key: cord-0740006-dy9a6x0e
authors: Isitt, Catherine; Sjöholm, Daniel; Hergens, Maria-Pia; Granath, Fredrik; Nauclér, Pontus
title: The early impact of vaccination against SARS-CoV-2 in Region Stockholm, Sweden
date: 2022-03-29
journal: Vaccine
DOI: 10.1016/j.vaccine.2022.03.061
sha: 96c4c575eedcf33632ef852d223c0ea56b0014fc
doc_id: 740006
cord_uid: dy9a6x0e

Vaccination against SARS-CoV-2 started in Region Stockholm, Sweden in December 2020 with those in long-term care facilities or receiving home care vaccinated first followed by those aged over 80 years. In this population-based, retrospective cohort study, we performed a Poisson regression to model the expected incidence of infections and deaths which we compared to the observed incidence and compared this to an unvaccinated control group of those aged 18-79 years. The aim of this study was to measure the early impact of the vaccination programme in Region Stockholm. Infections and deaths reduced substantially amongst the first two groups targeted for SARS-CoV-2 vaccination with an estimated total 3112 infections prevented, and 854 deaths prevented in these two groups from 4 weeks after the introduction of vaccination through to 2(nd) May 2021.

The Swedish COVID-19 vaccination programme prioritises those most at risk of severe disease from SARS-CoV-2 infection first with the aim of protecting those most vulnerable and safeguarding the healthcare system. The programme is divided into four phases: 1) adults living in long-term care facilities (LTCF), adults receiving home care and their household contacts, and healthcare staff working with this population, 2) adults over the age of 65 years as well as adults of any age who receive dialysis or are transplant recipients, with the oldest invited for vaccination first, 3) adults with other risk factors, 4) adults without risk factors over 18 years of age [1] . In Stockholm, vaccination of Phase 1 began on 27 th December and vaccination of Phase 2 started on 8 th March 2021 with adults >80 years invited for vaccination first [2] . Adults >75 years were invited starting on 29 th March 2021 [2, 3] .

To assess the early impact of the vaccination programme in Stockholm, we performed a population-based, retrospective cohort study to investigate the incidence of PCR-confirmed SARS-CoV-2 as well as deaths within 30 days of a PCR-confirmed SARS-CoV-2 diagnosis, in the cohorts targeted for vaccination compared to a control group not targeted for vaccination.

The Stockholm COVID-19 cohort study collects data from all residents living in Region Stockholm from 2015 onwards. We extracted data from the VAL database which collates information from more than ten other healthcare related databases in Region Stockholm and which has been previously described [4] . Age, and coding for resident in LTCF and home care were used to define the two groups first targeted for vaccination, those in LTCF/home care and those aged ≥80 years, plus a third group, those aged 18-79 years who were used as a composite unvaccinated control. Linkage to SmiNet, the Public Health Agency of Sweden notifiable infection reporting tool, was performed using unique personal identification numbers that each resident in Sweden has, to extract data on SARS-CoV-2 PCR-positive cases.

The study was approved by the Swedish Ethical Review Authority.

Analyses were restricted to 31 st August 2020 to 2 nd May 2021 in order to encompass the second and third waves of COVID-19 in Stockholm. Data from the first wave were not used given the lack of widely available testing at the start of the pandemic. Three mutually exclusive groups were defined, those in LTCF or receiving home care, those aged 80+ years and a composite control of those aged 18-79 years. The time point set for post-vaccination was 4 weeks after the start of vaccination, to account for the fact that the protective effect of a single dose of vaccine is not immediate and that not all individuals in the group will be vaccinated in the first week. Time up until 4 weeks after the vaccination campaign started was counted as pre-vaccination period. The pre-vaccination incidence rate ratio (IRR) between the three groups were estimated by Poisson regression based on the weekly incidences of cases (and deaths). Due to over dispersion the confidence intervals for the estimated IRRs were adjusted by a standard error inflation factor. The pre-vaccination IRRs (in relation to the composite control group) were used to prognosticate the post-vaccination incidence of infections (and deaths) in absence of vaccination, by multiplying the pre-vaccination IRRs with the observed post-vaccination incidence in the composite control group, using the confidence intervals of the pre-vaccination IRR to show uncertainty in the resulting estimate. Differences between the prognosticated and observed number of events were used to estimate the number of infections and deaths that may have been prevented up to 2 nd May 2021 due to the vaccination programme. Aggregated data on vaccine coverage was grouped by age and residence in LTCF (home care status was not available).

