Impacts of a delayed and slow-paced vaccination on cases and deaths during the COVID-19 pandemic: a modelling study


 Background: In Brazil, vaccination has always been cutting across party political and ideological lines, which have delayed its start and brought the whole process into disrepute. Such divergences put the immunisation of the population in the background and create additional hurdles beyond the pandemic, mistrust and scepticism over vaccines.Methods: We conduct a mathematical modelling study to analyse the impacts of late vaccination and with slowly increasing coverage, as well as how harmful it would be if part of the population refused to get vaccinated or missed the second dose. We analyse data from confirmed cases, deaths caused by COVID-19, and vaccination in the state of Rio de Janeiro in the period between March 10, 2020, and October 27, 2021. The classical SIR model is extended to consider the effect of vaccination (efficacy, interval between doses, and vaccination rate) and data sets are regularised using Gaussian Process Regression. The model parameter distributions are estimated using Bayesian inference, aiming to obtain credible intervals in the simulations.Findings: We estimate that if the start of vaccination had been 30 days earlier, combined with efforts to drive vaccination rates up, 31,657 (25,801–35,117) deaths could have been averted. Our results also indicate that the slow pace of vaccination and the low demand for the second dose could cause a resurgence of cases as early as 2022.Interpretation: The government's inaction and lack of a strategic plan to fight the pandemic meant that vaccination started late, leading to thousands of deaths that could have been prevented. Even when reaching the expected vaccination coverage for the first dose, it is still challenging to increase adherence to the second dose and maintain a high vaccination rate to avoid new outbreaks.Funding: Carlos Chagas Filho Foundation for Supporting Research in the State of Rio de Janeiro (FAPERJ) and Brazilian National Council for Scientific and Technological Development (CNPq).


Research in context
Evidence before this study Although vaccination rates against COVID-19 have increased over time, part of the population still harbours suspicions about the vaccine's benefits. Misinformation, shaped to a great extent by political and religious ideologies, makes many people unwilling to get vaccinated. Lack of effort and strategic planning caused delays in the availability of vaccines, also making the vaccination process often lethargic. Furthermore, many people have also neglected the second dose of vaccines, not achieving the expected protection. As these are global concerns, there is an immediate need to obtain answers on how these factors affect the mitigation of COVID-19. In particular, all these factors have been affecting the progress of vaccination against COVID-19 in Brazil. We searched PubMed up to October 27, 2021, with no language restrictions, using the search term ª((avertable deaths) OR (delayed start of vaccination) OR (low vaccination coverage) OR (resurgence of cases)) AND (COVID-19) AND (vaccines).º We found only nine studies that also analysed some impacts of delayed or slow-paced vaccination on the COVID-19 pandemic, inferring the number of possible cases and deaths resulting from low vaccination coverage. However, to our knowledge, no modelling analysis has sought to comprehensively study all of these factors and how they affect the number of infected and dead individuals in a population.

Added value of this study
This study provides a quantitative insight of an intuitive perception of global concern, the damage caused by a vaccination campaign that is sometimes negligent and inadequate, in the face of a pandemic. Since we consider achievable scenarios, from the point of view of the start date of vaccination and the efforts to increase vaccination rates and achieve greater vaccination coverage (based on the evidence before this study), the comparison with the factual situation expresses how such efforts could have mitigated the consequences of the pandemic. Estimating such consequences in terms of the number of dead and infected individuals is a pragmatic line of reasoning for analysing how bad decisions can affect the population as a whole. The added value of this study is also related to the computational framework proposed to conduct the analyses, which can be used in supplementary studies.

Implications of all the available evidence
An immediate implication of this study is to give support for driving the prioritisation of measures to enhance COVID-19 vaccination campaigns around the world, given the impacts that were quantified in this analysis. Furthermore, the eventual emergence of possibly more lethal variants, the antagonism between limited access to vaccines and their acceptance, and seasonality protection against the virus demand continuous immunisation supply chain improvement. in starting the vaccination campaign, compared to other coun- 9 tries, 1 which took place on January 17, 2021. 10 Although Brazilians' tendency towards vaccination compliance 11 is relatively high, 2 some factors were partly responsible for the 12 slowness of the mass vaccination campaign. The country is 13 paying a price for the slow pursuit of vaccines early on, espe-14 cially regarding the federal government's rejection of vaccines 15 from Pfizer in mid-2020, 3 in addition to the rebuke of the agree-16 ment signed with Sinovac; 4 millions of people are also missing 17 their second dose, especially because of misinformation, as- 18 suming that just one dose provides the expected immunity; 5,6 19 temporary interruptions of vaccination services, due to a lack 20 of shots, logistical problems or absence of supplies (particularly 21 active pharmaceutical ingredient); 7,8 furthermore, there are on 22 the one hand people who try to jump the queue to get vacci- 23 nated early, and on the other hand those who choose not to get 24 vaccinated, seemingly motivated by political ideology. 9 25 The epidemiological situation in some states is particularly 26 worrisome due to the level of government intervention, invest-27 ments in health, the pace of vaccination, and population mo-28 bility. 10 Political polarisation and the spread of fake news also 29 hamper the fight against COVID- 19 (1 − η). The portion of indi-69 viduals who do not receive the second dose is equal to α and we 70 assume an overall impaired efficacy, given by Å η. In general, the 71 model admits n classes of vaccinated individuals, depending on 72 the types of vaccines used. The schematic representation of the 73 model is shown in figure 1.

