COVID‐19 vaccine effectiveness against hospitalization due to SARS‐CoV‐2: A test‐negative design study based on Severe Acute Respiratory Infection (SARI) sentinel surveillance in Spain

Abstract Background With the emergence of SARS‐CoV‐2, influenza surveillance systems in Spain were transformed into a new syndromic sentinel surveillance system. The Acute Respiratory Infection Surveillance System (SiVIRA in Spanish) is based on a sentinel network for acute respiratory infection (ARI) surveillance in primary care and a network of sentinel hospitals for severe ARI (SARI) surveillance in hospitals. Methods Using a test‐negative design and data from SARI admissions notified to SiVIRA between January 1 and October 3, 2021, we estimated COVID‐19 vaccine effectiveness (VE) against hospitalization, by age group, vaccine type, time since vaccination, and SARS‐CoV‐2 variant. Results VE was 89% (95% CI: 83–93) against COVID‐19 hospitalization overall in persons aged 20 years and older. VE was higher for mRNA vaccines, and lower for those aged 80 years and older, with a decrease in protection beyond 3 months of completing vaccination, and a further decrease after 5 months. We found no differences between periods with circulation of Alpha or Delta SARS‐CoV‐2 variants, although variant‐specific VE was slightly higher against Alpha. Conclusions The SiVIRA sentinel hospital surveillance network in Spain was able to describe clinical and epidemiological characteristics of SARI hospitalizations and provide estimates of COVID‐19 VE in the population under surveillance. Our estimates add to evidence of high effectiveness of mRNA vaccines against severe COVID‐19 and waning of protection with time since vaccination in those aged 80 or older. No substantial differences were observed between SARS‐CoV‐2 variants (Alpha vs. Delta).


| INTRODUCTION
The emergence of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in late 2019 and the following coronavirus disease  pandemic had a great impact on influenza surveillance systems. 1 In Spain, influenza surveillance before the COVID-19 pandemic was based on a sentinel network of primary care physicians as well as a hospital network. 2 When SARS-CoV-2 emerged in 2020, the hospital-based system, which involved the reporting of confirmed influenza cases, was unable to detect early COVID-19 hospitalizations. In addition, the creation of SARS-CoV-2 testing centers outside of the usual primary care circuits and the reallocation of sentinel physicians to other centers led to the disruption of the sentinel primary care influenza surveillance network for the first time since it was established in 1996. 3 The World Health Organization (WHO) and the European Centre  19, with previous experience in influenza VE studies as well, was an essential guide for other hospitals and regions in the design and subsequent implementation of SARI surveillance in Spain. ECDC has also supported the creation of a European SARI surveillance network (E-SARI-NET) and multicountry COVID-19 VE studies in Europe. The first VE estimates against SARI associated with laboratory-confirmed SARS-CoV-2 were published in October 2021. 6 SARI surveillance was successfully implemented in 9 of 19 Spanish regions, with a total of 13 sentinel hospitals included in the national network during the 2020/21 SiVIRA pilot season. 7 COVID-19 vaccination in Spain began on December 27, 2020, initially prioritizing long-term care facilities and health care workers and progressively extending to the general population. By October 3, 2021, 77.6% of the total Spanish population was fully vaccinated. 8 Monitoring the real-world effectiveness of COVID-19 vaccines is essential to guide public health action and decision making, and even more so in a dynamic pandemic context with emerging new challenges such as new SARS-CoV-2 variants, or waning immunity.
With case-based data on a representative sample of SARI admissions, the SiVIRA surveillance system constitutes an appropriate platform for responding to these emerging challenges and for timely measuring of VE against severe forms of COVID-19.
The aim of this study was to provide COVID-19 VE estimates against COVID-19 hospitalization, by age group, type of vaccine, time since vaccination, and SARS-CoV-2 variant, using a test-negative design. As a data source, we used the information obtained from the SARI sentinel surveillance during season 2020-2021, the first season in which SiVIRA was implemented in Spain.

| Study design
A total of 13 sentinel hospitals from nine Spanish regions participated in the SARI surveillance. As described in the surveillance protocol, 9 case-based data were collected for a systematic weekly sample of patients who were hospitalized on Tuesday and/or Wednesday, depending on the Spanish region, meeting the SARI case definition (Appendix 1 in the Supporting Information). Patients were swabbed for RT-PCR or rapid antigen test for SARS-CoV-2, and demographic, clinical, and COVID-19 vaccination data were collected from hospital records and vaccine registries. We used a test-negative case-control study design to estimate COVID-19 VE. 10 Where feasible, SARS-CoV-2 viruses from RT-PCR positive cases were sequenced, and phylogenetic analysis was performed to identify the SARS-CoV-2 variant and lineage. If available, sequencing results were linked with epidemiological and clinical data merging through a unique sample identifier.

| Study period
The study period included data reported between Weeks 53/2020 and 39/2021, including SARI patients with swab dates between January 1, 2021, and October 3, 2021.

