COVID-19 transmission model
We developed a Susceptible-Exposed-Infected-Recovered (SEIR) model for COVID-19 transmission stratified by age (0–4, 5–9 ,10–17, 18–29, 30–44, 45–59, 60–69, 70 + years), and vaccination status by dose and type (Fig. 1). Detail on modelled vaccine effectiveness by vaccine type, dose, outcome, circulating strain, and time since vaccination is presented in the Supplementary Material S1.2.
Table 1
Model parameters with symbols, descriptions, values, and sources.
Parameter
|
Value
|
Source(s)
|
Latent period (λ)
|
3.71 days (delta)
2.22 days (omicron)
|
[6]
[7]
|
Infectious period (δ)
|
10.9 days (delta)
9.87 days (omicron)
|
[8]
|
Effectiveness of infection-derived immunity* (ρ)
|
95%
|
[9]
|
Infection acute-care bed hospitalisation rate**
|
0.13% (wild type)
0.26% (delta)
0.20% (omicron)
|
[10–12]
|
Infection severity rate**
|
0.27% (wild type)
0.90% (delta)
0.52% (omicron)
|
Infection fatality rate**
|
0.034% (wild type)
0.079% (delta)
0.045% (omicron)
|
Increased risk of severe outcomes
|
1.95 (adults with comorbidities)
2.40 (pregnant women)
|
[13]
[14]
|
* Protection against same variant at two weeks, see Supplementary Material S1.3 for effect of immune escape variants and waning over time |
** Population-level rates, see Supplementary Material S1.4 for age-specific rates |
Characterisation of study setting
Sierra Leone is a low-income nation in Western Africa [15], with high-levels of non-COVID respiratory disease in all age groups, and limited access to health services. This context makes the mitigation of severe COVID-19 outcomes through vaccination a priority since capacity to implement other control measures is limited.
We extracted vaccine coverage from the Africa Centres for Disease Control and Prevention COVID-19 Vaccine Dashboard with distribution over time from John Hopkins Coronavirus Data Repository [2, 16]. United Nations population estimates for 2022 inform the age structure of our model [17]. Contact patterns were adapted from Prem et al. [18]. The effectiveness of non-pharmaceutical interventions over time was informed by the Oxford COVID-19 Government Response Tracker’s quantification of the strictness of government policies such as school closures, travel bans, and mandated mask-wearing [19]. We fitted the model to our study setting using seroprevalence estimates from March and November 2021 (see Supplementary Material S3).
Severe outcome projections
We used country-specific severity estimates to project from cases to incidence of severe disease, acute-care bed hospitalisation, and deaths (Table 1) [10]. These wild-type estimates were adjusted by variant- and age-specific multipliers [11, 12]. We calculated years of life lost using United Nations Population Prospect estimates of age-specific life expectancy [17]. We considered vaccine-derived protection against both maternal and pregnancy outcomes when modelling pregnant individuals. See the Supplementary Material S1.4 and S1.5 for further details.
Existing vaccine coverage in study setting
Sierra Leone has achieved moderate vaccine coverage as of October 2022–34.1% single-dose coverage and 25.6% fully vaccinated using AstraZeneca, Johnson & Johnson, Pfizer, and Sinopharm vaccines [2, 16]. The Government of Sierra Leone’s intention is to achieve 51.6% COVID-19 primary schedule coverage, the entire adult population, by the end of 2022 [20]. Booster doses are not yet widely administered in Sierra Leone with 0.1% coverage as of October 2022 [16]. Hence, the initial state of the model included 1,381,537 (single dose vaccine) or 2,763,074 (double dose vaccine) remaining doses to reach the 51.6% target.
Hypothetical vaccine rollout
We assumed a constant daily rollout capacity of 11,075 doses/day, a conservative estimate slightly lower than the pilot vaccination campaign in Sierra Leone October-November 2021 [20]. We model these doses as distributed in order of priority to eligible age groups until a ‘ceiling’ of vaccine acceptance (1-vaccine hesitancy) is reached. Sierra Leone has reported high but not universal COVID-19 vaccine acceptance – 87.9% (86.2–89.6) [21]. We present results for a range of rollout speeds considering the 2,654 active vaccinators identified by the Sierra Leone Ministry of Health and Sanitation in the Supplementary Material [22].
Our baseline assumption is that future doses will be a single-dose Johnson & Johnson vaccine. The greatest supply of COVID-19 doses in Sierra Leone has been Johnson & Johnson − 46.8% Johnson & Johnson compared to 24.9% Pfizer, 18.9% AstraZeneca, 6.4% Sinopharm, and 3.0% Sinovac (as of October 2022) [16]. The Government of Sierra Leone has demonstrated a strong preference for this vaccine, negotiating with major donors, such as the Government of Italy, to change their donations of double-dose vaccines (such as Pfizer) to Johnson & Johnson vaccines for ease of deployment across Sierra Leone [20]. We present sensitivity analysis comparing the prioritisation of a single-dose vaccine (Johnson & Johnson) to a double-dose vaccine (Pfizer) in the Supplementary Material.
Vaccine prioritisation strategies
We modelled age-, dose- and risk-prioritisation strategies. We examined the benefit of expanding the existing vaccination program to include children, either during or after the rollout of vaccines to eligible adults. We considered scenarios with sufficient vaccine supply for the delivery of a primary schedule to 51.6% (baseline), 60%, 70%, and 75.5% of the population. Note that 75.5% coverage is the highest level of population coverage expected with 88% vaccine acceptance if all aged over five were eligible for COVID-19 vaccination in Sierra Leone.
We considered the prioritisation of two high risk groups – pregnant women and adults with comorbidities – for earlier primary doses or additional booster doses. When considering booster dose delivery, we assumed all adults were eligible three months after their primary schedule, regardless of the vaccine type of their primary schedule. We identified high risk adults as adults with comorbidities aged 30–59 (1.2%) and all adults aged over 60 (4.7%) [20]. We estimated that pregnant women made up 3.3% of the population over a year [23]. We modelled the opportunistic vaccination of pregnant women using data on their presentation to antenatal care from the 2019 Demographic Health Survey in Sierra Leone [23]. We assumed that all pregnant women could be vaccinated within the 4.2 months, the median time to first antenatal care visit for pregnant women. This resulted in an opportunistic vaccination rate of 2,148/day. We did not model the opportunistic vaccination of adults with comorbidities since there is low access to primary care for the management of chronic conditions in Sierra Leone [24, 25]. We present sensitivity analysis for reduced vaccine effectiveness in older adults or adults with comorbidities, and for increased vaccine hesitancy in pregnant women in the Supplementary Material.