COVID-19 vaccines are effective at preventing symptomatic and severe infection among healthcare workers: A clinical review

Introduction Health care workers (HCWs) have been at increased risk of infection during the SARS-CoV-2 pandemic and as essential workers have been prioritised for vaccination. Due to increased exposure HCW are considered a predictor of what might happen in the general population, particularly working age adults. This study aims to summarise effect of vaccination in this ‘at risk’ cohort. Methods Ovid MEDLINE and Embase were searched, and 358 individual articles were identified. Of these 49 met the inclusion criteria for review and 14 were included in a meta-analysis. Results Participants included were predominantly female and working age. Median time to infection was 51 days. Reported vaccine effectiveness against infection, symptomatic infection, and infection requiring hospitalisation were between 5 and 100 %, 34 and 100 %, and 65 and 100 % (respectively). No vaccinated HCW deaths were recorded in any study. Pooled estimates of protection against infection, symptomatic infection, and hospitalisation were, respectively, 84.7 % (95 % CI 72.6–91.5 %, p < 0.0001), 86.0 % (95 % CI 67.2 %-94.0 %; p < 0.0001), and 96.1 % (95 % CI 90.4 %-98.4 %). Waning protection against infection was reported by four studies, although protection against hospitalisation for severe infection persists for at least 6 months post vaccination. Conclusions Vaccination against SARS-CoV2 in HCWs is protective against infection, symptomatic infection, and hospitalisation. Waning protection is reported but this awaits more mature studies to understand durability more clearly. This study is limited by varying non-pharmacological responses to COVID-19 between included studies, a predominantly female and working age population, and limited information on asymptomatic transmission or long COVID protection.

Introduction: Health care workers (HCWs) have been at increased risk of infection during the SARS-CoV-2 pandemic and as essential workers have been prioritised for vaccination.Due to increased exposure HCW are considered a predictor of what might happen in the general population, particularly working age adults.This study aims to summarise effect of vaccination in this 'at risk' cohort.Methods: Ovid MEDLINE and Embase were searched, and 358 individual articles were identified.Of these 49 met the inclusion criteria for review and 14 were included in a meta-analysis.Results: Participants included were predominantly female and working age.Median time to infection was 51 days.Reported vaccine effectiveness against infection, symptomatic infection, and infection requiring hospitalisation were between 5 and 100 %, 34 and 100 %, and 65 and 100 % (respectively).No vaccinated HCW deaths were recorded in any study.Pooled estimates of protection against infection, symptomatic infection, and hospitalisation were, respectively, 84.7 % (95 % CI 72.6-91.5 %, p < 0.0001), 86.0 % (95 % CI 67.2 %-94.0 %; p < 0.0001), and 96.1 % (95 % CI 90.4 %-98.4 %).Waning protection against infection was reported by four studies, although protection against hospitalisation for severe infection persists for at least 6 months post vaccination.Conclusions: Vaccination against SARS-CoV2 in HCWs is protective against infection, symptomatic infection, and hospitalisation.Waning protection is reported but this awaits more mature studies to understand durability more clearly.This study is limited by varying non-pharmacological responses to COVID-19 between included studies, a predominantly female and working age population, and limited information on asymptomatic transmission or long COVID protection.

Introduction
Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has been responsible for the on-going worldwide pandemic since the virus was first identified in 2019.Rapid development, licensing and delivery of a number of different vaccine type [1] have altered the disease course of coronavirus disease 2019 .Vaccination is highly efficacious in preventing infection and severe disease in phase III trials [2][3][4][5] and has been evidenced by 'real world' effectiveness data showing reduced hospitalisation and improved survival, including with several SARS-CoV-2 variants of concern (VOCs) [6][7][8][9].However, despite huge improvements in outcome, rates of infection have remained high and reinfections are not uncommon.The reasons for this are not fully understood, but include rapid waning of immunity [10][11][12], immune escape of new SARS-CoV-2 variants [13][14][15][16], and suboptimal generation of mucosal protective immune responses [17][18][19].
An ideal population to study the relationship between SARS-CoV-2 infection and vaccination is healthcare workers (HCWs).HCWs are one of the few groups who been prioritised for vaccination and so evidence of (re)infection from vaccine waning is likely to occur first in this group.In addition, they have had regular access to screening and diagnostics enabling more rigorous estimates of rates SARS-CoV-2 infection.As a result, there have been many single centre and national healthcare worker studies which can be usefully summarised to understand the benefits of vaccination in this 'at-risk' cohort.
This systematic review with meta-analysis was conducted to assess the vaccine effectiveness (VE) of a primary vaccine course against COVID-19 infection, symptomatic infection, hospitalisation, and death among healthcare workers.Collating and understanding the relationship between infection and vaccination in a working age population is important to guide policy to maintain the health of this key workforce but also inform risk in the wider working age population.