Data was analysed using R version 4.1.0. 

The highest number of infections (182,695) were in the control group (18-79 year-olds) ( Table   2 ). More infections were recorded in the LTCF/home care group (6769 total; 5232 pre- Figure 1A ). Weekly observed versus expected incidences of infection per 100,000 days at risk are presented in Figure 1A .

The estimated number of infections prevented by vaccination was 2873 (2441-3337) for the LTCF/home care group and 239 (178-306) for the 80+ group (Table 2) . Figure 1B ). Weekly observed versus expected incidence for deaths are presented in Figure 1B . 

Our data suggests that vaccination of the LTCF/home care and 80+ groups succeeded in preventing a third wave in these groups of the same magnitude as the control group. Since the start of the vaccination programme there has been no national lock-down or increase in restrictions to visiting friends and relative in LTCF and as such the effect of prevention of infections and deaths is likely to be due to vaccination rather than to new social distancing measures alone [5] . The IRR for infections and deaths was much lower in the 80+ group compared to the LTCF/home care group even prior to vaccination, which is likely to be an effect of the Swedish recommendations, that is, that those over 70 years take extra precautions to avoid social contact with others outside their own household [6] . Those receiving home care who have regular visits by healthcare professionals and those living in LTCF have not been able to isolate to the same extent and have been much more vulnerable to the spread of infection [7] . to initial concerns about effectiveness in older adults [9] . Vaccination with Astra-Zeneca's vaccine was paused between 16 th and 25 th March whilst reports of links with a rare blood clotting disorder were investigated [3] . Since 25 th March, vaccination with Astra-Zenecas vaccine has been resumed amongst those over 65 [3, 10, 11] . It is likely therefore that the majority of those vaccinated in the LTCF/home care and 80+ groups received mRNA vaccines although data on the exact proportions was unavailable to the authors.

The effect observed is a combination of protection from a single dose and two doses. High Sars-CoV2 infection in Stockholm [13] . The Delta-VOC was not widely circulating at any point during the study period and first started to increase in prevalence from week 22 overtaking the Alpha-VOC by week 26 in 2021 [14] .

This study was limited to programmatic effect in that we were unable to link individual level vaccine data as has recently been done in Israel [15] however little has yet been published on the impact of programmes using heterologous vaccines such as this one which is one of the studies strengths. A recent internal report from Public Health England estimates 10,400 deaths have been prevented in England in the first 4 months of their COVID-19 vaccination programme using non-individual level data [16] . The advantage of before and after studies is that they take into account both the direct and indirect effects of vaccination. Yet, the assumption is that everything remains constant pre-and post-vaccination aside from vaccination itself. In Sweden, as previously discussed, the public health recommendations remained fairly constant throughout the study period and the testing recommendations were the same for the three groups. However, it is possible that personal behaviour changed differently according to age group and if so could influence our vaccination impact estimates in both directions. Also, the LTCF had media attention for being high-spread areas, which could have led to them working harder to prevent spread. This would result in an overestimation of the vaccination impact.

An additional limitation in our study is that vaccine coverage data was based on LTCF and age, since data for those receiving home care was not available whereas incidence of infections and deaths were calculated for the combined group (LTCF/home care). Some of the reduction observed in the LTCF/home care group is likely due to indirect effects of vaccination of healthcare workers. Reductions in observed vs. expected infections in the 80+ group started before vaccination indicating that the assumptions about symmetrical infection incidence ratio during the second and third wave might not be fully accurate. Vaccination of those aged 75-79 years began on 29 th March 2021. However, considering the principle that a protective effect starts 4 weeks after the start of vaccination, this group would not be considered to have immunity until a few days before the end of the study period and thus were kept in the control group. Figure 1C shows that the vaccination coverage in the control group did start to increase before the end of the study period which was because of healthcare worker vaccination as well as vaccination being offered to people living with a person receiving homecare. Increase in vaccination of the control group only started to substantially increase during April 2021 and even by the end of the study was less than 20%. Yet, it is possible that the vaccination in the control group may have led to an underestimation of the vaccine impact in the other groups. Waning immunity and time-since vaccination were not considered in this analysis. The overall trend, however, following initiation of vaccination is highly encouraging.

