Preplanned Studies: Association Between COVID-19 Vaccination Coverage and Case Fatality Ratio: a Comparative Study — Hong Kong SAR, China and Singapore, December 2021–March 2022

Hong Kong Special Administrative Region (SAR), China and Singapore have been facing the Omicron epidemic since January 2022. Despite their similar socioeconomic statuses, Hong Kong SAR, China encountered a higher case fatality ratio (CFR) accompanied by an overwhelmed medical system compared to Singapore. This study collected and compared the CFRs, vaccination coverage, infection rate, and medical resources occupancy rates from the two places. The vaccine coverage rates and CFRs in 52 Omicron variant dominated countries and regions were also collected and multiple linear regressions were used to explore associations. The random forest (RF) method was additionally used to estimate the importance of factors contributing to CFRs. The CFR in Hong Kong SAR, China was larger than in Singapore (0.53% vs. 0.06%), especially for people ≥70 years (4.46% vs. 0.48%). Correspondingly, the vaccine coverage was lower in Hong Kong SAR, China than in Singapore (full vaccine coverage for total population: 76.15% vs. 92.00%; for people ≥70 years: 56.82% vs. 95.00%). Based on the data from Omicron variant dominated countries and regions, a negative association between vaccine coverage and CFR was observed. The RF model indicated that full vaccination and ≥1 dose vaccination coverages were the top two contributing factors to CFR. These findings call on people without complete coronavirus disease 2019 (COVID-19) vaccinations, especially the elderly, to get any type of accessible vaccine, which could save lives.

Since the emergence of Omicron variant, due to its heightened transmissibility, the whole world has been suffering a new wave of the COVID-19 pandemic (1). As finance and trade centers with high international connectivity, Hong Kong SAR, China and Singapore have been experiencing a considerable increase in COVID-19 cases since January 2022. According to the Government of Hong Kong SAR, China, the epidemic was overwhelming the medical system with a relatively high CFR (2), while Singapore seemed to manage its epidemic within the capacity of its medical system contributing to a lower CFR (3). As we know, Hong Kong SAR, China and Singapore share many similar characteristics, such as developed economy, freeport international trade, and a similar population density. However, it is unclear why there is such a stark difference in CFR between the two locations. This study aims to explore this important issue, which could provide significant implications for China to develop and adjust future COVID-19 prevention strategies.

Hong Kong SAR, China had been applying the zero-COVID policy using strict non-pharmaceutical interventions (NPIs) such as border quarantine, social distancing, and mask wearing, and achieved low cases during the first four waves of epidemic. Unfortunately, several infected flight staff did not obey the quarantine rule and entered the community, which was the suspected origin of the Omicron variant epidemic in Hong Kong SAR, China. Because of the superior transmissibility of Omicron variant, the delayed identification, and the control of all infected people, the epidemic rapidly expanded. Together with low vaccination coverage among the elderly, the imperfect tiered diagnosis, and treatment strategy at the early stage, more severe cases may occur and a few patients with mild symptoms may occupy the hospital beds, which lead to overwhelmed medical systems.

Since the emergence of the first case of COVID-19 in January 2020 in Singapore, its containment strategy had been applied until August 2021, when 78% of its population had been vaccinated with at least one dose. Since then, Singapore has continued to promote COVID-19 vaccination and tried to transition to “co-existence with virus” strategy by gradually lifting NPI restrictions. Singapore has also been optimizing the tiered diagnosis and treatment strategy, which effectively relieved the pressures on the medical system.

CFRs, vaccination coverage, infection rate, medical resources occupancy rates for COVID-19, and other comparative variables, including age structure, the number of swab tests, and prevalence of chronic diseases were obtained and calculated from the Government of the Hong Kong SAR, China and Singapore. The detailed source websites were presented in the footnotes of tables and figures.