Search strategy
Ovid MEDLINE and Embase were searched on 11th November 2022 using combinations and variations of the terms "COVID-19", "SARS-CoV-2", "vaccination", "immunisation", "healthcare worker", "vaccine effectiveness", and "breakthrough infection" (Fig. 1a).Search terms were chosen based on exploratory searches and the supporting literature review found in Hall et al. [20] The papers were extracted and then independently assessed for inclusion by FA and OG, disagreements were resolved by consensus.This review identified HCW observational studies of COVID-19 VE against infection including the groups any infection (including asymptomatic and symptomatic), symptomatic infection alone, hospitalisation, and death.Studies were included if they compared HCWs that had received a full primary course of vaccination to unvaccinated HCWs.Time from second dose of primary vaccine schedule to infection and risk factors that might predict infection were extracted.Studies were included if the SARS-CoV-2 infection was confirmed by PCR or antibody seroconversion.This review included studies regardless of vaccine methodology but only included mRNA and adenoviral vector vaccines in the meta-analysis.In addition, only studies that define vaccine breakthrough (VBT) as SARS-CoV-2 infection ≥ 14 days after completing a primary vaccination course were eligible for meta-analysis.
No publications were excluded based on country, race, gender, time, SARS-CoV-2 variant, or language.However, articles were only included if full text was available.Any studies that did not present healthcare worker data separately from other workers or pooled data from partially and completely vaccinated people were excluded.Case reports, case series, controlled trials, systematic reviews, and meta-analyses were excluded.Editorials, commentaries, interviews, abstract only, and conference papers were excluded.Finally, studies that only assessed immunogenicity of vaccines were excluded.All studies from beginning of the COVID-19 pandemic in 2020 to the date of literature search were eligible.

Statistical methods
Studies that presented individual participant data were transformed to a hazard ratio (HR) using a complimentary log-log link equation and pooled with reported HRs and incidence rate ratios using a three level random effects model to produce an estimate of HR [21][22][23], which was Sub-group analyses by vaccine type are presented.Studies with empty cells had a treatment arm continuity correction with sum of one prior to transformation with complimentary log-log [24].Unless otherwise specified, point estimates are presented with 95 % confidence intervals (CI) and medians are presented as medians and interquartile range (IQR).
Estimates of overall proportions are an average of the estimates provided by studies weighted by their population size.Median age and time to infection are calculated using a median of medians approach.Studies presenting mean was converted to a median using techniques described by Wan et al. [25].All analyses were conducted using R (ver.4.2.2) [26], RStudio [27], and the meta [28] and metafor [29] packages.
The mean follow-up of all included studies was 5.7 months (range 1-16.5),74 % (range 32-87 %) of participants were women, and the

