Modelling the Potential Public Health Impact of Different COVID-19 Vaccination Strategies with an Adapted Vaccine in Singapore

ABSTRACT

Background

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causing COVID-19 has been a dynamically changing virus, requiring the development of adapted vaccines. This study estimated the potential public health impact alternative vaccination strategies for COVID-19 in Singapore.

Research Design and Methods

The outcomes of alternative vaccination strategies with a future adapted vaccine were estimated using a combined Markov decision tree model. The population was stratified by high- and standard-risk. Using age-specific inputs informed by local surveillance data and published sources, the model estimated health (case numbers, hospitalizations, and deaths) and economic (medical costs and productivity losses) outcomes in different age and risk subpopulations.

Results

Booster vaccination in only the elderly and high-risk subpopulation was estimated to avert 278,614 cases 21,558 hospitalizations, 239 deaths, Singapore dollars (SGD) 277 million in direct medical costs, and SGD 684 million in indirect medical costs. These benefits increased as vaccination was expanded to other subpopulations. Increasing the booster vaccination coverage to 75% of the standard-risk population averted more deaths (3%), hospitalizations (29%), infections (145%), direct costs (90%), and indirect costs (192%) compared to the base case.

Conclusions

Broader vaccination strategies using an adapted booster vaccine could have substantial public health and economic impact in Singapore.

KEYWORDS:

• Booster vaccine
• COVID-19
• Omicron-adapted vaccine
• SARS-CoV-2
• Singapore

1. Introduction

Coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a highly contagious infectious disease with widespread health and economic consequences worldwide. Globally, as of 11 October 2023, over 771 million confirmed cases and nearly 7 million deaths due to COVID-19 have been reported. Locally, in Singapore, approximately 2.5 million cases and nearly 2,000 deaths attributed to COVID-19 have been reported from 3 January 2020 to 11 October 2023 [1].

Singapore was among the first nations to implement non-pharmaceutical intervention policies and launch a vaccination campaign against COVID-19 [2]. As of November 2022, Singapore’s vaccination coverage rate reached 81% [3], and over 15 million vaccine doses had been administered as of August 2023. Specifically, over 5 million people have received at least one dose, and over 5 million people have been vaccinated with a complete primary series [1].

While the emergence of the Omicron variant resulted in a large wave of COVID-19 cases with a peak from February to March 2022, severe outcomes were less than in other countries due to high vaccination coverage [4]. Recently, Singapore witnessed a spike in cases with a peak in April 2023 [5]. Given the waning of the effectiveness of original monovalent vaccines and their reduced effectiveness against Omicron variants [6–8], in 12 December 2022, vaccination efforts in Singapore was one of the first countries in Asia to approve and authorize the bivalent BNT162b2 vaccine targeting original and Omicron BA.4/BA.5 strains [9]. Bivalent vaccines confer protection against previous and current variants in circulation [10]. Therefore, booster vaccination with the Omicron-adapted bivalent vaccine has been suggested to be the optimal vaccination strategy in Singapore during the Omicron-predominant period [9]. In 2023, the Omicron subvariant XBB was the dominant variant in Singapore [11].

Given the increase in COVID-19 cases in Singapore that peaked in April 2023, policymakers making decisions regarding vaccination strategies could benefit from data estimating the impact of adapted booster vaccines in different age groups and risk subpopulations.

In this study, a health economic model was developed to estimate the outcomes of increasing the booster vaccination coverage of an adapted vaccine in Singapore. The aim of this study was to model the public health impact of alternative age (age ≥65, age ≥ 50, age ≥ 50 and age 5–11) and risk (high-risk, standard risk) based vaccination strategies on the number of infections, hospitalizations, and deaths as well as its economic impact on direct medical and indirect (i.e. productivity losses) costs. A hypothetical one-year time period was evaluated based on historical epidemiology data (Table S2). Results will be applicable to future adapted vaccines that have similar efficacy and duration of protection [12].

2. Methods

2.1. Model overview

A combined decision tree-Markov model was developed using Microsoft Excel as previously described [13] to compare vaccination with an adapted vaccine versus no booster vaccination in Singapore (Figure 1). At the start of the simulation, a single dose of the booster vaccine was administered, and the outcomes of infection with and without the booster dose over a 1-year time horizon were captured via the decision-tree component of the model. Patient outcomes were tracked over a time horizon of 1 year due to the limited vaccine duration of protection (DoP) and continuous evolution of the SARS-CoV-2 virus.

Figure 1. Model structure combining (a) Markov cohort model and (b) decision tree component.

