Comparing COVID-19 vaccine allocation strategies in India: A mathematical modelling study

Abstract

Background: The development and widespread use of an effective SARS-CoV-2 vaccine could prevent substantial morbidity and mortality associated with COVID-19 and mitigate the secondary effects associated with non-pharmaceutical interventions.

Methods: We used an age-structured, expanded SEIR model with social contact matrices to assess age-specific vaccine allocation strategies in India. We used state-specific age structures and disease transmission coefficients estimated from confirmed incident cases of COVID-19 between 1 July and 31 August 2020. Simulations were used to investigate the relative reduction in mortality and morbidity of vaccine allocation strategies based on prioritizing different age groups, and the interactions of these strategies with concurrent non-pharmaceutical interventions. Given the uncertainty associated with COVID-19 vaccine development, we varied vaccine characteristics in the modelling simulations.

Results: Prioritizing COVID-19 vaccine allocation for older populations (i.e., >60 years) led to the greatest relative reduction in deaths, regardless of vaccine efficacy, control measures, rollout speed, or immunity dynamics. Preferential vaccination of this group often produced relatively higher total symptomatic infections and more pronounced estimates of peak incidence than other assessed strategies. Vaccine efficacy, immunity type, target coverage, and rollout speed significantly influenced overall strategy effectiveness, with the time taken to reach target coverage significantly affecting the relative mortality benefit comparative to no vaccination.

Conclusions: Our findings support global recommendations to prioritize COVID-19 vaccine allocation for older age groups. Relative differences between allocation strategies were reduced as the speed of vaccine rollout was increased. Optimal vaccine allocation strategies will depend on vaccine characteristics, strength of concurrent non-pharmaceutical interventions, and region-specific goals.

Keywords: COVID-19; Immunization; Mathematical modelling; SEIR.

Figure 1

Schematic of model transmission dynamics. Subjects may move from susceptible to exposed to symptomatic or asymptomatic infectious. Asymptomatic infectious are assumed to always recover, while symptomatic infectious first quarantine, before recovering or dying and (if recovered) eventually losing immunity. Each major compartment comprises eight sub-compartments, comprising age groups (0–10, 10–20, […] 60–70, ≥70 years). Rates of symptomatic infection ($p$) and death ($\delta$) vary by age group. Contact between susceptible and infectious populations is age-structured, proportional to the estimated contact pattern matrix ($C$). Under a progressive rollout scheme, $M$ individuals are vaccinated each week, at an efficacy of $\epsilon$. If the vaccine confers non-sterilizing immunity, individuals can become exposed and develop asymptomatic infections, before recovering and eventually losing infection-driven immunity. For those vaccines that do not confer sterilizing immunity, ${\beta }_2=0$; meaning vaccinated individuals no longer contribute to transmission dynamics.

Figure 2

Simulated infection curves and cumulative deaths with four vaccination strategies. Each simulation assumed that 3% of the population was vaccinated each month, with a vaccine efficacy of 75%, no control measures, and an $R_0$ of 2.4. Strategies 1–4 corresponds to no prioritization of any age group in vaccination, while strategies 2–4 correspond to prioritizing those 20–40, 40–60, and >60 years old respectively. Strategy 4 leads to the greatest reduction in deaths, though all strategies perform better than no vaccination.

Figure 3

Comparison of benefit for four different vaccination strategies, against no vaccination. The relative reduction in deaths (A, B) and symptomatic infections (C, D) over a 5-year period are presented for four vaccination strategies, under varying speeds of vaccine dispensation. Results are stratified using three different vaccine efficacies, three types of control measure policy, and assuming a vaccine grants either sterilizing or non-sterilizing immunity. All simulations assumed vaccination did not exceed a population coverage of 75% and used an $R_0$ of 2.4. Baseline deaths were calculated assuming no control measures and the same R0 value.

Figure 4

Relative reduction in deaths using vaccination strategy four. Effectiveness of strategy 4 comparative to no vaccination and no control measures is given under varying dispensation speeds, and to different maximum population coverage levels, with and without control measures. Contour lines represent 25%, 50% and 75% reductions in cumulative deaths, comparative to no vaccination and no control measures. All simulations were performed using an $R_0$ of 2.4.