A spatial and dynamic solution for allocation of COVID-19 vaccines when supply is limited.

Background: Since most of the global population needs to be vaccinated to reduce COVID-19 transmission and mortality, a shortage of COVID-19 vaccine supply is inevitable. We propose a spatial and dynamic vaccine allocation solution to assist in the allocation of limited vaccines to people who need them most.
Methods: We developed a weighted kernel density estimation (WKDE) model to predict daily COVID-19 symptom onset risk in 291 Tertiary Planning Units in Hong Kong from 18 January 2020 to 22 December 2020. Data of 5,409 COVID-19 onset cases were used. We then obtained spatial distributions of accumulated onset risk under three epidemic scenarios, and computed the vaccine demands to form the vaccine allocation plan. We also compared the vaccine demand under different real-time effective reproductive number (Rt) levels.
Results: The estimated vaccine usages in three epidemiologic scenarios are 30.86% - 45.78% of the Hong Kong population, which is within the total vaccine availability limit. In the sporadic cases or clusters of onset cases scenario, when 6.26% of the total population with travel history to high-risk areas can be vaccinated, the COVID-19 transmission between higher- and lower-risk areas can be reduced. Furthermore, if the current Rt is increased to double, the vaccine usages needed will be increased by more than 7%. Conclusions: The proposed solution can be used to dynamically allocate limited vaccines in different epidemic scenarios, thereby enabling more effective protection. The increased vaccine usages associated with increased Rt indicates the necessity to maintain appropriate control measures even with vaccines available.

Introduction

COVID-19 has become a global challenge for human beings[1,2]. As of 5 May 2021, COVID-19 has spread rapidly in 222 countries and regions[3], and 153,526,293 people have been diagnosed with this disease, with a total of 3,213,701 deaths[3]. Since the outbreak of the COVID-19 pandemic[4], many countries have successively adopted social distancing and other transmission mitigation measures to reduce the risk of others being infected[5,6]. Most people still lack immunity to the COVID-19 virus (SARS-CoV-2) and are thus still vulnerable to SARS-CoV-2 infection[7-10].

Therefore, ensuring effective SARS-CoV-2 vaccination against COVID-19 is of the highest priority[11,12]. Such a move could possibly end the COVID-19 pandemic and accelerate the recovery of the global economy[13].

Currently, 280 candidate vaccines are being developed around the world, of which 96 vaccines, according to WHO data[14] on 4 May 2021, are at the clinical trial stage. Although over 194 countries and states have received more than 841 million vaccine doses and have already started strategic vaccination distribution[15], not all countries and states are equally well-equipped with medical resources. More than 700 million doses of vaccine have been given globally, but, it appears that more than 87% of vaccinations are concentrated in high-income and upper-middle-income countries, while low-income countries account for only 0.2%. In high-income countries, on average, close to 1 in 4 people have been vaccinated, whereas in low-income countries the figure is only 1 in 500[16,17]. Disparities were further discovered to exist within the nation. It also found evidence of divides along socio-economic lines[18]. On this basis, the WHO has led the launch of the COVID-19 Vaccines Global Access Facility (COVAX Facility)[19] to ensure an equitable allocation of vaccines around the world to cover the most vulnerable 20%[20] of the population of each country participating, in particular the lower-income countries. Currently, 190 counties/regions[21] have participated in the COVAX Facility. Based on this fact, the Strategic Advisory Group of Experts on Immunisation, World Health Organisation (WHO SAGE) has also been proposed to various countries and regions for reference, a vaccine allocation roadmap[22] compatible with the COVAX Facility. However, the specific feasibility of the allocation roadmap, especially regarding how to carry out dynamic and reliable spatiotemporal vaccine allocation, is still an important consideration.

