The Effectiveness of a Diverse COVID-19 Vaccine Portfolio and Its Impact on the Persistence of Positivity and Length of Hospital Stays: The Veneto Region’s Experience

The vaccination campaign for the Veneto region (northeastern Italy) started on 27 December 2020. As of early December 2021, 75.1% of the whole Veneto population has been fully vaccinated. Vaccine efficacy has been demonstrated in many clinical trials, but reports on real-world contexts are still necessary. We conducted a retrospective cohort study on 2,233,399 residents in the Veneto region to assess the reduction in the COVID-19 burden, taking different outcomes into consideration. First, we adopted a non-brand-specific approach borrowed from survival analysis to estimate the effectiveness of vaccination against SARS-CoV-2 in preventing infections, hospitalizations, and deaths. We used t-tests and multivariate regressions to examine vaccine impact on breakthrough infections, in terms of the persistence of positivity and the length of hospital stays. Evidence emerging from this study suggests that unvaccinated individuals are significantly more likely to become infected, need hospitalization, and are at a higher risk of death from COVID-19 than those given at least one dose of vaccine. Cox models indicate that the effectiveness of full vaccination is 88% against infection, 94% against hospitalization, and 95% against death. Multivariate regressions suggest that vaccination is significantly correlated with a shorter period of positivity and shorter hospital stays, with each step toward completion of the vaccination cycle coinciding with a reduction of 3.3 days in the persistence of positivity and 2.3 days in the length of hospital stay.


Introduction
The SARS-CoV-2 virus has infected more than 250 million people worldwide, causing over 5 million confirmed deaths [1]. Although COVID-19 carries a higher risk of complications and death for older adults and individuals with comorbidities [2][3][4][5], the involvement of younger sections of the population has also to be taken into account [6,7].
The Veneto region in northern Italy was one of the first areas of Europe to be severely hit by COVID-19 outbreaks [8], and it has reached over 500,000 cases and 11,000 deaths among a population of 4.9 million [9].
Much effort has been made into developing vaccines, with the aim of containing the virus's transmission, the pressure on critical-care facilities, and the risk of death. Veneto's vaccination campaign began on 27 December 2020 [10], initially focusing on health workers,

Materials and Methods
Anonymized data on our study population were retrieved from regional databases managed by Azienda Zero (Padua, Italy), a regional institution whose mission consists of ensuring the rationalization, integration, and efficiency of health, social, technical, and administrative services of regional structures. These databases were compiled through a mandatory reporting system, and they contain general information about all the molecular and antigen SARS-CoV-2 tests and the anti-COVID-19 vaccinations performed by the Veneto region, as well as the monitoring of confirmed SARS-CoV-2 cases taken under the charge of the same region.
The latest available updates were 8 October 2021 for molecular and antigen tests and the monitoring of individuals testing positive, and 1 September 2021 for vaccinations. We considered individuals as infected when they tested positive on either an antigen or molecular nasopharyngeal swabs. More than half (55.2%) of these cases had initially been tested with a molecular swab, while the remainder were first identified by a positive antigen swab. More than 86% of the latter cases were confirmed by a positive PCR test within 48 h. When subsequent molecular swabs were negative, the case was recorded as a "false positive", as suggested by FDA [19]. False-positive tests were then excluded from our analysis.
Our study period extended from 27 December 2020, when the regional vaccine campaign began, and covered 254 days, up to 7 September 2021. For the purposes of our analysis, we only considered individuals who had not tested positive before the starting date of the study and who had at least one swab during the study period. This study population then amounted to 2,233,399 individuals, with 7,019,827 tests (60.1% of which were antigen swabs) taken during the period considered (averaging 3.1 swabs each), and 213,469 individuals testing positive for the virus.
At any time during the period considered, individuals in our study sample could be unvaccinated, partially vaccinated, or fully vaccinated. There is a well-known delay between inoculation and artificial immunization, and many scientific articles and reports take this into account [20][21][22][23][24]. We considered individuals as partially vaccinated as of 14 days after their first dose of vaccine, and fully immunized as of 7 days after their second or as of 14 days after being vaccinated with the mono-dose Ad26.COV2.S.
According to our data, some individuals in our sample who never tested positive for SARS-CoV-2 did not appear to have had their second dose of vaccine within a reasonable timeframe. Since subjects who tested positive before the beginning of the vaccination campaign were excluded, we believe that some of these latter cases are due to individuals who refused or deferred their second dose, while some others may be probably due either to missing data or individuals receiving their second dose outside the Veneto region. We then decided to limit the follow-up of partially immunized individuals to the arbitrary threshold of 150 days to take into account potentially belated second doses but, at the same time, not excessively overestimate the effectiveness of the partial immunization.
When necessary, we drew on official public data from the permanent census of the Veneto population, referring to Italian Statistics Institute (ISTAT) records updated to 2020 [25].
The data were summarized using percentages and incidence rates, or mean values with their 95% confidence intervals (95% C.I.) as appropriate. Kaplan-Meier curves were used to compare the survival probability among different groups, and log-rank tests were used to identify any significant differences between these curves. A multivariate Cox model was computed, adjusting for covariates such as sex and age, estimating vaccine effectiveness VE as where OR stands for the adjusted odds ratio between the risk of an "event" (infection, hospitalization, or death) in immunized and unvaccinated individuals. The persistence of individuals' positivity to the virus was computed as the number of days between their first positive swab and a first negative molecular test performed afterward. Only individuals with such before-and-after swabs were considered when computing the persistence of positivity. Length of hospital stay was computed as the time elapsing between the date of hospitalization and the date of discharge home; patients who died in hospital were excluded from the analysis. When considering the persistence of positivity and length of hospital stay, we compared mean values with t-tests, and we conducted multivariate linear regressions with sex, age, and vaccination status as independent variables, and the persistence of positivity or length of hospital stay as the dependent variable.
A p-value below 0.05 was always considered statistically significant. Data linkage, cleaning, and visualization were performed with specific libraries for data analysis in Python 3.7.0, Kaplan-Meier curves and Cox models were obtained with statistical packages from Python. Linear regressions and t-tests were conducted using Excel 2013 and SPSS 27.0.