In the Markov model component, individuals transition weekly through health states related to COVID-19 protection and infection. If individuals transition into any infection state, they immediately enter (B) the decision tree component for sorting. Green circles indicate decision points, with probabilities informed by the level of protection against COVID-19 in the particular infected state from which an individual arrives. Red triangles indicate decision tree end points, from which participants reenter the Markov model component.

A susceptible-infected-recovered (SIR) structure was adopted for the Markov component of the model similar to previous modeling studies aiming to predict the transmission and impact of infectious diseases, including COVID-19 [14–17]. At the start of the simulation, individuals lacking immunity were compartmentalized to the susceptible health state, individuals with vaccine-induced immunity (i.e. primary vaccination series) were compartmentalized to the vaccinated health state, and individuals with infection-induced immunity were compartmentalized to the recovered health state. The model assumed that infection-induced immunity and immunity induced by completing a primary vaccine series waned at a rate of the reciprocal of the DoP; subsequently, individuals transitioned to the susceptible health state. The transition from the susceptible health state to the infected health state was simulated based on age-specific yearly attack rates. The transition from the vaccinated health state to the infected health state was simulated based on age-specific attack rates adjusted according to the vaccine efficacy.

The adapted booster vaccine could be administered in all health states at a rate based on assumptions regarding vaccine coverage. Over a 1-year time horizon, individuals’ transitions across the health states were simulated according to their vaccination status, and individuals could transition to an alternative health state in each 1-week cycle based on the model parameters, which were derived from various sources as described in the Model Inputs section below.

In the infected health state, probabilities of infection, hospitalization, intensive care unit (ICU) admission, ventilation, and post-COVID condition were applied, and individuals exhibiting symptoms could receive outpatient treatment or be hospitalized in a regular ward or ICU with or without ventilation. Following recovery, individuals could develop post-COVID condition regardless of whether their infection was symptomatic in the infected health state. COVID-19 infection was considered the cause of any death occurring in the infected health state, while death occurring in another health state was considered all-other-cause mortality.

The model was constructed according to the International Society for Pharmacoeconomics and Outcomes Research and the Society for Medical Decision Making (ISPOR-SMDM) guidelines. Modeling experts involved in the model construction evaluated all inputs and results to ascertain face validity. Furthermore, experts in the field who were not involved in the model construction evaluated the model equations, inputs, and outputs to ensure internal validity.

2.2. Model inputs

2.2.1. Population inputs

The age distribution of the Singapore population was obtained from the Singapore Department of Statistics [18]. The population was stratified into eight groups by age as follows: 6 months-4 years, 5–11 years, 12–17 years, 18–29 years, 30–49 years, 50–64 years, 65–74 years, and ≥75 years (Table S1). To determine the impact of vaccination in different risk populations, the entire population was additionally divided into a standard-risk population (defined as the population comprising individuals aged <65 years with no comorbid conditions) and a high-risk population (defined as the population comprising individuals with at least one comorbid condition and individuals aged ≥65 years) Table S1). The proportion of the population assumed to be at a high risk in each age group was approximated by the comorbidity most prevalent by age as follows: asthma in the 5–29 years age group, obesity in the 30–49 years age group, and hypertension in the groups aged ≥50 years. Although this approach could result in an underestimation of the high-risk population, using an alternative approach may result in overcounting.

The prevalence of the considered comorbidities was informed by the literature describing the risk of COVID-19 in patients with comorbid conditions. The prevalence of obesity, smoking, hypertension, and chronic kidney disease was derived from the Ministry of Health Singapore [19]. The prevalence of asthma in children and adults was informed by Yang et al. [20] and Picco et al. [21], respectively. The prevalence of chronic obstructive pulmonary disease was obtained from George et al. [22]. Finally, the prevalence of cancer was informed by the World Cancer Research Fund International [23].

2.2.2. Infection inputs

At the start of the model horizon, the percentage of the population with natural immunity (i.e. infection-induced immunity) by age was informed by the Ministry of Health Singapore [24]. Due to the waning of the effectiveness of original monovalent vaccines, the proportion of the population with vaccine-induced immunity was assumed to be 10% of the vaccinated population as informed by the Ministry of Health Singapore [24]. The remaining population represented the population susceptible to COVID-19, which was calculated as follows: 100%─percentage of the population with natural immunity─10% of the vaccinated population. The yearly attack rate in the susceptible population was informed by He et al. [25] (the details of the infection inputs are shown in Table S2).

2.2.3. Vaccine coverage inputs

All individuals in Singapore’s population aged ≥5 years were eligible to receive the adapted vaccine. The shares of the eligible population receiving booster doses by age group were informed by the Ministry of Health Singapore [24] (Table S3).

