20, a 70% reduction in the importation rate corresponds to a 10% reduction in the probability of hospitalization. A graphical representation of Eq. (8) as a function of infection prevalence at the initial time is given in Fig. 2. The importance of vaccinating rotational workers when I0 > 0 decreases more rapidly when the importation rate is low, when viral transmission is high and when shorter time intervals are considered (see Appendix B, Figs. B.1 and B.2). An analogous relationship to Eq. (8) can be written to compare a reduction in the probability of ICU admission and a reduction in the importation rate:
Application to Newfoundland and Labrador
Fig. 3 shows the reduction in the risk of hospitalization obtained when 6,000 rotational workers (corresponding to 1% of the total population) or 6,000 individuals between 80 and 84 years (i.e., half of the 80-84 age group in NL) are vaccinated. We can see that if infection prevalence is initially zero and if rotational workers are responsible for 60% of the importations, vaccinating rotational workers decreases the risk of hospitalization and ICU admission by 42%. Alternatively, vaccinating older adults would lead to a 7% reduction in the risk of hospitalization and a 12% reduction in the risk of ICU admission for the susceptible population. Even if non-self-isolating rotational workers are responsible for only 20% of importations, vaccination would lead to a 16% reduction in the risk of hospitalization and ICU admission, which is still higher than the reduction achieved by vaccinating half of the 80-84 age group.
Using Eqs. (3) and (13) we can show that, if infection prevalence at the time of vaccination is zero, vaccinating rotational workers may ensure a large reduction in the risk of hospitalization and ICU admission with a lower number of vaccines. We estimate that if vaccination reduces viral transmission by 67% (Wise, 2021), and if rotational workers are responsible of 60% of the importations (with 1.2 importations per month due to non-self-isolating rotational workers and 0.8 non-self-isolated importations per month due to other sources, i.e., Wη = 1.2/30 and ω = 0.8/30), vaccinating 6,000 rotational workers would lead to a 42% reduction in the importation rate, and thus to an equivalent reduction in the risk of hospitalization and ICU admission (see Eq. (13) and Fig. 3). In contrast, vaccinating all individuals aged 75 and older in NL (about 43,300 individuals, or 8% of the total population) would lead to only a 24% reduction in the risk of hospitalization, and a 45% reduction in the risk of ICU admission.
The benefits of vaccinating rotational workers decrease rapidly with increasing number of infectious cases in the community. With 50 infectious cases in the community (not in self-isolation), vaccinating half of the 80-84 age group has a nearly equivalent impact on reducing the risk of hospitalization as vaccinating rotational workers, even if rotational workers are responsible for 80% of the imported cases. With only 20 infectious cases, the reduction in the risk of ICU admission obtained when vaccinating older adults is nearly equivalent to what obtained by vaccinating rotational workers.
Discussion
Most studies on vaccine priorities found that an efficient strategy to reduce severe illness due to COVID-19 is to first vaccinate groups that are likely to experience severe illness (Bubar et al., 2021; Babus et al., 2020; Jentsch et al., 2020; Buckner et al., 2020). However, these studies did not consider the impact of disease importations on the infection dynamics. Importations can constitute a significant infection source in regions such as the Atlantic Provinces and the Territories, where most COVID-19 cases are travel-related. We consider both community spread and virus importations by travelers to evaluate the impact of vaccinating rotational workers on reducing the risk of hospitalization and ICU admission. We show that, when disease prevalence at the time of vaccination is low, community spread can be significantly reduced by vaccinating individuals that are likely to introduce the disease in the community, thereby indirectly protecting vulnerable individuals from infection.
We compare the expected number of hospitalizations and ICU admission over four months in the absence of vaccination to this same quantity if rotational workers or vulnerable groups had been vaccinated. We found that, when infection prevalence at the time of vaccination is zero, an X% decrease in the importation rate and an X% decrease in the average probability of hospitalization (or ICU admission) reduce the risk of hospitalization (and ICU admission) in the community by equal amounts. However, if the number of rotational workers is low, reducing disease importations would require significantly less vaccines than reducing the average probabilities of hospitalization and ICU admission. We show that the reduction in the risk of hospitalization and ICU admission when vaccinating 6,000 rotational workers (or 1% of the NL population) can be 3-7 times larger than that obtained if the same number of vaccines were directly given to individuals aged 80 years and older instead. Indeed, vaccinating a small number of older adults would not be sufficient to significantly lower the average hospitalization and ICU admission probabilities amongst the remaining susceptible population when the number of elderly individuals vaccinated is too few. This strategy, however, relies on being able to correctly identify rotational workers, so as to not vaccinate individuals who falsely identify themselves as such.
Vaccinating rotational workers becomes less desirable when infection prevalence in the community at the time of vaccination is high. Our findings are in agreement with the analysis of Russell et al. (2021), showing that imported cases greatly contribute to local epidemic outbreaks in countries with low COVID-19 prevalence, but not in regions with high prevalence. Our analysis shows that the prioritization of rotational workers for vaccination will reduce the importation rate and is likely to prevent community outbreaks that might otherwise have occurred, but we note that it will not be possible to observe these averted community outbreaks. Therefore, while the prioritization of rotational workers should be considered for communities with low infection prevalence to protect their vulnerable groups, the effect of this vaccination strategy can only be assessed with scenario modelling.