Vaccine Coverage
Judging from data collected by the Veneto Regional Authority, the number of people vaccinated with at least one dose before 1 September 2021 amounted to 3,504,221, and 98% of them were residents of the region. While the proportion of the region's population that had received at least one dose of vaccine was 70.4%, only 59.2% had been fully vaccinated ( Figure 1).
where OR stands for the adjusted odds ratio between the risk of an "event" (infection, hospitalization, or death) in immunized and unvaccinated individuals. The persistence of individuals' positivity to the virus was computed as the number of days between their first positive swab and a first negative molecular test performed afterward. Only individuals with such before-and-after swabs were considered when computing the persistence of positivity. Length of hospital stay was computed as the time elapsing between the date of hospitalization and the date of discharge home; patients who died in hospital were excluded from the analysis. When considering the persistence of positivity and length of hospital stay, we compared mean values with t-tests, and we conducted multivariate linear regressions with sex, age, and vaccination status as independent variables, and the persistence of positivity or length of hospital stay as the dependent variable.
A p-value below 0.05 was always considered statistically significant. Data linkage, cleaning, and visualization were performed with specific libraries for data analysis in Python 3.7.0, Kaplan-Meier curves and Cox models were obtained with statistical packages from Python. Linear regressions and t-tests were conducted using Excel 2013 and SPSS 27.0.

Vaccine Coverage
Judging from data collected by the Veneto Regional Authority, the number of people vaccinated with at least one dose before 1 September 2021 amounted to 3,504,221, and 98% of them were residents of the region. While the proportion of the region's population that had received at least one dose of vaccine was 70.4%, only 59.2% had been fully vaccinated ( Figure 1). Several factors contribute to the gap between the numbers of people who were partially vaccinated, as opposed to fully vaccinated. First, anyone given their second dose on schedule after the end of the study period, as well as people considered as fully vaccinated after testing positive for the virus and those receiving one dose of vaccine, would be classified as only partially vaccinated. On the other hand, people who were given a single Several factors contribute to the gap between the numbers of people who were partially vaccinated, as opposed to fully vaccinated. First, anyone given their second dose on schedule after the end of the study period, as well as people considered as fully vaccinated after testing positive for the virus and those receiving one dose of vaccine, would be classified as only partially vaccinated. On the other hand, people who were given a single dose of the Ad26.COV2.S vaccine were classified as fully vaccinated. Approximately 51% Vaccines 2022, 10, 107 5 of 15 of Veneto residents who had been at least partially vaccinated are female, a proportion consistent with the M/F ratio in the region's general population [25].
Vaccine coverage and vaccine brand distributions differed across different age groups (Table 1). Among those vaccinated with the first dose of BNT162b2, 82.6% had completed their vaccination cycle, almost all of them with a second dose of the same vaccine. Among those inoculated with mRNA-1273 first, 68.1% had received a second dose, almost always (99.8%) with a second dose of the same vaccine. Among those given the first dose of ChAdOx1, only 57.9% had completed their vaccination cycle, with a second dose of the same vaccine in 92.5% of cases. Thus, 0.7% of all the complete vaccination cycles were heterologous, and these cycles mainly involved younger people given the first dose of ChAdOx1, for whom heterologous vaccination became mandatory after 11 June 2021 [26].