2.2.4. Vaccine effectiveness inputs

The effectiveness of vaccines in conferring protection against infections, symptomatic infections, and severe disease as well as their duration of protection were constructed based on assumptions and real-world data [26–31] (Table S4).

2.2.5. Duration of protection inputs

Additionally, infection-induced immunity was assumed to confer protection against reinfection for a duration of three months based on evidence suggesting that following infection with Omicron, the time to potential reinfection ranges from 60–90 days [26,32,33].

2.2.6. Hospitalization inputs

The age-specific probabilities of an infection being symptomatic were derived from Ngiam et al. [34]. The age-specific probabilities of symptomatic patients being hospitalized were informed by data from the Ministry of Health Singapore [35]. Among hospitalized patients, the age-specific probabilities of admission to a critical care/ICU, the probabilities of receiving ventilation in either a regular ward or an ICU (Table S5), and the mortality rate (Table S6) were informed by the Ministry of Health Singapore [35].

2.2.7. Mortality inputs

The mortality rate among patients treated in the outpatient setting was assumed to be 0%. Due to limited data availability and the lack of granular data, the mortality rate could not be differentiated by ward (i.e. regular ward vs. ICU) or whether the patients received ventilation.

2.2.8. Relative risk for high-risk population

Because individuals with comorbid conditions are more likely to experience severe COVID-19, relative risk parameters deriver derived from the odds/hazard ratios take from the literature were applied to adjust for the increased risk of complications in this high-risk subpopulation. The risk of hospitalization was adjusted by 3.39 according to Mattey-Mora et al. [36], the risk of severe COVID-19 requiring hospitalization was adjusted by 1.96 according to Sim et al. [37], and the risk of mortality was adjusted by 4.50 according to Surendra et al. [38].

2.2.9. Long COVID inputs

Data related to long COVID in Singapore are scarce; therefore, US data derived from Menges et al. [39] were used to inform the probabilities of long COVID among asymptomatic patients (36.9%) and symptomatic patients receiving care in the outpatient (36.9%) and inpatient (45.7%) settings. The probability of developing long COVID did not vary by age.

2.2.10. Adverse events inputs

Vaccination with the adapted vaccine could result in adverse events (AEs). The probabilities of developing AEs following vaccination were informed by Hause et al. [40] and Klein et al. [41] (Table S7). Due to a lack of granular age-specific data, the same AE rates were applied to all age groups.

2.2.11. Direct medical cost inputs

The model considered the direct medical costs associated with COVID-19, including testing, treatment, hospitalization, AE, and post-COVID condition costs, from the Singapore payer perspective. As this study did not include a cost-effectiveness analysis, vaccine acquisition and administration costs were not considered. All costs are shown in Singapore Dollars (SGD). At the time of writing, 1 SGD was equivalent to 0.74 United States Dollars (USD) [42]. All costs were inflated to 2022 prices.

The model assumed that two COVID-19 tests, including one test to confirm the diagnosis and another test to confirm treatment, were administered to all patients regardless of their treatment setting at a cost of SGD 198.00 per test [43]. The model further assumed that asymptomatic patients did not attend any general practitioner (GP) visits, while symptomatic patients treated in the outpatient setting attended two GP visits at a cost of SGD 30 per visit [44]. These patients were further assumed to purchase over-the-counter pain medication at a cost of SGD 6.55 [45].

The costs of hospitalization were differentiated by ward. Inpatients treated in a regular ward were assumed to accrue total hospitalization costs of SGD 5,152 as informed by Cai et al. [46] and list of inpatient charges by Tan Tock Seng Hospital [47]. Inpatients treated in an ICU were assumed to accrue total hospitalization costs of SGD 15,840 based on Chew et al. [48] and list of inpatient charges by Tan Tock Seng Hospital [47]. Due to a lack of granular data, the hospitalization costs were not differentiated by age, whether the patient received ventilation, or whether the patient was susceptible or protected.

The long COVID costs include four GP visits, four COVID-19 tests, and one specialist visit at a cost of SGD 100 [44]. Of the patients experiencing post-COVID condition, 4.2% were assumed to be treated in a hospital setting and accrue total hospitalization costs of SGD 5,152 (i.e. the same total costs of inpatient admission to a regular ward).

The costs associated with the management of AEs caused by vaccine administration were informed by the Healthcare Cost and Utilization Project (HCUP) [49] and converted from USD (Table S7).