In conclusion, the programmatic early effects of vaccination in two of the most vulnerable groups are both reduced SARS-CoV-2 infections and deaths.

Word count: 2020

All authors have seen and approved the manuscript and contributed significantly to the work.

PN devised the study, CI wrote the manuscript and contributed to the analysis, DS performed the main analysis, FG assisted with statistical analyses and MPH analysed and provided the vaccination data.

☐ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

☒The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: CI has previously received research funding from Pfizer.

Stockholm, Sweden Katie.isitt@gmail.com +4676155056

Dear Professor Reingold, Thank you for the opportunity to reply to the reviewers comments. We have divided the comments into individual points which we have addressed below in our point-by-point reply. See our replies in italics. We believe the manuscript has improved after the update and hope that it is now suitable for publication in Vaccine.

Authors use the comparison of observed and expected incidence rates after and before introduction of vaccination in various population categories (LTCF and >= 80y of age) using incidence rates in unvaccinated age groups (18-79 years) as a reference. AS such they correct the observed incidence in the post vaccination period using the ratio of incidences observed in the reference group after and before introduction of vaccination. Authors suggest that vaccination has substantially reduced infection and deaths in the Stockholm population.

The study question is highly relevant to guide public health strategies chosen all along vaccination against SARS-CoV-2. However, the manuscript raises several questions that need clarification before a potential publication.

Authors suggest that indicators of impact of a vaccination programme can be computed using the comparing of incidence rates of SARS CoV 2 after to before introduction of vaccination in a population. As mentioned by authors in their discussion section, the best design would have been a person time cohort study in which the vaccine status (doses, dates, intervals, brand) of each individual is taken into account and in which person with SARS CoV 2 infection are counted as cases and then excluded from the person time denominator at risk when infection or death occurs. My understanding is that, in the current study, those exclusion criteria were used. This potential bias should be further discuss and its consequences quantifies as possible.

Reply: Thank you for your comment. It is correct that deaths and infections were censured at the point they occurred. As a result, re-infections were not included and so possible cases would be lost from both the numerator and denominator. Yet, we believe that censuring individuals that tested positive gives the best possible estimate of the susceptible population, as re-infections were rare with the circulating variants. "Yet, the assumption is that everything remains constant pre-and post-vaccination aside from vaccination itself. In Sweden, as previously discussed, the public health recommendations remained fairly constant throughout the study period and the testing recommendations were the same for the three groups. However, it is possible that personal behaviour changed differently according to age group and if so could influence our vaccination impact estimates in both directions. Also, the LTCF had media attention for being high-spread areas, which could have led to them working harder to prevent spread. This would result in an overestimation of the vaccination impact."

Authors have computed the incidence rates (person-time) among the three groups after and before the introduction of vaccination. My understanding is that different time cut-offs were used for each of the computation of incidence rates and incidence rates ratio in LTCF (4w after 27th December 2020, Early Feb. on the graph) and in >= 80y (4w after 8th March 2021, early April 2021 on the graph). I understand that the after/before time periods used to compute incidence rates in the control group (18-79 years) match those of the above (i.e. "reference" in table 2 is not the same incidence rate per line). However, I do not understand why, in Table 2 , the number of cases occurring after and before introduction of vaccination are missing for the control group.

Reply: The time cut-offs for post-vaccination are as the reviewer describes. We have now added to tables 2 and 3 the numbers of infections and deaths that occurred in 18-79 year group for the LTCF 'post-vaccination' period and the 80+ 'post-vaccination' period.