The formulation of vaccine allocation plans during an infectious disease pandemic is critical in terms of public health response[23]. During the past few influenza pandemics, some countries have chosen the strategy of allocating available vaccine supplies to each region in proportion to the population of that region[23]. Currently, there are a number of guidelines on how to allocate vaccines fairly and thus protect the rights and interests of various groups of people, including those from the WHO SAGE; Johns Hopkins School of Public Health[24] (JHSPH); National Academies of Sciences, Engineering, and Medicine[23] (NASEM), US; Advisory Committee on Immunisation Practices[25] (ACIP), US; and many other authoritative organisations. The degree of application of these guidelines in specific areas, especially at the urban community scale, is still unknown. Further, COVID-19 shows a typical spatial variation trend which changes dynamically[26,27]. Therefore, knowing how to adjust vaccine allocation spatially and dynamically, to cope with the dynamic changing trend of the COVID-19 pandemic is also essential.

The overall aim is to help low-income countries achieve the following goals by allocating the limited supply of vaccines both spatially and dynamically, but with the following in mind: (i) effectiveness: Care must be taken to ensure vaccines are limited to people in direct need, thereby providing the most effective protection by avoiding further rapid spread, as indicated above; (ii) fairness: To ensure the most vulnerable or susceptible subgroups are considered during each stage of the vaccine allocation; (iii) reasonableness: To assess the impact of vaccination, control measures, and behaviour changes on the spatiotemporal distribution of onset risk to better ensure effective vaccine allocation.

In addition, to the above, the costs of vaccines and cold chain logistics with spatial variation, spatiotemporal factors should be considered in the formulation and implementation of vaccine distribution strategies[28]. Thus, if the supply is limited it is urgent to formulate a spatial and dynamic allocation solution for the COVID-19 vaccine at the urban community scale, and hence better enable the achievement of overall vaccination effectiveness with a limited vaccine supply.

The trend of the symptom onset risk of potential patients can better reflect the risk level of the COVID-19 pandemic[29] due to the lengthy incubation period[30] of SARS-CoV-2, and the normally delayed corresponding treatment[31]. Therefore, to spatially and dynamically formulate an effective vaccine allocation plan, appropriate data-driven spatial models need to be adopted to dynamically identify the onset risk level in each region. The weighted kernel density estimation (WKDE) model is one such model[32] and performs retrospective analysis based on the spatiotemporal information of the cases to infer the date of infection of each onset case[32]. In this way, the risk of infection in a specific location caused by the spatial movement of infected individuals can be predicted. Furthermore, based on the characteristics of human-to-human transmission of COVID-19, dynamic mobility data were introduced into the model to enable reliable and location-specific predictions. This adapted model is referred to as the intercity-scale extended WKDE model[29]. This data-driven spatial model can reduce dependence on theoretical assumptions and parameters, suitable for COVID-19, the mechanism of transmission of which is still under study. Currently, however, there are two areas to be further developed for the extended WKDE model: (a) consideration of vaccination, control measures and behavioural changes in the new normal of the COVID-19 pandemic, and (b) further examination of the effectiveness of utilising finer intraurban community scales, even though the current models are performing well, in general, regarding county/intercity scales. As a result, an urban-community-scale WKDE model has been developed during this study. The aim, when taking into account the vaccination, control measures, and behavioural changes in cities worldwide is to influence the symptom onset risk of COVID-19. The newly developed WKDE model will further include the real-time effective reproduction number (Rt) to both indicate real-time transmissibility and quantify[33] the effect of vaccination, control measures, and behavioural changes in the model.

In-line with the above, the COVID-19 vaccine allocation plan for Hong Kong (one of the 190 COVAX Facility counties/regions) is to be formulated taking three epidemiologic setting scenarios into account[22]: (i) no local onset cases, (ii) sporadic cases or clusters of local onset cases, and (iii) community transmission. The newly developed urban-community-scale WKDE model is used to predict COVID-19 symptom onset risk in each of the 291 Tertiary Planning Units (TPUs) in Hong Kong, covering a total of 1106 km2[34] (Fig. 1). Five thousand four hundred and nine COVID-19 onset cases[35] data in Hong Kong from 18 January 2020 to 22 December 2020 is used in this study. The vaccine demand is then estimated for two different situations: (a) before and or (b) during the process of vaccine allocation, under the impact of daily vaccination, or control measures and behavioural changes. Furthermore, the vaccine usage (in proportion to the population vaccinated) needed under different real-time effective reproduction number (Rt) levels, is also be simulated and estimated.