2.2.12. Indirect cost inputs

From the societal perspective, the model incorporated indirect costs defined as productivity losses caused by COVID-19 infection, post-COVID condition, or premature death. The inputs used to calculate productivity losses included age-specific workforce participation rates and labor costs per week derived from the Manpower Research and Statistics Department of the Ministry of Manpower [50] (Table S8). The productivity loss of asymptomatic patients was calculated as a working time loss of 0.4 days based on the following assumptions: 9% of asymptomatic patients were diagnosed with COVID-19, 44% of the asymptomatic patients diagnosed with COVID-19 could not work from home [51], and the working time lost of these patients was 10 days [52]. Because Singapore-specific data are unavailable, the productivity loss of symptomatic patients receiving care in the outpatient setting was assumed to be 10 days of working time lost based on the isolation guidelines by the Centers for Disease Control and Prevention [53]. The working days lost among hospitalized patients were assumed to be the same as the duration of a COVID-19 inpatient stay. Given the lack of country-specific data and granularity, the working time lost among patients receiving care in a regular ward or an ICU was assumed to be 8 days (regardless of whether ventilation was received) and 16 days (11 days with ventilation). Indirect costs (i.e. productivity losses) due to premature death were computed based on market productivity across patients’ lifetime based on their age group, estimated life expectancy based on life tables, workforce participation rate, and average wage rate.

2.3. Model outcomes

To determine the public health impact of booster vaccination with the adapted booster vaccine, the model estimated the number of COVID-19 cases (derived based on the attack rate), the number of hospitalizations due to COVID-19 (derived based on the hospitalization rate of infected patients), the number of outpatient cases (derived based on the rate of non-hospitalized cases), and the number of deaths (derived based on the mortality rate of hospitalized patients).

From the Singapore payer perspective, to determine the economic impact of booster vaccination, the model estimated the number of vaccine doses administered and COVID-19 related direct medical costs, including testing, AE management, treatment, and post-COVID condition-related costs. From the societal perspective, the model estimated productivity losses due to COVID-19.

Following the guidelines by the Agency for Care Effectiveness, the lifetime productivity losses due to premature death were discounted at an annual rate of 3% [54].

2.4. Analysis

In this study, the health and economic benefits of increased booster vaccination coverage of an adapted vaccine were evaluated in the Singapore population. Specifically, we evaluated the impact of increased booster vaccination in specific age and risk groups. The base case strategy comprised individuals with comorbid conditions and individuals aged ≥65 years (i.e. high-risk population), and coverage in this group remained constant. Two alternative base case strategies were explored: individuals with comorbid conditions and individuals aged ≥50 years; individuals with comorbid conditions and individuals aged ≥50 years and individuals age 5–11 years. Furthermore, the incremental benefits of increasing coverage in the standard-risk subpopulation to 25%, 50%, and 75% compared to no vaccination were estimated.

A sensitivity analysis was conducted to evaluate the impact of alternative burden of disease scenarios. In this analysis, the age-specific attack rate and hospitalization rate during Wave 2 (Delta dominant wave from 23 August 2021 to 19 September 2021) were compared with those during Wave 3 (Omicron dominant wave from 9 November 2021 to 6 December 2021) (Table S9). Wave 1 (from January 2021 to March 2021) was not included due to a lack of data.

A one-way deterministic sensitivity analysis (DSA) was conducted to explore parameter uncertainty in the base-case strategy. The DSA was conducted by varying over 350 parameters by ± 20% of the base-case value.

3. Results

3.1. Base-case analysis and expanding vaccination to other age and risk groups

This analysis sought to estimate the impact of alternative vaccination strategies based on age and risk. First, only the base case comprising individuals in the high-risk subpopulation (i.e. individuals aged ≥65 years and individuals considered at a high risk due to comorbid conditions) were considered eligible for a booster dose. In this analysis, 1,359,269 vaccine doses were delivered. Compared to no booster vaccination, the model estimated incremental health gains of 278,614 cases 21,558 hospitalizations, and 239 deaths averted. Furthermore, compared to no booster vaccination, the model estimated incremental direct medical cost savings of SGD 50 million in testing costs, SGD 132 million in inpatient treatment costs, SGD 17 million in outpatient treatment costs, and 79 million in long COVID treatment costs and an increase in the cost of treatment for AEs of SGD 168 thousand, resulting in direct medical cost savings of SGD 277 million. Finally, compared to no booster vaccination, the model estimated incremental indirect cost savings of SGD 683 million in productivity loss due to infection, SGD 1 million in productivity loss due to death, and < SGD 1 million in education loss, resulting in indirect cost savings of SGD 684 million (Table 1).

Table 1. Outcomes of vaccination in various age and risk groups.