Nationell plan för vaccination mot covid-19 (delrapportering 3) 2021

Information om vaccination med Astra Zenecas vaccin till personer som är 65 år och äldre

Patterns of multimorbidity and pharmacotherapy: a total population cross-sectional study

Tillfälliga lokala besöksförbud på äldreboenden upphör

Folkhälsomyndighetens föreskrifter och allmänna råd om allas ansvar att förhindra smitta av covid-19 m.m

Leveraging epidemiological principles to evaluate Sweden's COVID-19 response

Samtliga godkända vacciner mot covid-19 skyddar mot sjukdom

Om vaccinerna mot covid-19

Rekommendation om åldersgräns på 65 år för AstraZenecas vaccin kvarstår

Om vaccinerna mot covid-19

Mortality trends among hospitalised COVID-19 patients in Sweden: A nationwide observational cohort study. Lancet Reg Health Eur

Statistik om SARS-CoV-2 virusvarianter av särskild betydelse

Impact and effectiveness of mRNA BNT162b2 vaccine against SARS-CoV-2 infections and COVID-19 cases, hospitalisations, and deaths following a nationwide vaccination campaign in Israel: an observational study using national surveillance data

Impact of COVID-19 vaccines on mortality in England

The research was supported by grants from Stockholm County Council and the Swedish Research Council (Dnr 2021-04809).

It is strange that, in the control group, the SARS CoV 2 infection incidence rate after vaccination is not available while the infection incidence rate ratio is computed (and I understand that the denominator of the IRR is the incidence rate in the control group???). To fully understand table two, it would be extremely useful to readers if authors could provide for each group (LTCF, >+ 80, and controls, by time periods after / before) the number of individuals, their time at risk (denominator of the incidence rate), and the number of cases.Reply: In accordance with the reviewers comment we have now added to table 2; Individuals at risk at the start of the study, individuals at risk at the end of the study, risk-time in 100,000 person days before vaccination, risk time in 100,000 person-days after vaccination and the number of cases (sum observed). We have also given similar information in Table 3 for deaths.Authors mention that, for the post vaccination period, the study period started 4 weeks after the start of vaccination in a specific group. Could author specify if the 4 weeks data (person time at risk and cases) were dropped from the pre vaccination period?Reply: No, these data have not been dropped from the pre-vaccination period, they are included as part of the pre-vaccination period because prior to 4w after vaccination start it is assumed that immunity of a first vaccine dose has not been achieved. This is now clarified in the method as follow: In addition could authors confirm that the study period for the reference group also started 4 weeks after vaccination start in each (LTCF and 80+) of the post vaccination study period. Same issue for the pre vaccination period.Reply: Yes, we can confirm that the same study period was used for the control group (18-79 year olds) when compared with the two test groups. This has been clarified in footnotes to Table 2 

To that mater, it would be useful to have a graph showing the date vaccination started in each group, the date duration of the pre and post vaccination periods for each group. Figure 1 that indicate where we cut between pre-and postvaccination periods.

In addition, authors mention that vaccination for the 75+ (in addition to the 80+) started on March 29th 2021. Could authors clarify if this small age group (70_74 years) was kept or not in the unvaccinated reference group?Reply: Yes, this small group was included in the unvaccinated reference group. This was because the entire study period finished on the 2 nd May 2021 and if we use the same principle of vaccination being effective only 4 weeks after vaccines started to be offered to the 70-74 year old group then we wouldn't class their postvaccination cut off as being until the 26 th April 2021 which is only 6 days before the end of the study period. Therefore it is unlikely to affect the model in any significant way. As is shown in Figure 1C the vaccination coverage in the control group did start to increase before the end of the study period which is due to the effect of healthcare worker vaccination as well as vaccination being offered to people living with a person receiving homecare. However, using the principle again that it would take 4 weeks after vaccination to achieve immunity then we can say that less than 10% of the control group would in effect have vaccination coverage by the end of the study period. This may however lead to an underestimate of the vaccine effect. We have added this to the discussion as follows:"Vaccination of those aged 75-79 years began on 29 th March 2021. However, considering the principle that a protective effect starts 4 weeks after the start of vaccination, this group would not be considered to have immunity until a few days before the end of the study period and thus were kept in the control group. Figure 1C shows that the vaccination coverage in the control group did start to increase before the end of the study period which was because of healthcare worker vaccination as well as vaccination being offered to people living with a person receiving homecare. Increase in vaccination of the control group only started to substantially increase during April 2021 and even by the end of the study was less than 20%. Yet, it is possible that the vaccination in the control group may have led to an underestimation of the vaccine impact in the other groups."Authors state that to estimate the post vaccination number of expected cases among LTCF and 80+ y, they multiply the IRR before vaccination with the observed post vaccination incidence rate in the control group. This is understandable as it is. However, I wonder if it would be easier to say that to compute, for example, the expected number of cases and incidence rate in LTCF in the period after vaccination (if vaccination had not occurred) authors multiply the incidence rate before vaccination in LTCF by the ratio of the observed incidence in the control group after to before vaccination cut-offs. I believe this would better explain that authors computed expected incidence rate in the post vaccination period by correcting observed incidence in that period with the change in incidence rates in the before / after time periods in the control group (IRR).