Fig. 1

The daily variation in COVID-19 symptom onset cases[35] in Hong Kong from 18 January 2020 to 22 December 2020 for the three epidemiologic setting scenarios.

The blue, orange, and grey dots represent daily onset cases of Hong Kong in the three scenarios of community transmission, sporadic cases and clusters of local onset cases, and no local onset cases, respectively.

We show that the estimated vaccine usages (30.86–45.78%) in Hong Kong in three epidemiologic scenarios are within the limit of total vaccine availability. The vaccine usages needed will be increased if the real-time effective reproduction number Rt is increased. Our study demonstrates the feasibility of the proposed solution for allocating limited COVID-19 vaccines spatially and dynamically in different epidemic scenarios.

Methods

Data sources

A total of 3977 COVID-19 onset cases with spatiotemporal information in Hong Kong during the period 18 January 2020 to 16 November 2020 were collected from official reports[33] of the Department of Health of Hong Kong. Excluding the imported onset cases receiving compulsory quarantine and onset cases with unknown location information, 3316 local onset cases were used for this study (Fig. 1). For these 3316 cases, we obtained the available information on dates of onset and reporting as well as the community-level locations, where these patients had stayed prior to diagnosis. According to the transmission categories corresponding to epidemic setting scenarios, the 3316 local onset cases in Hong Kong were classified into three epidemiologic scenarios[22]: (i) no local onset cases (31%), (ii) sporadic and cluster of local onset cases (46%), and (iii) community transmission (23%).

In addition, in order to further simulate the impact of the protection of the population formed by vaccination on the spatiotemporal trend of onset risk, the 2093 local onset cases[35] with community-scale geographic location from 17 November 2020 to 22 December 2020 were used for this study (Fig. 1). Currently, the main COVID-19 vaccine used in Hong Kong is Pharma/BioNTech Comirnaty COVID-19 mRNA Vaccine-BNT162b2[36] (also one of the main vaccines allocated by the COVAX facility). It has a vaccine efficacy of 95.0%. Forty-two thousand people can be vaccinated every day[36]. It takes time after vaccination for antibodies to develop in the body and offer protection. Individuals may not be fully protected until 7 days after their second vaccine dose[36]. Hence this study assumes that (i) the interval between the first dose and the second dose is 21 days[36]; (ii) It takes time after vaccination for antibodies to develop in the body and offer protection against COVID. Individuals will be fully protected after 7 days after their second dose of vaccine[36]; (iii) the vaccine allocations in Stage I (very limited vaccine availability accounting for 1–10% of the city’s population) would be in Hong Kong from 16 November, 2020. So accordingly, the Stage II (limited vaccine availability accounting for 11–20% of the city’s population) would start on 4 December 2020, and the Stage III (moderate vaccine availability accounting for 21–50% of the city’s population) would start on 22 December 2020.

Based on the daily traffic flow data[37] of 575 closed circuit televisions (CCTV) and traffic detectors covering all Hong Kong strategic routes in Hong Kong, during the same period in 2020 (from 18 January 2020 to 22 December 2020), the traffic flow data within a TPU and between TPUs are used in this research to indicate human mobility within a particular TPU and that from other TPUs to this TPU. These data mainly include two parts: traffic flow data within a particular TPU and that between this TPU and other TPUs. In addition, community-scale daily human mobility data[38,39] (Fig. 2) during the same period in 2020 (from 18 January 2020 to 22 December 2020), provided by Apple Maps and Google, were used to improve the traditional SIR model to calculate real-time effective reproduction number Rt for local cases in Hong Kong.

Fig. 2

The daily variation in human mobility[38,39] in Hong Kong from 13 January 2020 to 22 December 2020.

The blue dots represent daily community-level human mobility of Hong Kong.

The other 18 categories of the statistics data[40] in 291 TPUs (Supplementary Table S1), such as the number of medical workers, elderly individuals, school staff, other essential workers outside the health and education sectors, low-income groups, and immigration staff in Hong Kong, were also used in this study.