	Base Case: Age ≥65 Years and High Risk*			Age >50 Years and High Risk**			Age >50 Years, Age 5–11 Years, and High Risk***
	No Vaccination	Vaccination	Incremental Impact	No Vaccination	Vaccination	Incremental Impact	No Vaccination	Vaccination	Incremental Impact
Vaccine doses administered (million)	0	1.36	1.36	0	1.61	1.61	0	1.67	1.67
Deaths	879	641	239	879	631	248	879	631	248
Hospitalizations	107,341	85,783	21,558	107,341	84,273	23,067	107,535	84,401	23,134
Infections (millions)	3.50	3.22	0.28	3.50	3.17	0.33	3.50	3.16	0.34
Direct medical costs (million SGD)	2,877	2,600	277	2,877	2,565	312	2,951	2,630	321
Testing cost (million SGD)	787	737	50	787	728	59	813	751	62
AE management cost (thousand SGD)	0	168	−168	0	199	−199	0	0	0
Inpatient treatment cost (million SGD)	642	510	132	642	501	141	643	501	142
Outpatient treatment cost (million SGD)	226	209	17	226	205	20	232	211	21
Post-COVID condition treatment cost (million SGD)	1,223	1,144	79	1,223	1,130	92	1,263	1,166	97
Indirect costs (million SGD)	11,975	11,292	684	11,975	11,156	819	11,975	11,153	823
Productivity loss due to illness (million SGD)	11,967	11,284	683	11,967	11,148	819	11,967	11,145	822
Productivity loss due to death (million SGD)	0.1	0.1	0.0	0.1	0.1	0.0	0.1	0.1	0.0
Education loss (million SGD)	8.1	7.9	0.2	8.1	7.9	0.2	8.4	8.0	0.4

*Base case coverage: standard-risk population: 0% (6mo-64 years), 88% (65–74 years), and 85% (≥75 years); high-risk population: 29% (5–11 years), 79% (12–17 years), 85% (18–29 years), 89% (30–49 years), 90% (50–64 years), 88% (65–74 years), and 85% (≥75 years).

**Base case coverage: standard-risk population: 0% (6mo-49 years), 90% (50–64 years), 88% (65–74 years), and 85% (≥75 years); high-risk population: 29% (5–11 years), 79% (12–17 years), 85% (18–29 years), 89% (30–49 years), 90% (50–64 years), 88% (65–74 years), and 85% (≥75 years).

***Base case coverage: standard-risk population: 29% (5–11 years), 90% (50–64 years), 88% (65–74 years), and 85% (≥75 years); high-risk population: 29% (5–11 years), 79% (12–17 years), 85% (18–29 years), 89% (30–49 years), 90% (50–64 years), 88% (65–74 years), and 85% (≥75 years).

Subsequently, to determine the impact of expanding coverage, the base case was expanded to include individuals aged ≥50 years in addition to those included in the base case above (high-risk population), which increased the number of vaccine doses delivered by 18%, averted 17%, 7%, and 4% more cases, hospitalizations, and deaths, respectively, and reduced the direct and indirect costs by 13% and 20%, respectively. The model estimated incremental direct medical cost savings of SGD 59 million in testing costs, SGD 141 million in inpatient treatment costs, SGD 20 million in outpatient treatment costs, and SGD 92 million in post-COVID condition treatment costs and an increase in the cost of treatment for AEs of SGD 199,000, resulting in direct medical cost savings of SGD 312 million. Furthermore, the model estimated incremental indirect cost savings of SGD 819 million in productivity loss due to infection, SGD 0.05 million in productivity loss due to death, and SGD 0.2 million in education loss, resulting in indirect medical cost savings of SGD 819 million (Table 1).

Finally, booster vaccine eligibility was expanded to include children aged 5–11 years in addition to those aged ≥50 years and the high-risk population, which increased the number of vaccine doses delivered by 23%, averted 23%, 7%, and 4% more cases, hospitalizations, and deaths, respectively, and reduced the direct and indirect costs by 16% and 20%, respectively. The model estimated incremental direct medical cost savings of SGD 62 million in testing costs, SGD 142 million in inpatient treatment costs, SGD 21 million in outpatient treatment costs, and SGD 97 million in post-COVID condition treatment costs and an increase in the cost of treatment for AEs of SGD 207,000, resulting in direct medical cost savings of SGD 321 million. Furthermore, the model estimated incremental indirect cost savings of SGD 822 million in productivity loss due to infection, SGD 0.05 million in productivity loss due to death, and SGD 0.4 million in education loss, resulting in indirect medical cost savings of SGD 823 million (Table 1).