Vaccine Effectiveness against Infection
As summarized in Table 2, unvaccinated people had both a higher positivity rate and a higher risk of infection than those who were partially or fully vaccinated. The fully vaccinated were also considerably better protected against infection than the partially vaccinated. Kaplan-Meier curves and log-rank tests confirmed these findings ( Figure 2): at all times, any form of immunization was significantly preferable to no vaccination, and, where comparable, survival probability was significantly lower for the partially vaccinated than for the fully vaccinated population. Kaplan-Meier curves and log-rank tests confirmed these findings ( Figure 2): at a times, any form of immunization was significantly preferable to no vaccination, and where comparable, survival probability was significantly lower for the partially vac cinated than for the fully vaccinated population. The cumulative probability of not contracting the virus decreased at different rate over time in the three subgroups considered, and by 5 months after the second dose, th effectiveness of a full vaccination cycle affords much the same protection as the first dose In a Cox multivariate model, vaccine effectiveness against contracting the infection after adjusting for the effect of sex and age, was 82% (82-83%) for partial vaccination an 88% (87-88%) for full vaccination. The cumulative probability of not contracting the virus decreased at different rates over time in the three subgroups considered, and by 5 months after the second dose, the effectiveness of a full vaccination cycle affords much the same protection as the first dose.
In a Cox multivariate model, vaccine effectiveness against contracting the infection, after adjusting for the effect of sex and age, was 82% (82-83%) for partial vaccination and 88% (87-88%) for full vaccination. Table 3 shows the incidence of hospitalizations and deaths for the three subgroups, which were both lower the higher the level of vaccination. According to the Cox multivariate model, the vaccine effectiveness against hospitalization, after adjusting for sex and age, was 82% (81-84%) for partial vaccination and 94% (94-95%) for full vaccination.

Vaccine Effectiveness against Hospitalization and Death
We were not able to distinguish the particular type of hospitalization (intensive care, semi-intensive care, etc.), but for some individuals, the initial clinical status was recorded: we should underline that more than 95% (N = 1405) of those for whom the clinical status is registered as "critical" (N = 1476)-to which those admitted in ICU most likely belong-are unvaccinated.
Vaccine effectiveness against COVID-19-related deaths, after adjusting for sex and age, was 54% (48-60%) for partial vaccination and 95% (94-96%) for full vaccination. Figures 3 and 4 show the Kaplan-Meier survival curves representing the probability of not being hospitalized and surviving to COVID-19 among unvaccinated, partially vaccinated, and fully vaccinated people: for both outcomes, log-rank tests confirmed that each of the three probability distributions differed significantly from the others.

Impact of Vaccination on the Persistence of Positivity and Length of Hospital Stay
Impact of Vaccination on the Persistence of Positivity As shown in Figure 5, among the people who contracted the virus, those who had been vaccinated remained positive for a shorter period, on average, than those who had not, which means that the former were also likely to recover more quickly than the latter.

Impact of Vaccination on the Persistence of Positivity and Length of Hospital Stay
Impact of Vaccination on the Persistence of Positivity As shown in Figure 5, among the people who contracted the virus, those who had been vaccinated remained positive for a shorter period, on average, than those who had not, which means that the former were also likely to recover more quickly than the latter. On average, positivity seemed to persist for longer the older the individual affected ( Figure 6). Our findings also suggest that the time to the first negative molecular swab became shorter with each step toward completion of the vaccination cycle. These results are presented in Table 4, by vaccination status and age group. The results of t-tests showed that the unvaccinated, partially vaccinated, and fully vaccinated subpopulations varied considerably in terms of mean persistence of positivity, with all p-values lower than 0.001.  On average, positivity seemed to persist for longer the older the individual affected ( Figure 6). Our findings also suggest that the time to the first negative molecular swab became shorter with each step toward completion of the vaccination cycle. These results are presented in Table 4, by vaccination status and age group. The results of t-tests showed that the unvaccinated, partially vaccinated, and fully vaccinated subpopulations varied considerably in terms of mean persistence of positivity, with all p-values lower than 0.001.