Table 1
Characteristics of included studies.† These periods refer to the time periods studies report as contributing to their calculation of VE, which was not necessarily the length of the while study (e.g Larese-
Four studies provide estimates of VE across time.Bianchi et al. [59], Hall et al. [69], and Poukka et al. [33] show waning VE, particularly after six months, independent of vaccine type.Consonni et al. [70] demonstrate rising VE in HCWs with evidence of SARS-CoV-2 infection prior to vaccination and falling VE in HCWs without a prior history of SARS-CoV-2 infection (Fig. 2).Fig.3 Only thirteen studies [31,33,42,47,51,53,55,[57][58][59]63,65,69] were suitable for meta-analysis estimating VE against any infection.These studies demonstrate a significant VE of 84.7 % (CI 72.6-91.5 %, p < 0.0001) despite the relatively short follow up period of these studies (2.5-10 months).The estimated variance components are τ 2 = 0.83 between studies and τ 2 = 0.16 within studies.This corresponds to 81.9 % of the total variance being attributable to differences between studies and 15.3 % to differences within studies.Sub-group analysis by vaccine methodology showed an improved VE of mRNA over adenovirus vector or heterologous vaccination, although this does not reach statistical significance (Table 2).Studies that performed head-to-head analyses had similar conclusions [33,69].
The only study that specifically acknowledged the difference between asymptomatic and symptomatic infection was Knobel et al. [52] which found mRNA VE against asymptomatic infection to be 90.6 % in the four and a half months following vaccination.
One study compared the time of symptoms between vaccinated and unvaccinated HCWs and found a trend toward shorter illness (median length of illness of 5 days vs 9 days).[37] Symptoms among vaccinated HCWs were typically milder than unvaccinated HCWs [37,43,48].
Data suitable for pooling were provided by five studies [31,33,58,59,71]; with a pooled estimate of VE against hospitalisation of 96.1 % (CI 90.4-98.4%, p < 0.0001).The estimated variance components are τ 2 = 0 between studies and τ 2 = 0.69 within studies.This corresponds to 52.2 % of the total variance being attributable to within study heterogeneity.Sparse data prevented a subgroup analysis by vaccine type.
The effect of job role was variable.Basso et al. [76] found that VBTs were more frequent among nurse aids and auxiliary personnel (39 %), followed by nurses (36 %) and physicians (14 %).Technicians (7 %) and administrative and support staff (5 %) were less likely to have VBT.Their logistic regression identified only administrative and support staff and technicians to be at decreased risk of VBT when compared to physicians (OR 0.34 CI 0.14-0.70 and 0.32 CI 0.14-0.84,respectively); this was only maintained for technicians in the multivariate analysis (aOR 0.29 CI 0.13-0.67).Likewise Allen et al. [34] found that nurses and healthcare assistants had highest risk of VBTs (1.4 CI 1.0-1.8 and 1.8 CI 1.3-2.3;compared to administrative staff).Alishaq et al. [75] found that compared to nurses and midwives, clinical support services, administrators, and physicians were at increased risk of VBT (HR 9.15 (CI 1.32-63.64),4.10 (CI 1.13--14.90),6.27 (CI 1.20-32.82)).Allied Health Professionals and Non-clinical Support Services appeared to be at similar or slightly increased risk to nurses and midwives (HR 2.99 (CI 0.94 -9.50) and 4.95 (CI 1.00 -23.45)).However, Anshory et al. [78] Table 2 Subgroup analyses of Vaccine Effectiveness against Any Infection and Symptomatic Infection by Vaccine Type.Heterologous refers to two dose courses containing both mRNA and adenoviral vector vaccines.a Lan et al [53] removed from subgroup analysis as they pooled multiple vaccine types; b Gaio et al [64] removed from subgroup analysis as they pooled multiple vaccine types.VE=Vaccine Effectiveness; 95 % CI=95 % Confidence Interval.All subgroup estimates considered significant at p < 0.05.found neither resident doctors nor nurses were at higher risk of vaccine breakthrough than specialist doctors (aRR 5.01 CI 0.76-32.94and aRR 3.15 CI 0.70-14.14,respectively).Almufty et al. [73] identified dentists to have highest chance of VBT (43 %), followed by pharmacists and physicians (35 % and 33 % p < 0.001); paramedics had the lowest chance of VBT (8 % p < 0.001), although they had the highest chance of breakthrough requiring ventilation (33 % vs between 1-10 %, p < 0.05).Kale et al. [31] found patient-facing roles (doctors 22 % and nurses 24 %) had increased risk of infection compared to non-patient facing roles (technicians 7 %, non-medical 4 %, and housekeeping staff 3 % p < 0.001).Likewise, Rivelli et al. [44] found that staff in clinical roles were at increased risk of breakthrough when compared to non-clinical roles (aOR 2.29 CI 1.36-3.84)and this risk was increased when comparing COVID clinical roles to non-clinical ones (aOR 7.36 CI 2.99-18.09).Likewise, Allen et al [34] found that, when compared to HCWs with no patient contact, both HCWs with contact to patients without COVID-19 and HCWs with contact to patients with COVID-19 had increased risk of VBT (aRR 1.3 CI 1.1-1.5 and aRR 1.4 CI 1.1-1.7,respectively).Alsihaq et al [75] identified that HCWs with VBT were more likely to reported contact with a COVID-19 case than those without, particularly their spouse (27 % vs 6 % p < 0.00001), other close family member (10 % vs 4 % p < 0.017), or patient (14 % vs 3 % p < 0.00001).Uninfected were more likely to deny contact with individuals with COVID-19 (80 % vs 44 % p < 0.0001).However, Almufty et al [73] reported no difference in chance of VBT between those working in COVID centres and other healthcare locations (30 % vs 27 %, p = 0.551).Anshory et al [78] identify contact with COVID-19 as increasing risk of breakthrough when outside the home or hospital (aRR 6.82 CI 1.97-47.98).They assess the effect of Personal Protective Equipment (PPE) use on infection risk.Although the amount of PPE worn had no effect, N95 and KN95 masks were more protective than surgical masks (aRR 0.05 CI 0.01-0.45and 0.06 CI 0.01-0.51,respectively).Likewise, reporting never wearing a mask increased risk of breakthrough (aRR 7.12 CI 1.88-26.96)[78].