75
Daily data on confirmed cases and dead individuals due to 76 COVID-19 in RJ are divided into two subsets, from before and 77 during vaccination. The former, considered the training set, 78 ranges between March 10, 2020, the first day with at least five 79 cases diagnosed, and January 19, 2021, the last day before the 80 start of vaccination. The latter subset contains the data between 81 January 20, the day the vaccination started, and October 27, 82 2021. Cumulative data on individuals vaccinated with the first 83 dose and immunised (with both doses and with the single-dose 84 vaccine) are also adopted from the same period. These data 85 are provided by the Ministry of Health public repository 12 13 In these circum-92 stances, the regularisation of data emerges as an alternative to 93 reduce the noise level, without misrepresenting data behaviour, 94 to streamline the task of fitting model responses to the data set. 95 In particular, Gaussian Process models are a probabilistic ap-96 proach to representing arbitrary functions by means of a prob-97 ability distribution over all possible functions that fit a set of 98 points. 14  We consider that the target of individuals to be immunised in RJ 116 is proportional to 80%, which also corresponds to the number 117 of inhabitants aged 12 years or over. 16 The vaccination process were not used to estimate the model parameters.  We now simulate the model considering only 70% vaccination 230 coverage, a reduced amount due to people who are unwilling 231 to be vaccinated. Figure 4A shows

283
More ambitious vaccination targets and avertable deaths 284 We attempt to infer deaths that could have been averted simply 285 by having vaccination started days earlier or if the daily rate 286 of vaccination had been higher. Figure 4B shows the relation-287 ship between vaccinated individuals and cumulative deaths over 288 time. We simulate the model using the benchmark vaccination 289 rate (ν = 0·40%) and compare the outcomes in the context of 290 a faster vaccination (ν = 0·50%), making allowance for differ-291 ent days for the start of vaccination from the day it actually 292 started. Simulations show that presumably not-so-challenging 293 measures, such as having anticipated the vaccination campaign 294 roll-out by just ten days, combined with an average vaccina-295 tion rate approximately 25% faster, could have averted 15,129 296 deaths (12,029±16,986) in relation to the actual scenario; from 297 a more optimistic, yet still realistic, perspective on the vaccina-298  Such delays could also affect the prevalence of the disease, 316 causing the incidence of cases to be longer-lasting at high lev-317 els and, consequently, causing the peak of deaths to be shifted 318 Brazilian population in such measures has always been be-351 low expectations. 22 A very relevant fact is that only 45·5% of 352 Brazilians say they wear a face mask outside the home. 23 Our 353 findings show that the possibility of an eventual resurgence of 354 cases in 2022 should not be overlooked, even though most of 355 the population has been vaccinated. This concern even brings 356 up discussions about possible loss of immunity and the need 357 for extra doses, 24 although vaccines may remain limited, es-358 pecially in low-incoming countries, 25,26 making NPIs essential 359 even after achieving adequate vaccine coverage.

360
The analysed scenarios reflect current knowledge about vacci-361 nation in RJ, from the perspective of available data. The per-362 sistence of such predictions depends to some extent on the con-363 firmation of the hypotheses put forward. Particularly regarding 364 vaccine hesitancy (whether for both doses or just the second), 365 the inaction of certain people depends a lot on facts that can-366 not be predicted. ple. 28,29 In Brazil, the oscillations regarding the feelings anal-

Declaration of interests
The authors declare no competing interests.

Data sharing
Code to replicate analysis and figures supporting the findings of the manuscript are available via the project GitHub repository at https://github.com/gustavolibotte/vaccines-COVID-19. The code is licensed under the MIT license. Source data are provided with this paper and all data used in this study can be downloaded from the cited sources. Figure 1 Schematic representation of the model.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supplementarymaterial.pdf