Our work provides an analytical criteria that can be easily adopted by public health officials to evaluate whether protection of vulnerable individuals can be achieved by prioritizing members of the mobile labour force for vaccination. Our criteria relates basic elements, such as infection prevalence at the time of vaccination, importation rate, viral transmission after vaccination, and population age structure, to the risk of hospitalization and ICU admission before and after vaccination. Our simple model disentangles the basic relationships between the importation rate and the consequent community infection dynamics, to capture features relevant for planning vaccines prioritization strategies. The analysis presented here can be easily applied to compare vaccinating rotational workers and any vulnerable group known to be at high risk from COVID-19, consisting for example of individuals with health conditions or individuals of other age groups, provided that their vaccination leads to a reduction in the average probabilities of hospitalization and ICU admission of the susceptible population.
For more accurate predictions, future research should consider rotational worker compliance with quarantine, self-isolation and testing requirements (Arino et al., 2020), disease prevalence at work location (affecting the importation rate) (Lopez et al., 2016), the risk of importing new variants of concern (Du et al., 2021), waning immunity (Bubar et al., 2021), non-pharmaceutical interventions, population heterogeneity, and, more generally, stochasticity. Variants of concern are particularly relevant to consider, as immediately after their identification, the prevalence of the particular variant will be low in all other regions. Therefore, our results may apply to all regions of the world with a substantial mobile work force when variants of concern with significant novel effects emerge. Additionally, the dichotomy we consider is to illustrate the value of vaccinating rotational workers, which has been under-explored, however, in practice a multi-pronged approach where vulnerable groups and rotational workers are vaccinated simultaneously is sensible, particularly as more vaccines arrive. Finally, the focus of our work is epidemiological, although economic implications, moral and ethical considerations, and consequences for mental health should also be considered (Giubilini, 2019; Premji, 2020).
Data Availability
The manuscript contains a mathematical analysis and all data values are provided in the text.
Acknowledgement
We acknowledge Randy Giffen for suggesting that we conduct this study. AH is supported by grants from the National Science and Engineering Research Council (Discovery Grant), the Atlantic Association for Research in the Mathematical Sciences Centre for Disease Modelling, and the Newfoundland and Labrador Centre for Health Information. AH and MM acknowledge financial support from CANMOD (Canadian Network for Modelling Infectious Diseases - Réseau canadien de modélisation des maladies infectieuses).
Appendix A: Supplementary Data
Population
Age groups considered:
Population of Newfoundland and Labrador by age group (Population and Demographics):
Proportion of the Newfoundland and Labrador population in each age group (where the total population size N = 522, 103 (Population and Demographics)):
Proportion of the susceptible Newfoundland and Labrador population in each age group (if individuals aged 85 and above, 10,256 people (Population and Demographics), are vaccinated. We assume that vaccination reduces viral transmission by 0.67% (Knoll and Wonodi, 2021)):
Proportion of the unvaccinated Newfoundland and Labrador population in each age group (if half of the 80-84 age group, 6,309 people, and individuals aged 85 and above, 10,256 people, are vaccinated. We assume that vaccination reduces viral transmission by 67% (Knoll and Wonodi, 2021)):
Probabilities of hospitalization and ICU admission
Age-stratified probabilities of hospitalization given symptomatic infection (Ferguson et al., 2020):
Age-stratified probabilities of asymptomatic infection (adapted from Kronbichler et al. (2020)):
Age-stratified probabilities of hospitalization (assuming that asymptomatic infections do not result in hospitalization.) (ph0 = ps(1 − pa)):
Age-stratified probabilities of ICU admission given hospitalization (Ferguson et al., 2020):
Age-stratified probabilities of ICU admission (pu0 = ph0 puh0):
Probabilities of hospitalization and ICU admission after vaccination of older individuals
To calculate those probabilities we assume that vaccination prevents severe illness (Voysey et al., 2021; Knoll and Wonodi, 2021), and that the probability of hospitalization and ICU admission of vaccinated individuals is 0. In NL, there is 10,256 individuals aged 85 and above, and 12,617 are between 80 and 84 years old (Population and Demographics). Vaccinating half of the 80-84 age group corresponds therefore to vaccinating 6,309 individuals.
Age-stratified probabilities of hospitalization given infection if individuals aged 85 and above are vaccinated:
Average probability of hospitalization if individuals aged 85 and above are vaccinated:
Age-stratified probabilities of ICU admission if individuals aged 85 and above are vaccinated:
Average probability of hospitalization if individuals aged 85 and above are vaccinated:
Age-stratified probabilities of hospitalization if half of the 80-84 age group and individuals aged 85 and above are vaccinated:
Average probability of hospitalization when half of the 80-84 age group and individuals aged 85 and above are vaccinated:
Age-stratified probabilities of ICU admission if half of the 80-84 age group and individuals aged 85 and above are vaccinated:
Average probability of ICU admission if half of the 80-84 age group and individuals aged 85 and above are vaccinated:
Most of the rotational workers of Newfoundland and Labrador are between the ages of 25 and 39 (Hewitt et al., 2018). As the number of individuals in those age classes is high, and as the probabilities of hospitalization and ICU admission for those age classes are low, we assume that vaccinating rotational workers does not affect the average probabilities of hospitalization and ICU admission of the susceptible population.