| Study inclusion criteria
We included patients aged 20 years and older who were part of an age-specific target group for vaccination at time of swab (Table S1), with positive or negative SARS-CoV-2 results and known COVID-19 vaccination status. We excluded those who were swabbed more than 10 days (RT-PCR tests) or 5 days (rapid antigen tests) after symptom onset. We excluded patients vaccinated with first dose on or after onset date and those who did not meet the complete vaccination schedule in terms of number of doses and delay between doses. We further excluded those with symptom onset within 1-13 days of latest dose of COVID-19 vaccine ( Figure 1).

| Case definitions
Cases and controls were defined as SARI patients testing positive or negative for SARS-CoV-2 in an RT-PCR or rapid test, in the first 10 or 5 days, respectively, since symptom onset. There were no influenza positive cases among SARS-CoV-2 negative controls.
In the variant-specific analyses, a case was defined as RT-PCR positive with SARS-CoV-2 Alpha (B.1.1.7 lineage) or Delta variants (B.1.617.2 or AY lineages) confirmed through next generation sequencing (NGS). We only used negative controls from weeks with sequenced cases, removing all controls from weeks before and after the first and last case. We also removed controls from hospitals with no reported sequencing information.
We classified eligible cases and controls as either completely vaccinated or unvaccinated, dropping those with partial vaccination schedules. Definitions used for complete vaccination can be found on Appendix 1 of Supporting Information.

| Statistical analysis
We compared the odds of complete COVID-19 vaccination between cases and controls using a logistic regression, and VE was estimated as 1-OR. We adjusted for age, sex, and presence of at least one chronic condition (hypertension, heart disease, chronic respiratory disease, diabetes, liver disease, renal disease, immunodeficiency or other chronic conditions). Age was modeled as restricted cubic splines (RCS), and swab date was modeled as RCS or month of swab, depending on the analysis. For the age-specific analyses, we stratified the data into the following age groups: 20-39, 40-59, 60-69, 70-79, and ≥80 years. For some analyses, we stratified age into wider groups to increase sample size.
We measured VE, overall, and for mRNA vaccines, by time between vaccination and onset of symptoms with cut-off points

| Informed consent
All data used for this study were collected as part of routine surveillance, and informed consent or official ethical approval was not Although individual informed consent was not required, all data were pseudoanonymised to protect patient privacy and confidentiality.

| Characteristics of cases and controls
We included 1772 SARI patients aged 20 and older, of which 1104 were positive to SARS-CoV-2 (cases) and 668 were negative (controls) ( Figure 1). Among the cases and controls, 770 (43%) had received complete COVID-19 vaccination at least 14 days before symptom onset ( Figure 2).
More than 75% of controls and 58% of cases were aged 70 and older, and the median age was 81 for controls and 71 for cases. The prevalence of underlying conditions was higher among controls than cases, with significant differences for all chronic conditions, except for chronic liver disease. Clinical presentation was more severe among cases than controls: 85% cases versus 51% controls had pneumonia, and 8% cases versus 1% controls required mechanical ventilation support. Cases had significantly higher proportion of ICU admission (10% vs. 2%) and death (18% vs. 10%) than controls (Table 1).
PCR was the most commonly used diagnostic test, although rapid antigen test was also used for case confirmation (25%). Rapid tests were less frequently used among controls (3%), in line with the protocol recommendation of a PCR test for confirmation if the initial rapid antigen test was negative.  Figure 3B).
Compared with mRNA vaccines, Janssen VE was lower among those aged 20 to 59 years, and Vaxzevria and Janssen VE were lower among those aged 60 years, although precision is low in these analyses due to the small sample size (Table S4).

| Characteristics of Alpha and Delta hospitalizations
Among cases with sequencing data notified through SARI sentinel surveillance, 35 were Alpha cases, swabbed between Weeks 7 and 32, and 71 were Delta cases, swabbed between Weeks 27 and 39.
Severe outcomes like pneumonia, mechanical ventilation, ICU admission, and death were more frequent among Alpha than Delta cases, and no differences were observed in underlying chronic conditions (Table 3). We found no differences between periods with circulation of Alpha or Delta SARS-CoV-2 variants, although variant-specific VE was slightly higher against Alpha.

| VE against Alpha and Delta
The overall VE estimate is lower than the one estimated soon after the implementation of the vaccination program in Israel, [11][12][13] Canada, 14

| CONCLUSION
In summary, surveillance data from the first season of the SiVIRA hospital network demonstrate the usefulness of sentinel syndromic surveillance systems to describe clinical and epidemiological characteristics of SARI hospitalizations and to monitor the circulation of SARS-CoV-2, influenza, and other respiratory viruses, while also providing data to measure the effectiveness of vaccination in the population under surveillance. Our study adds to the evidence of waning of protection against severe COVID-19 with time since vaccination in those 80 years or older, but with no substantial differences between SARS-CoV-2 variants (Alpha or Delta). In addition, this study provides more data on the higher effectiveness of mRNA vaccines compared with Janssen or Vaxzevria. Our results endorse the policy, already approved in Spain, 43 of administering additional doses, particularly in the population over 80.

ACKNOWLEDGMENTS
We thank all the participants in the SARI sentinel surveillance system, including everyone involved in data collection and