Impact of Vaccination on the Persistence of Positivity
As shown in Figure 5, among the people who contracted the virus, those who ha been vaccinated remained positive for a shorter period, on average, than those who ha not, which means that the former were also likely to recover more quickly than the latte On average, positivity seemed to persist for longer the older the individual affecte ( Figure 6). Our findings also suggest that the time to the first negative molecular swa became shorter with each step toward completion of the vaccination cycle. These result are presented in Table 4, by vaccination status and age group. The results of t-tests showe that the unvaccinated, partially vaccinated, and fully vaccinated subpopulations varie considerably in terms of mean persistence of positivity, with all p-values lower than 0.001   The length of hospital stays also correlated strongly with patients' age ( Figure 7).  The length of hospital stays also correlated strongly with patients' age ( Figure 7).

Figure 7.
Mean length of hospital stay with respect to patients' age. Linear trend. Table 5 shows the mean length of hospital stays for COVID-19 by vaccination status and age. The results of t-tests revealed no significant differences between the unvaccinated and the partially or fully vaccinated groups (p = 0.92 and p = 0.11, respectively), whereas fully vaccinated patients' hospital stays were significantly shorter than those of unvaccinated patients (p = 0.01). Table 5. Mean length of hospital stay (days) by age groups and vaccination status.  Table 5 shows the mean length of hospital stays for COVID-19 by vaccination status and age. The results of t-tests revealed no significant differences between the unvaccinated and the partially or fully vaccinated groups (p = 0.92 and p = 0.11, respectively), whereas fully vaccinated patients' hospital stays were significantly shorter than those of unvaccinated patients (p = 0.01).   Table 6 shows the results of multivariate linear regressions were run with the persistence of positivity or length of hospital stay as the dependent variable, and sex, age, and vaccination status as independent covariates. While sex was not significantly associated with the persistence of positivity, this variable showed a linear positive correlation with age and a linear negative correlation with vaccination status. The length of hospital stay showed a significant association with sex, age, and vaccination status: females tended to have shorter hospital stays, older people tended to remain in the hospital for longer, and vaccination status again showed a linear negative association with this variable.