3.2. Expanding vaccination in the Standard-risk Population

This analysis further sought to estimate the impact of booster vaccination strategies aiming to increase coverage in the standard-risk population (i.e. individuals aged 5–64 years with no comorbidities). Compared to the base case, increasing coverage to 25%, 50%, and 75% in the standard-risk population increased the number of vaccine doses delivered (42%, 85%, and 127%), prevented more deaths attributable to COVID-19 (1%, 2%, and 3%), averted more hospitalizations (10%, 19%, and 29%) and infections (48%, 97%, and 145%), and averted more direct (30%, 60%, and 90%) and indirect (64%, 128%, and 192%) costs (Table 2). Similar results were obtained in analyses considering alternative base cases expanding coverage in the standard risk population (Tables 3 and 4).

Table 5. Scenario analysis results varying the disease burden in analysis of the full Singapore population.

	Delta Variant (July 2021-September 2021)	Omicron Variant (January 2022-March 2022)*
Vaccine doses administered	3,241,970 (0%)	3,241,970
Deaths averted	10 (−96%)	248
Hospitalizations averted	1,673 (−94%)	28,639
Infections averted	29,912 (−96%)	721,807
Direct medical costs averted (million SGD)	26 (−95%)	552
Indirect costs averted (million SGD)	100 (−96%)	2,230

*base case.

Table 2. Base case results and expanding vaccination in the standard-risk population.

	Base case: Age ≥ 65 and High Risk*	Standard Risk Coverage: 25%	Standard Risk Coverage: 50%	Standard Risk Coverage: 75%
Vaccine doses administered	1,359,269	1,935,740 (42%)	2,512,212 (85%)	3,088,684 (127%)
Deaths averted	239	241 (1%)	244 (2%)	246 (3%)
Hospitalizations averted	21,558	23,616 (10%)	25,674 (19%)	27,732 (29%)
Infections averted	278,614	413,513 (48%)	548,411 (97%)	683,309 (145%)
Direct medical costs averted (million SGD)	277	361 (30%)	444 (60%)	527 (90%)
Indirect costs averted (million SGD)	684	1,120 (64%)	1,557 (128%)	1,994 (192%)

* Base case coverage: standard-risk population: 0% (6 months-65 years), 88% (65–74 years), and 85% (≥75 years); high-risk population: 0% (6 months-4 years), 29% (5–11 years), 79% (12–17 years), 85% (18–29 years), 89% (12–49 years), 90% (50–64 years) 88% (65–74 years), and 85% (≥75 years).

Standard-risk population comprises individuals aged <65 years with no comorbidities.

Table 3. Alternative base case results and expanding vaccination in the standard-risk population.

	Base case: Age > 50 and High Risk*	Standard Risk Coverage: 25%	Standard Risk Coverage: 50%	Standard Risk Coverage: 75%
Vaccine doses administered	1,605,692	2,113,974 (32%)	2,622,256 (63%)	3,130,538 (95%)
Deaths averted	248	248 (0%)	248 (0%)	248 (0%)
Hospitalizations averted	23,067	24,708 (7%)	26,348 (14%)	27,989 (21%)
Infections averted	327,314	448,737 (37%)	570,159 (74%)	691,581 (111%)
Direct medical costs averted (million SGD)	312	386 (24%)	459 (47%)	533 (71%)
Indirect costs averted (million SGD)	819	1,218 (49%)	1,618 (97%)	2,017 (146%)

*Base case coverage: standard-risk population: 0% (6 months-4 years), 0% (5–49 years), 90% (50–64 years) 88% (65–74 years), and 85% (≥75 years); high-risk population: 0% (6 months-4 years), 29% (5–11 years), 79% (12–17 years), 85% (18–29 years), 89% (12–49 years), 90% (50–64 years) 88% (65–74 years), and 85% (≥75 years).

Standard-risk population comprises individuals aged 5–49 years with no comorbidities.

3.3. Disease Burden Scenario Analysis

In a scenario analysis, two waves of COVID-19 occurring in Singapore with varied disease burden were compared. The alternative burden of disease scenarios varied in the attack rate and hospitalization rate. We compared wave 2, which was attributable to the Delta variant and occurred from 23 August 2021 to 19 September 2021, and wave 3, which was attributable to the Omicron variant and occurred from 9 November 2021 to 6 December 2021. The first wave of COVID-19, which occurred from January 2021 to March 2021, was not included in the analysis due to a lack of data.

In all age groups, the attack rate in wave 3 driven by the Omicron variant was remarkably higher than that in wave 2 driven by the Delta variant. In contrast, the hospitalization rate in wave 3 was substantially lower than that in wave 2 in most age groups, except for the 65–74 and ≥75 years age groups (Table S9).