The solution to COVID-19 vaccine allocation at the community scale

A solution for how to allocate the COVID-19 vaccine spatially and dynamically in the context of its limited early supply is proposed in this study. The solution is composed of (1) the principles of the allocation, which is based on the WHO SAGE roadmap[22] for the allocation of COVID-19 vaccines; (2) an urban-community-scale WKDE model for predicting COVID-19 symptom onset risk for different epidemiologic scenarios: (i) no local onset cases, (ii) sporadic cases or clusters of local onset cases, and (iii) community transmission; and (3) an urban-community-scale COVID-19 vaccine allocation strategy for the three scenarios based on these principles and the onset risk prediction results. Furthermore, the impact of the different real-time effective reproduction number levels on vaccine demand and allocation are also evaluated.

Principles of vaccine allocation

In this study, the WHO SAGE roadmap[22] for prioritising the administration of the COVID-19 vaccine in the context of its limited supply is adopted as the overall strategy for COVID-19 vaccine distribution at the urban-community-scale. The reason for this adoption of the WHO SAGE roadmap[22] as our principle is that we share its goals and can realise these goals quantitatively: (a) reduce the mortality, morbidity, and infection rate brought by the COVID-19 pandemic to cities; (b) prioritise ensuring that key population subgroups have equal access to the vaccine; and (c) reduce the total cost of vaccination and transportation. It should be noted that the assumptions in the WHO SAGE roadmap[22] are also applicable in this solution.

A community-scale WKDE model for predicting the onset risk of COVID-19 symptoms

As a further development of the intercity-scale extended WKDE model[29], the community-scale WKDE model proposed, includes the following three steps:

(a) Conducting a retrospective analysis of the historical existence likelihood of the infection in each community location, in which an onset case occurred,

(b) Making inferences concerning the historical existence of the likelihood of the infection spreading throughout the entire city, and

(c) Predicting the entire city’s future potential epidemic onset risk on any one given day.

The main difference between the urban-community-scale WKDE model and the extended intercity-scale WKDE model is that at step b) of the model, the historical existence likelihood of the infection occurring at a random location in the entire city has been formulated as follows:where PInfection(S, ti) is the probability of any individual, l in the city on day ti, infected with COVID-19 infecting others in random locations, S, Lj is the jth location places where onset cases have occurred[29]. PInfection(L, ti) denotes the probability that one onset case had been infected on day ti in location L[29]. Kh(S – Lj) denotes a Gaussian kernel[29].

Rt(ti) denotes the real-time effective reproductive number for local cases in the city on day ti[41]. The values of Rt(ti) indicates real-time transmissibility. As there is a delay of several days between an infection and the case report, the traditional framework is unable to provide a real-time estimate of Rt. Leung et al. proposed a new method[41], whereby human mobility data was used to improve the traditional SIR models to conduct real-time monitoring of community-level transmissibility to obtain real-time estimates of daily Rt in the following three steps[41]:

Estimate the instantaneous reproduction number Rt of local cases in the city from day t1 to day ti.

Correlate the time series (post-average value) of the obtained empirical Rt estimation with the daily variations of human mobility obtained from different types of data sources in the city.

Select the human mobility values with a high correlation coefficient with Rt to represent the overall human mobility trend of the city to improve the traditional SIR (susceptible-infected-recovered) epidemic model. That is, on the t day, by using the overall human mobility trend as the scaling factor of the SIR model contact matrix, the epidemic curve from day t1 to day ti is further obtained.

However, it needs to be pointed out that in step iii, when the vaccination rate in the city is constantly changing, the removals R will increase accordingly, and the susceptibility of S will increase. Therefore, the impact of daily variation in the vaccination rate on the epidemic curve from day t1 to day ti is noted.