To investigate the effects of vaccination coverage after controlling for confounding factors, we collected the data of 52 Omicron dominated countries and regions (Omicron proportion ≥90%) from Our World in Data (https://ourworldindata.org/). Nine explanatory variables, including full vaccination, ≥1 dose, booster vaccination, gross domestic product (GDP) per capita, population density, hospital beds per 1,000 people, human development index, proportion of people aged 70 and older, and cardiovascular mortality per 100,000 people were collected. One dependent variable (CFR) was collected. The detailed definitions of these variables are presented in Supplementary Table S1. Due to the 28-day efficacy assessment period (4), the proportion of people with at least one dose, with full vaccination, with booster vaccination before February 7, 2022 were collected and then averaged for analysis.

Multiple linear regression models were employed to explore the effects of vaccination coverage (full vaccination, ≥1 dose, and booster vaccination) on CFR separately, after controlling for potential confounding variables. In addition, a random forest (RF) regression model was used to qualitatively assess the feature importance of nine explanatory variables contributing to CFR. RF regression analysis, a tree-based machine learning algorithm, works well with multicollinearity and nonlinear problems between predictors and makes decisions by constructing a multitude of decision trees (5,000 trees in this study) for training the data included in a model (5–6). Predictions for each variable are aggregated across all trees and percentage increase in mean square error (%IncMSE) is calculated, which indicates how much the model accuracy decreases if that variable were omitted.

All statistical analyses were performed using R software (version 4.0.5, R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value of <0.05 represented statistically significant.

Table 1 shows that from December 31, 2021 to March 22, 2022, the CFR of COVID-19 in Hong Kong SAR, China (0.58%; age standardization: 0.53%) is much larger than in Singapore (0.05%; age standardization: 0.06%). The CFRs increase with the rise of age in both Hong Kong SAR, China and Singapore, and in most age groups, CFRs were higher in Hong Kong SAR, China than in Singapore, especially for people ≥70 years.

As of March 22, 2022, the full vaccination and booster coverage in Hong Kong SAR, China was 76.15% and 34.50%, respectively, which is much lower than in Singapore with 92.00% full vaccination and 71.00% with a booster vaccination. Especially for people aged ≥70 years in Hong Kong SAR, China, the full vaccination rates are far lower (56.82%) than in Singapore (95.00%).

Table 2 presents the CFRs by vaccination status and age groups. The CFRs of cases with full vaccination were far lower than those with partial vaccination, especially for the elderly. In Singapore, the CFR for people ≥70 years ranges 7.20%–16.00% for non-fully vaccinated cases, 0.76%–2.80% for fully vaccinated cases, and 0.08%–0.41% for booster vaccinated cases. The CFRs of fully vaccinated cases were also lower than those without full vaccination in Hong Kong SAR, China.

In order to provide more solid evidence for the effects of vaccination coverage, we further collected data from 52 Omicron variant dominated countries/regions and conducted multiple linear regressions. After adjusting for GDP per capita, proportion of people aged 70 and older, population density, hospital beds per 1,000 people, human development index, and cardiovascular mortality per 100,000 people, we found that a higher coverage of ≥1 dose vaccination, full vaccination, and booster vaccination were all negatively associated with lower CFRs (Supplementary Table S2). Moreover, full vaccination and ≥1 dose vaccination coverages were the top two contributing factors to CFR (Figure 1).

Figure 1. 

The importance rank of contribution variables to CFR using a RF model.

Note: Full vaccination refers to people who received all doses prescribed by the initial vaccination protocol (commonly 2 doses and sometimes 3 doses, depending on the definition by the corresponding countries) per 100 people in the total population of the country or region; ≥1 dose refers to people who received at least 1 dose vaccination per 100 people in the total population of the country or region.

Booster vaccination refers to people with doses administered beyond the number prescribed by the initial vaccination protocol (commonly 3 doses and sometimes 4 doses, depending on the definition by the corresponding countries or regions) per 100 people in the total population of the country or region.

The detailed information of other variables is presented in Supplementary Table S1.

Abbreviation: CFR=case fatality ratio; RF=random forest regression model; %IncMSE=percentage increase in mean square error.