Discussion
The Veneto region has already reached a good level of vaccination coverage, which will hopefully continue to grow. It is not homogeneous among different age groups because older people were given priority at the start of the vaccination campaign. That said, our findings show that vaccinations are clearly mitigating the effects of COVID-19 regardless of the age group considered. Vaccination has produced favorable results in both clinical and public health terms.
First, we confirmed previous evidence to indicate that SARS-CoV-2 infection poses less of a threat if people are vaccinated. We also found that, even when artificial immunization fails and vaccinated people become infected, they are less prone to serious COVID-19 complications such as hospitalization and death. The risk of both SARS-CoV-2 infection and COVID-19 complications was also found significantly lower in the fully vaccinated than in the partially vaccinated group, confirming the need for a complete vaccination cycle whenever possible. Concerning the risk of infection, Italian Government health policies adopted from 6 August 2021 onwards established that anyone unvaccinated or who had received the first dose of vaccine less than 14 days earlier was obliged to undergo molecular or antigen testing for the virus, in order to access public places (and a negative swab was considered valid for 48 h). This would mean that relatively few positive cases were likely to emerge from the test frequently taken by such people. This could lead to the positivity rate in this subpopulation being underestimated, with a consequent potential underestimation of vaccine effectiveness against infection as well.
It can be inferred that vaccine effectiveness against infection, hospitalization, and death due to COVID-19 decreases over time, especially beyond 150 days after the second inoculation. This is consistent with growing evidence of the efficacy of COVID-19 vaccines waning with time [27]. Given recurrent waves of the pandemic, it is, therefore, fundamentally important to delivering a booster dose of vaccine to those most in need, who were also the people vaccinated first when the vaccination campaign began, so their protection will have faded the most. Some evidence about the association between booster dose and an increase in the IgG titers have already been made [28], and, in our opinion, further research on the efficacy of third doses and on the optimal timing of their administration is now needed.
Our findings suggest a significant reduction in the persistence of positivity in people who have been artificially immunized. The overall average reduction in our sample amounted to 5 days, and according to multivariate regression, each step toward the completion of a vaccination cycle coincide with a reduction of 3.3 days. This aspect is important for two reasons. Reducing the amount of time an infected person remains positive means lowering the overall social costs (less morbidity, fewer school days lost, etc.). It also has a direct influence on the likelihood of a person transmitting the virus to others: Assuming the same social interactions, the longer they remain positive, the greater their chances of infecting someone else. It has also been demonstrated that a shorter period of positivity correlates closely with a lower viral load, meaning that a more persistent positivity could coincide with greater infectivity [29].
Another contribution of our work is the finding that length of hospital stay is also influenced by vaccination status, as it was significantly shorter in our immunized population. The average difference between unvaccinated and fully vaccinated patients' hospital stays in our sample as a whole was only 2 days, but among people aged 60 or more (the age group most likely to be hospitalized with COVID-19), it ranges from 4 to 5 days. In a multivariate model, each step toward completion of the vaccination cycle coincided with a hospital stay 2.3 days shorter. This is another fundamental issue, as the length of hospital stays strongly influences the burden on healthcare systems. Shorter hospital stays reduce the economic and human resources needed to cope with COVID-19, enabling a faster return to the hospitals' ordinary activities so heavily affected during the pandemic. One of the aspects to which health authorities have been paying particular attention in assessing the impact of the pandemic (and introducing restrictive measures) is ICU occupancy and ordinary hospitalization rates. Clearly, if further studies confirm that vaccinated people becoming infected and ill enough to need admitting to hospital have shorted stays, then a population with a better vaccination coverage would take longer to overstretch the inpatients care services, and the bar for introducing restrictions on the public's freedom of movement could be raised.
We are well aware that our work is not without its limitations. For instance, the study was based entirely on information obtained from official regional databases, which meant that the analysis had to be kept at a rather general level. We could not adjust or stratify our analysis on the strength of more specific information, such as the clinical picture or particular risk of individuals. We could not distinguish between types of hospitalization (in intensive care, semi-intensive care, etc.). On other hand, the use of such databases enabled us to conduct statistically significant and robust analyses on large volumes of records. It also gave us a reliable snapshot of the overall impact of vaccination against COVID-19 on the population of the Veneto region as a whole.
It is important to keep in mind that all our findings are influenced by the demographic structure and the vaccine coverage of the Veneto region, as well as by specific viral factors such as the predominant SARS-CoV-2 mutation among our study population. During the beginning of our study period, the distribution of variant of concerns (VOCs) among confirmed SARS-CoV-2 cases in the Veneto region has been quite variable; for instance, in January 2021 at least six different VOCs were circulating, even if the prevailing ones were B.1.177-rel and Alpha (B.1.1.7) e sublineages (Q.x). However, from April 2021, Delta (B.1.617.2) e sublineages (AY.x) started spreading, and from August 2021, almost all the confirmed cases were due to this latter VOC [30]. In the last period, we are witnessing the extremely rapid spreading of the Omicron (B.1.1.529) VOC: since viral variant factors, such as the antigenic mismatch with vaccines or increased transmissibility could affect vaccine effectiveness [31], further studies based on the latest available data are needed.

Conclusions
In conclusion, part of our work confirms the content of previous reports with a specific focus on experience gained in the Veneto region. Other findings, such as the significant correlations for vaccination status with the persistence of positivity and length of hospital stay, are new, as these issues had yet to be thoroughly investigated. Our work could hopefully serve as a starting point for further research on these variables, which we consider far from secondary aspects of the beneficial consequences of the vaccination campaign. Funding: This study was supported by the Veneto Regional Authority (DGR 1643 del 24/11/2020. "Project to assess the prevalence and spread of SARS-CoV-2 in the population" in: Bollettino Ufficiale Regione Veneto n. 193 del 15/12/2020. Available online: https://bur.regione.veneto.it/BurvServices/ pubblica/DettaglioDgr.aspx?id=435266 accessed on 5 December 2021).

Institutional Review Board Statement:
According to the Italian national guidelines (DM 18/03/1998), anonymized data may be analyzed and used in aggregate form for scientific studies without further authorization, meaning that no formal ethics committee approval was needed for the present study. This study complies with the requirements of the latest version of the Declaration of Helsinki. All patients gave their consent, and all the data were anonymized before the analysis. The study was requested by the Veneto Regional Authority (DGR 1643 del 24/11/2020). The data were treated with full respect for confidentiality, in accordance with Italian legislation.

Informed Consent Statement:
The need for patients' consent was waived due to the anonymous nature of the data, and the fact that its recording is mandatory. The data were treated with full respect for confidentiality, in accordance with Italian legislation. Before the database was made available to the authors, all sensitive data concerning the patients considered in the study were replaced with anonymous codes, making it impossible to identify the individuals concerned.

Data Availability Statement:
The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.