Of interest, in this study, the real-time effective reproductive number (Rt(ti)) for local cases in Hong Kong from 18 January 2020 to 22 December 2020 was estimated (Fig. 3). In Hong Kong, the 95% uncertainty interval of the daily Rt value was also estimated (Fig. 3). Note that Rt(ti) after implementation date t239 in Hong Kong of the current real-time effective reproductive number level is calculated differently in the following three situations: (a) maintaining the current real-time effective reproductive number level, (b) decreasing the current real-time effective reproductive number level, and (c) increasing the current real-time effective reproductive number level. Thus, Rt(ti) after t239 is set equal to an actual real-time effective reproductive number, while maintaining the current real-time effective reproductive number level. Rt(ti) after t239, is set equal to half of the real-time effective reproductive number, while decreasing the current real-time effective reproductive number level. Rt(ti) after t239, is set equal to double the value of the real-time effective reproductive number, while increasing the current real-time effective reproductive number level.

Fig. 3

The daily variation in real-time effective reproductive number (Rt(ti))[41] for local cases of Hong Kong from 18 January 2020 to 22 December 2021.

The blue line and shades represent the mean and 95% uncertainty interval of daily Rt[41].

Kh(S – Lj) denotes a Gaussian kernel[29]:where h denotes the d × d bandwidth matrix. By generalising Scott’s rule of thumb, the d × d bandwidth matrix is chosen here[32,42-44] with n being the total number of onset cases in all 339 days, and being the covariance matrix of the onset cases sample[29]. For the bivariate distributed onset case samples, the value of d is set as 2[29]. So the bandwidth matrix h is equal to .

Mintra_TPU(S, ti) denotes a human mobility factor[29] within a TPU containing location S on day ti, calculated as follows:where Wk denotes the daily traffic flow within the TPU containing location S on day tk prior to ti.

MinterTPU(S, ti) denotes a human mobility factor[29] from other TPUs to the TPU containing location S, calculated as follows:where Vk denotes the daily traffic flow from other TPUs to the TPU containing location S on day tk prior to ti.

Finally, the daily predicted risk in each location was standardised[29] to a value between 0 and 1 on a specific date. Different levels of onset risk were set as follows[29]: low onset risk [0–0.2], low-medium onset risk [0.2–0.4], medium onset risk [0.4–0.6], medium-high onset risk [0.6–0.8], and high onset risk [0.8–1]. Similar to the extended intercity-scale WKDE model, the reliability of the predicted COVID-19 onset risk was evaluated by its spatial significance[29], i.e., the percentage of onset cases on a future date to be predicted that occur in the high onset risk areas[29] (identified onset hotspot).

Next, based on the daily onset risk prediction results in a specific time period, the illness onset risk prediction was derived for the different epidemiologic scenarios: (i) no local onset cases, (ii) sporadic cases or clusters of local onset cases, and (iii) community transmission. When a new daily onset risk prediction result is added later, the accumulated onset risk in each scenario can be adjusted dynamically.

COVID-19 vaccine allocation at the urban community scale

The final allocation of the vaccine at the urban-community-scale is based on the spatial prediction model of COVID-19 onset risk. The allocation strategy was set for three epidemiologic setting scenarios: (i) no local onset cases, (ii) sporadic cases or clusters of local onset cases, and (iii) community transmission. The key subgroups in terms of high infection risk and the predicted spatiotemporal distribution of COVID-19 onset risk in each scenario were comprehensively considered. As a result, spatial and dynamic vaccine allocation at the urban-community-scale was achieved.

In this study, an overall urban-community-scale COVID-19 vaccine allocation plan for the three scenarios (Supplementary Table S2–4) was formulated. For each scenario, a vaccine allocation plan was made according to three possible vaccine supply stages: Stage I: very limited vaccine availability accounting for 1–10% of the city’s population; Stage II: limited vaccine availability accounting for 11–20% of the city’s population; and Stage III: moderate vaccine availability accounting for 21–50% of the city’s population.

Compared with the roadmap of WHO SAGE, based on the characteristics of the proposed urban-community-scale onset risk model, the following improvements have been made to the allocation plan: (a) the consideration of the spatiotemporal distribution of onset risk was highlighted (the susceptible population in high-risk communities was protected first); (b) when considering a high-risk community, other communities with close contacts were also considered; and (c) in Stage III of the sporadic cases or clusters of local onset case scenarios, people travelling to high-risk areas for work were also considered (cross-district staff within communities with high onset risk and with close contacts to communities with high onset risk).

