Vaccination is the best tool to control the spread of SARS-CoV-2 and minimize the burden of COVID-19 globally. Because Ghana primarily relies on multilateral donations for their COVID-19 vaccine supply, there is a need to determine the best vaccine optimization strategies to minimize deaths, cumulative case counts, and epidemic peaks over a relatively short period. Using 2 contact matrices, we used an age-stratified mathematic model to answer the question of who should get vaccinated first when the vaccine supply is limited and when supplies are exhausted over 3 and 6 months. Our findings suggest that, for both the initial strain and the Delta variant, prioritizing persons <25 years of age for vaccination would avert the most cumulative infections and prioritizing the elderly (≥65 years of age) would result in the lowest death counts.

Optimization of vaccine prioritization strategy is sensitive to the population structure. Prioritizing younger persons to avert cumulative infections is a finding that has been reported in other studies (13,38,39). Bubar et al. concluded in their multicountry research that the cumulative incidence of COVID-19 was lowest when adults 20–49 years of age were prioritized, especially with a highly effective transmission-blocking vaccine (13). In Senegal, Diarra et al. used an age-structured dynamic mathematical model to explore various vaccination strategies and reported that prioritizing persons <60 years of age was associated with the lowest case burden (40). Those authors argued that countries with younger populations, such as Ghana, should prioritize vaccinating younger persons to minimize hospital costs and productivity loss.

As was the case for our team, most research teams conducting previous studies concluded that prioritizing the elderly was associated with the lowest mortality. However, Bubar et al. reported that persons 20–49 years of age should be prioritized to minimize mortality when transmission is low, when vaccine efficacy is lower in older adults, and when the vaccine is highly effective in blocking transmission. Buckner et al. reported results similar to those in our study and found that, to control deaths directly, the elderly should be vaccinated first, after stratifying young adults by essential worker status (41). Although the conclusions in that study and our study were similar, Buckner et al. used a dynamic approach in modeling vaccine allocation strategies that accounted for changes in the epidemiologic status of the population (shares of the population in different disease states) over 6 months using stochastic nonlinear programming techniques. In a vaccine optimization modeling study in India, Foy et al. concluded that prioritizing older adults (>60 years of age) was associated with the most significant reduction in deaths, regardless of vaccine efficacy, control measures, and rollout speed (38). Another modeling study by Chapman et al., using COVID-19 data from California, reported similar results (42); however, that study focused on identifying the groups to prioritize after healthcare workers and long-term facility residents received initial vaccine doses.

The differences in outcomes between the contact matrices in our study might be due to the much lower reported contact rates among the younger population in the matrix adapted from Trentini et al. (i.e., second matrix) (35). The study by Bubar et al. on vaccine optimization strategies across multiple countries, including South Africa, concluded the best vaccination strategy depended on the extent of mixing patterns (13). The use of 2 contact matrices reflects the degree to which decision makers should consider social interactions in the population before optimizing vaccination strategies when vaccine supplies are limited. Our findings demonstrate that the mixing pattern is relevant when the goal of the vaccination program is to minimize infection burden and the vaccine rollout takes place over an extended period. Thus, a population with lower contact rates among the older population would need to prioritize younger persons. However, contact patterns in the population may not be relevant if the goal of the vaccination program is to minimize deaths and vaccine uptake is high. Future studies might consider exploring differences observed using matrices of different settings; for example, rural versus urban and household versus community mixing.

As reported by Ko et al., the question of who should receive vaccinations first depends also on the objective for vaccination (minimizing cumulative infections or deaths) and the effective reproduction number (39). Thus, policymakers might need to consider compromises in deciding the best vaccine allocation strategies. For example, prioritizing the elderly may lead to fewer deaths but higher case numbers, which could exacerbate economic loss due to a high case count in a younger population. The transmissibility of the circulating variant also might inform a vaccine optimization strategy. We did not see evidence of this effect in our study because the priority group remained the same for the Delta variant, which carried a higher reproduction number. Another study concluded vaccine optimization depended on the vaccine supply (42).

Although our study demonstrates the need to prioritize certain age groups to minimize the burden of COVID-19 in Ghana, depending on the objective of the program, other factors need to be considered to ensure people receive vaccinations when they become eligible. Employing targeted vaccine campaigns to minimize hesitancy among the prioritized group might be a necessary part of the program. Acheampong et al. reported the level of reluctance among older adults was lower than that for younger adults in Ghana (43). Likewise, a survey among persons >65 years of age in the United States found that 91% of the elderly were willing to get vaccinated (44). This reported vaccine hesitancy across different age groups suggests a need for campaigns to create an enabling environment and engage younger populations about their role in mitigating the pandemic.

The first limitation of our study is that our model was age-stratified only. Other demographic variables (e.g., occupation and comorbidity) might change a person’s COVID-19 infection risk and clinical prognosis (45). Second, our model did not include a hospitalization compartment. Thus, we could not evaluate the effects of the Omicron variant, for which vaccines demonstrated effectiveness against severe infections but modest effectiveness against infection. Third, the contact matrices we used were adapted from other countries in Africa. Those countries had similar demographic distributions as Ghana, and we assumed that the frequency of contact in the population would be similar. Fourth, we did not have data to represent the rural and urban differences in contact matrices in Ghana. Fifth, our model design specified the symptomatic period to be the same for persons who recovered from COVID-19 and those who died from it. Sixth, the highly transmissible Omicron variant was not included in our study because of limited evidence of the effectiveness of vaccines against infection from that variant (46). Last, our model accounted for vaccine effectiveness against infection but did not account for the reduction of the case-fatality ratio if a person was vaccinated and still became infected.

In conclusion, we used an age-stratified compartmental model to assess the impact of various COVID-19 vaccine allocation strategies in Ghana. Our study reiterates the need to increase vaccination rates by ensuring increased vaccine supplies and faster rollout speed. Vaccinating persons <25 years of age was associated with the highest numbers of cumulative infections averted for the initial strain and the Delta variant. Prioritizing persons ≥65 years of age was associated with the lowest deaths in the population. Our findings indicate that vaccine prioritization strategies in Ghana, or in any country, depend on the country’s policy objectives, population structure, mixing patterns, and vaccine supply.