Booster vaccination under the disease burden estimates (attack and hospitalization rates) from wave 2 was estimated to prevent 10 deaths, 1,673 hospitalizations 29,912 infections, SGD 26 million in total direct costs, and SGD 100 million in total indirect costs (Table 5). Meanwhile, booster vaccination under the disease burden estimates from wave 3 was estimated to prevent 248 deaths 28,639 hospitalizations, 721,807 infections, SGD 552 million in total direct costs, and SGD 2,230 million in total indirect costs.

Table 4. Alternative base case results and expanding vaccination in the standard-risk population.

	Base case: Age 5 mo-11 years or >50 years and High Risk*	Standard Risk Coverage: 25%	Standard Risk Coverage: 50%	Standard Risk Coverage: 75%
Vaccine doses administered	1,676,751	2,129,078 (33%)	2,581,405 (61%)	3,033,731 (89%)
Deaths averted	248	248 (0%)	248 (0%)	248 (0%)
Hospitalizations averted	23,134	24,750 (7%)	26,365 (14%)	27,981 (21%)
Infections averted	342,280	451,849 (38%)	561,417 (72%)	670,986 (105%)
Direct medical costs averted (million SGD)	321	388 (24%)	455 (46%)	522 (67%)
Indirect costs averted (million SGD)	823	1,222 (49%)	1,621 (98%)	2,020 (147%)

*Base case coverage: standard-risk population: 0% (6 months-4 years), 29% (5–11 years), 0% (12–49 years), 90% (50–64 years) 88% (65–74 years), and 85% (≥75 years); high-risk population: 0% (6 months-4 years), 29% (5–11 years), 79% (12–17 years), 85% (18–29 years), 89% (12–49 years), 90% (50–64 years) 88% (65–74 years), and 85% (≥75 years).

Standard-risk population comprises individuals aged 12–49 years with no comorbidities.

3.4. Deterministic sensitivity analysis

A one-way DSA was conducted to assess the impact of uncertainty in the parameters on the clinical and economic outcomes using the base case comprising the high-risk population and individuals aged ≥50 years. The top 10 parameters with the greatest influence on the results are shown in a tornado diagram in Figure 2. The most influential parameters include the workforce participation rate and labor cost per week in the 30–49 years age group, the cost of testing, and the probability of outpatients in the 30–49 years age group developing long-COVID condition.

Figure 2. Tornado diagram showing the 10 most impact parameters identified in the deterministic sensitivity analysis varying parameters by ± 20% on the total incremental cost savings in the high-risk population and individuals aged ≥ 50. Color of the bar indicates whether the parameter is at its lower or upper bound.

4. Discussion

Due to the limited duration of protection of COVID vaccines and the changing dynamics of the virus, vaccines adapted to the latest virus strains are an important public health strategy. The results of this study illustrated that booster vaccination with an adapted vaccine will have a significant positive public health and economic impact in Singapore, especially among those at a high risk of developing severe COVID-19.

The elderly and high-risk individuals with comorbid conditions have a higher probability of severe outcomes [36–38], so vaccination campaigns have traditionally targeted these populations [55]. However, the age cutoff to use for this approach is unclear with a range of cutoffs used such as 65 years [56] and 60 years [57]. In this study, we examined using both age 65 and age 50 as cutoffs for vaccination strategy. Compared to the base case defining the elderly as those aged ≥65 years, expanding vaccination to those aged ≥50 years and high-risk individuals increased the estimated benefits of booster vaccination. Additionally, our results demonstrated the potential public health benefits of expanding vaccination programs to target standard-risk individuals aged 5–64 years. We further explored different burden of disease from an alternative wave of COVID-19 to determine the sensitivity of results to the disease burden. The results were impacted by disease burden, highlighting the importance of estimating the future burden of disease. If future public health surveillance reveals major shifts in the burden of disease, it will be important to conduct additional analysis reflecting these changes. Similarly, as variants continue to shift and vaccines are adapted to the latest variant, it is important to monitor adapted vaccine performance. If adapted vaccine effectiveness changes for future variants, new analysis must be conducted reflecting these changes.

Because inputs in the model were primarily based on local data from Singapore, results may not be generalizable to other countries unless they have similar disease burden, vaccine environment, and variant circulation. However, similar analysis from other countries in the region such as Malaysia and Thailand have produced similar results [58,59].