Note that in the application of the Hong Kong solution, given above, the community refers to TPUs. In addition, according to the mortality rate of COVID-19 cases in various age groups in Hong Kong, elderly individuals were defined in the age-based risk group as those over 80 years old. The mortality rates for the different groups were 29.86% (aged 80 years or above), 17.13% (aged 70 years or above), and 11.58% (aged 65 years or above).

Table 1

COVID-19 vaccine usage in the no local onset cases scenario.

Vaccine supply scenario	Priority groups	Vaccine usage in each substage (%)	Vaccine usage in each stage	Total vaccine usage (%)
Stage I (Very limited vaccine availability accounting for 1–10% of the city’s population)	Stage Ia: Front-line medical workers in communities with high onset risk and with close contacts to communities with high onset risk	0.28	0.40% (With 9.60% for essential travellers at risk + emergency reserve in Stage I)	0.40
	Stage Ib: Border protection staff and workers for outbreak management	0.12
	Stage Ic: Essential travellers facing risk of infection outside Hong Kong Stage Id: Emergency reserve utilisation for focused outbreak response	9.60
Stage II (Limited vaccine availability accounting for 11–20% of the city’s population)	Stage IIa: Front-line medical workers in the remaining low-risk to medium-high-risk communities	0.69	3.35% (With 6.65% for remaining travellers at risk + emergency reserve in Stage II)	3.75
	Stage IIb: Elderly individuals aged 80 years or above with medium-high or higher onset risk and communities with close contacts to communities with medium-high or higher onset risk	2.66
	Stage IIc: Remaining travellers facing risk of infection outside Hong Kong Stage IId: Emergency reserve of vaccines utilisation for outbreak mitigation	6.65
Stage III (Moderate vaccine availability accounting for 21–50% of the city’s population)	Stage IIIa: Elderly individuals aged 80 years or above in communities with low to medium onset risk	1.98	27.11	30.86
	Stage IIIb: School staff	1.19
	Stage IIIc: Other essential workers outside the health and education sectors	23.94

Table 2

COVID-19 vaccine usage in the sporadic or clusters of local onset cases scenario.

Vaccine supply scenario	Priority groups	Vaccine usage in each substage (%)	Vaccine usage in each stage	Total vaccine usage (%)
Stage I (Very limited vaccine availability accounting for 1–10% of the city’s population)	Stage Ia: Front-line medical workers in communities with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	0.61	3.67% (With 6.33% for emergency reserves)	3.67
	Stage Ib: Elderly individuals aged 80 years or above in communities with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	3.06
	Stage Ic: Emergency reserve of vaccines for utilisation in outbreak response or mitigation	6.33
Stage II (Limited vaccine availability accounting for 11–20% of the city’s population)	Stage IIa: Front-line medical workers in communities with low to medium onset risk	0.36	10.74%	14.41
	Stage IIb: Elderly individuals aged 80 years or above in communities with low to medium onset risk	1.58
	Stage IIc: Groups with comorbidities within communities with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	0.41
	Stage IId: Low-income groups in communities with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	1.46
	Stage IIe: Other essential workers outside the health and education sectors in communities with high onset risk and with close contacts to communities with high onset risk	6.93
Stage III (Moderate vaccine availability accounting for 21–50% of the city’s population)	Stage IIIa: School staff in communities with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	0.47	25.30%	39.71
	Stage IIIb: Remaining low-income groups in communities with low to medium onset risk	1.56
	Stage IIIc: Remaining essential workers outside the health and education sectors in communities with low to medium-high onset risk	17.01
	Stage IIId: Cross-district staff within communities with high onset risk and with close contacts to communities with high onset risk	6.26

Table 3

COVID-19 vaccine usage in the community transmission scenario.