I do not understand why authors compute expected incidence rates in the pre vaccination period? Why should we correct a pre vaccination rate for what will happen in the post vaccination period? As a consequence, I do not understand the quantified data behind the final statement in the conclusion, "Reductions in observed vs expected infections in the 80+ group started before vaccination". This does not seem to match with Figure 1 A.Reply: The expected incidence rates in the pre-vaccination period are presented as a visualization of the model fit only. We have added a footnote to Figure 1 and a text in the methods section to clarify this for the reader. In Figure 1A , the line representing observed infections (block green) starts to turn downwards before the cut off between pre and post vaccination whereas the dashed line representing expected infections turns down after the cut off. It is this that the comment refers to.The principle of after / before studies if that everything but vaccination remains constant between the two time periods. Is correcting for changes in incidence rates in the control group after/before enough?Reply: This is a known limitation of the study but as we have previously discussed, during the time period in the study, the recommendations for social distancing and working from home were constant as well as advice regarding testing and isolation with symptoms of illness. As far as possible we believe that other factors remained constant and we have added a point in the discussion to address this as replied to a previous comment by the reviewer.The changes in incidence in the control group may not reflect other changes than vaccination that may have also happened in the LTCF or >=80y. For example, in absence of vaccination, probability to be in contact with the virus may have differently changed over time in each group; % of remaining susceptibles in each of the two groups and in the controls may have been different due to very different transmission patterns (associated to age); testing policies may have differently changed over time in each group. Authors clearly touch that point when discussing the lower impact in >=80y compared to LTCF (the group being more likely to stay at home). Authors need to further discuss the validity of using the unvaccinated age group (18-79y) as a reference group and its consequences on the magnitude of the impact (over or underestimated).

Authors mention that impact studies allow to take into account direct and indirect effect. The direct effect can only be measured in a population with a vaccination programme (here corresponding to the second time period). Even if effectiveness (direct effect) contributes to the impact, it is not quantified, by study design, part of the impact measurements in this after/before studies.

The authors present an analysis of the short-term effects of COVID vaccine in 2 high risk populations (80+, LCTF) compared to disease rates expected from pre-vaccination waves and compared to the unvaccinated 18-79 population in 1 region in Sweden. Because non-pharmaceutical interventions were not different, the authors attribute the decrease in morbidity and mortality to vaccine effects.The analysis was well-performed and the authors described it well. The findings match similar studies performed in other countries but there is value in replicating the findings in different healthcare settings and policy recommendations (vaccine and other).

The authors should state which variants were prevalent during the study period.Reply; Thank you, we have now included this in the discussion as follows:

In addition, it is important to note that waning immunity and time since immunization were not evaluated in the present analysis.Reply: Thank you, this is now noted in the discussion as follows:"Waning immunity and time-since vaccination were not considered in this analysis."The authors mention that the programmatic effects were studied. Given the relatively small number of deaths in the "vaccinated" population -it would be useful to see if the deaths were in full immune persons or if they were diagnosed with COVID before or shortly after the second dose.Reply: We agree that this would be very interesting. Unfortunately, we are unable to provide this since we did not have access to vaccine data on the individual level.The authors should provide some indication of the relative frequency of each vaccine (Pfizer, Moderna, Astra-Zeneca). It sounds as most were Pfizer based on the rapidity of vaccination but it should be stated or alternatively that the information is not available.Reply: Thank you we agree this woulc be interesting. Unfortunately the authors were unable to obtain this information. A comment has been added to the discussion to this effect.The references should be formatted according to journal requirements, some of the English language references were missing article titles.Reply: Thank you, this has now been now corrected.Yours sincerely, Associate Professor Pontus Naucler and Dr Catherine Isitt Table 2 since the at risk population for 30-day mortality was based on information from 2020-08-01. **According to time point of vaccination in LTCF and 80+ respectively ***During period of vaccination in LTCF and 80+ respectively