5. Limitations

This study has several limitations. First, this study used a Markov structure to model infection. Studies modeling the impact of infectious diseases sometime use a dynamic transmission structure, but others have used a Markov approach [13,60,61]. Dynamic models are preferred to capture the indirect effect of vaccination on disease transmission, but little is known about the dynamics of disease transmission, so this study took the simpler approach of modeling probability of infection with and input parameter based on observed surveillance data. It has been noted that static models can be more suitable for capturing the economic benefits of vaccination programs [62]. SIR models omit several components that may affect disease dynamics such as reinfections due to overexposure, the use of nonpharmacological interventions, infection among vaccinated individuals, etc. Nevertheless, the SIR models have been used frequently to model COVID vaccination [14–17]. This study did not include vaccine costs for dose and administration, so no cost-effectiveness analysis was conducted. Inputs for vaccine effectiveness and duration of protection were derived based on sources that examined mRNA vaccines generally, however differences across brands will change results based on the uptake of each brand in Singapore. Further, it was assumed that vaccine effectiveness and duration of protection of a future adapted vaccine would be equivalent to that observed in real-world data from the Omicron-adapted vaccine [26–31], however vaccines adapted to future strains may have different effectiveness and duration of protection. Additionally, vaccine effectiveness was assumed to be the same for standard-risk and high-risk populations because estimates based real-world data do not stratify effectiveness by risk level [26–31]. However, if data emerges on differences, this may impact results. Only strategies using a single dose of the vaccine were examined based on the latest strategies. However, if multiple doses each year are considered, additional analysis would be necessary. For some parameters, Singapore-specific inputs were not available, resulting in the use of US data or assumptions. Costs related to AEs were converted from US data, potentially resulting in an overestimation. The prevalence of long COVID is an emerging area of study and exact estimates remains unclear, so US data was used. Additionally, estimating long COVID-related costs is challenging because its definition is unclear and the treatment for relies on patient symptoms rather than a diagnosis of long COVID. If true long COVID costs are higher or lower than assumed here, results will change accordingly. The comorbidity prevalence was estimated based on the most common risk factor per age group, which may have resulted in an underestimation of the high-risk prevalence by assuming that there is perfect overlap of comorbid conditions. The hospitalization and mortality data used to generate inputs for the model were taken from a period in which medications for COVID treatment were available and so these rates incorporate the effect of these medications. However, results may be sensitive to future advances in treatment that reduce hospitalization or mortality rates. The probability of hospitalization among the pediatric population was low and more recent data have suggested that this is increasing. Finally, it was assumed that the working time loss of asymptomatic patients was based our assumptions on the isolation guidelines by the US Centers for Disease Control and Prevention [53]. However, isolation recommendations may change in the future.

Despite these limitations, our results highlight the significant public health impact of expanding the vaccination strategy in Singapore. Expanding booster vaccination programs can reduce the burden on Singapore’s health system and avert substantial direct medical costs and societal costs. Based on these results, policy makers making decisions about vaccination campaign strategies should consider increasing vaccination coverage among standard-risk and younger populations to reduce cases, hospitalizations, medical costs and productivity losses.

6. Conclusions

The introduction and expansion of a vaccination campaign in the standard-risk population in Singapore would have substantial public health impact. The results of our analyses provide important estimates that can help decision makers optimize COVID-19 vaccination campaigns. Specifically, in Singapore a broader recommendation for vaccination to younger age groups and the standard-risk population would drive higher uptake in a larger population and this has the potential to have large public health and economic benefits. As the virus continues to evolve, it will be important to conduct additional studies to provide updated insight to policymakers.

Declaration of interest

K Thakkar, J Spinardi, MH Kyaw, J Yang, E Ozbilgili, B Taysi, and C Mendoza are employees of Pfizer and may hold stock or stock options of Pfizer. J Dodd and B Yarnoff are employees of Evidera, which received financial support from Pfizer in connection with the study and the development of this manuscript. Medical writing and editorial support was provided by Dr. Ruth Sharf-Williams at Evidera and was funded by Pfizer. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or material discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have received honoraria for their review work. Peer reviewers on this manuscript have no other relevant financial or other relationships to disclose.

Author contributions

All named authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship for this article. All authors contributed to study conception and design, data acquisition, analysis, and interpretation, drafting and revising of the manuscript.

Acknowledgments

We would like to thank Ervin Cheong, Medical Scientific Specialist, Pfizer Singapore for his support. Assistance with model conceptualization and development and input collection was provided by Solene De Boisvilliers and Lucie Bouin (Evidera). Medical writing was provided by Dr. Ruth Sharf-Williams (Evidera) and was funded by Pfizer, Inc.

Supplemental Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14760584.2023.2290931

Additional information

Funding

This paper was funded by Pfizer.