Vaccine supply scenario	Priority groups	Vaccine usage in each substage (%)	Vaccine usage in each stage (%)	Total vaccine usage (%)
Stage I (Very limited vaccine availability accounting for 1–10% of the city’s population)	Stage Ia: Front-line medical workers	0.97	5.61	5.61
	Stage Ib: Elderly individuals aged 80 years or above	4.64
Stage II (Limited vaccine availability accounting for 11–20% of the city’s population)	Stage IIa: Elderly individuals not covered in the first stage (aged 65 years or above)	9.33	14.21	19.82
	Stage IIb: Groups with comorbidities	2.91
	Stage IIc: Low-income groups in communities with high onset risk and with close contacts to communities with high onset risk	1.89
	Stage IId: Medical workers engaged in immunisation delivery	0.08
Stage III (Moderate vaccine availability accounting for 21–50% of the city’s population)	Stage IIIa: School staff	1.19	25.96	45.78
	Stage IIIb: Remaining low-income groups with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	0.83
	Stage IIIc: Other essential workers outside the health and education sectors	23.94

Table 4

COVID-19 vaccine usage in the community transmission scenario after vaccines start to be allocated.

Vaccine supply scenario	Priority groups	Vaccine usage in each substage (%)	Vaccine usage in each stage (%)
Stage I (Very limited vaccine availability accounting for 1–10% of the city’s population)	Stage Ia: Front-line medical workers	0.97	5.61
	Stage Ib: Elderly individuals aged 80 years or above ≥65 years	4.64
Stage II (Limited vaccine availability accounting for 11–20% of the city’s population)	Stage IIa: Elderly individuals not covered in the first stage (aged 65 years or above)	9.33	14.45
	Stage IIb: Groups with comorbidities	2.91
	Stage IIc: Low-income groups in communities with high onset risk and with close contacts to communities with high onset risk	2.13
	Stage IId: Medical workers engaged in immunisation delivery	0.08
Stage III (Moderate vaccine availability accounting for 21–50% of the city’s population)	Stage IIIa: School staff	1.19	25.69
	Stage IIIb: Remaining low-income groups with medium-high or higher onset risk and with close contacts to communities with medium-high or higher onset risk	0.56
	Stage IIIc: Other essential workers outside the health and education sectors	23.94

Table 5

COVID-19 vaccine usage in the no local onset cases scenario for situations of different real-time effective reproduction number levels.

Vaccine supply scenario	Total vaccine usage—Maintaining the current real-time effective reproduction number level	Total vaccine usage—Increasing the current real-time effective reproduction number level	Total vaccine usage—Decreasing the current real-time effective reproduction number level
Stage I (Very limited vaccine availability accounting for 1–10% of the city’s population)	0.36% (With 9.64% for essential travellers at risk + emergency reserve in Stage I)	0.64% (With 9.36% for essential travellers at risk + emergency reserve in Stage I)	0.23% (With 9.77% for essential travellers at risk + emergency reserve in Stage I)
Stage II (Limited vaccine availability accounting for 11–20% of the city’s population)	3.48% (With 6.88% for essential travellers at risk + emergency reserve in Stage II)	4.54% (With 6.10 % for remaining travellers at risk + emergency reserve in Stage II)	2.10% (With 8.13% for remaining travellers at risk + emergency reserve in Stage II)
Stage III (Moderate vaccine availability accounting for 21–50% of the city’s population)	30.86%	30.86%	30.86%

Table 6

COVID-19 vaccine usage in the sporadic cases or clusters of local onset cases scenario for situations of different real-time effective reproduction number levels.

Vaccine supply scenario	Total vaccine usage—Maintaining the current real-time effective reproduction number level	Total vaccine usage—Increasing the current real-time effective reproduction number level	Total vaccine usage—Decreasing the current real-time effective reproduction number level
Stage I (Very limited vaccine availability accounting for 1–10% of the city’s population)	2.47% (With 7.53% for emergency reserve)	4.15% (With 5.85% for emergency reserve)	1.14% (With 8.86% for emergency reserve)
Stage II (Limited vaccine availability accounting for 11–20% of the city’s population)	12.02%	18.14%	7.77%
Stage III (Moderate vaccine availability accounting for 21–50% of the city’s population)	39.28%	46.60%	34.59%