Discussion
We performed a clinical review and meta-analysis of observational studies of vaccine effectiveness against SARS-CoV-2 infection among HCWs.Reviewing observational studies provides essential data of the real-world effectiveness of COVID-19 vaccines and by studying HCWs we have limited the influence of social distancing measures on VE estimates.Vaccination is found to be effective against any infection, symptomatic infection, and hospitalisation.Of 866,163 vaccinated HCWs included in this review no individual died of COVID-19.In other similarly age cohorts, small numbers of deaths are recorded, and these are restricted to those over 40s years of age [80][81][82].There was waning protection against infection, particularly past six months, but protection against symptomatic and severe infection persisted.VE reduced with VOCs, particularly omicron BA.1.Sub-group analysis by vaccine type suggests mRNA vaccines are more effective than adenovirus vector vaccines at preventing symptomatic or asymptomatic SARS-CoV-2.In addition, vaccines reduce the severity of disease; vaccinated HCWs were less likely to have symptoms and more likely to have milder symptoms compared to their unvaccinated colleagues.Lowest VE estimates came from India, adenoviral vector or whole inactivated virus vaccines, and VOCs; It is difficult to tease apart the individual effects of these factors as their effects are confounded with each otherparticularly the low VE of adenoviral vector and whole inactivated virus vaccines and low VEs reported from India as the studies conducted in India all used adenoviral vector and whole inactivated virus vaccines.
This demonstrates the effectiveness of COVID-19 vaccines in an essential workforce.The waning protection against non-hospitalising infection supports the decision to roll out booster programmes among working age adults at high exposure risk to SARS-CoV-2.This correlates with waning antibody titres seen in similar populations [10][11][12][83][84][85] and the longer lasting protection against hospitalising infection suggests an important cellular component to immunity against severe SARS-CoV-2 infection.
Our mixed findings on the effect of risk factors on VE may be explained by the restricted population included.Observational studies of the general population indicate increasing risk with increasing age [81,86], Hippisley-Cox et al [82] found an exponential increase in risk beyond middle age; our inconclusive findings are largely due to all included participants being of a low risk age for breakthrough and the possibility that older HCWs being deployed away from high exposure areas.Hippisley-Cox et al [82] reported the risk of vaccine breakthrough being highest at the extremes of BMI, and highest at a BMI of 15 [82]; which may explain the conflicting results of Basso et al [76] and Alishaq et al [75], who reported BMI>25 and BMI<30 increase risk of VBT (respectively).In the general population, male gender appears to increase risk of VBT and it is difficult to explain the mixed results seen in this review; but it may be due to the gender skew of the studies assessing this (Table 1).
Likewise, the inconsistent effect of profession may be because professional background (e.g.physician or radiographer) does not reliably predict SARS-CoV-2 exposure.Direct contact with a confirmed COVID-19 case increased vaccine breakthrough risk in HCWs and this is similar to the finding that increased background SARS-CoV-2 prevalence increases risk of VBT.[82] The evidence provided by Anshory et al [78] supports that the quality and availability of PPE, not professional background, is more directly linked to vaccine breakthrough risk.In addition, job role is confounded by socioeconomic status and education level, both have been linked to increased likelihood of SARS-CoV-2 infection [34,87,88].

Limitations
We acknowledge several limitations to this systematic review.We were unable to account for differences between studies in nonpharmacological interventions at work or home (e.g.lockdowns or social distancing) and PPE quality and availability.We are also unable to account for the effect of variable testing regimens.This is particularly important when assessing protection against asymptomatic infection, where studies with more proactive testing regimens are more likely to detect infection (particularly mild or asymptomatic infection).Recent modelling work suggests that in observational studies testing infrequently or only symptomatic participants, there would be a tendency to underestimate VE [89].The most common testing regimen of included studies included in our review was regular nasopharyngeal PCR testing or monitoring for seroconversion, which are least vulnerable to this effect [89].Together, these limitations may explain (in part) the high heterogeneity of the meta-analysis estimates and wide variation in reported VE estimates.
Our comparison of VE between different vaccine types and VOCs is limited by a relative lack of studies conducting direct comparisons.In addition, the cohorts reviewed were typically working age and femalewhich makes the study results less transferable to other 'at-risk' populations.In addition, we may be vulnerable to the healthy worker effect, which can be more pronounced in women [90].
The included literature did not assess vaccination's effect on nosocomial infection or asymptomatic transmission of SARS-CoV-2, but a systematic review of observational studies suggests that vaccination can reduce onward SARS-CoV-2 transmission (VE against transmission 16-95 %), however this is lower with more recent variants and they could not assess the VE of vaccines targeted to more recent VOCs [91].In addition, none of the studies included in this review reported VE against long COVID, therefore we could not assess this; however, other work suggests that risk and severity of long COVID is increased by severity of infection [92][93][94][95][96] and our finding of reduced severity of infection in vaccinated HCWs suggests that, by extension, vaccination may lessen the burden of long COVID among HCWs (a finding seen in other populations [97]).We were unable to assess the effect of race and ethnicity on risk VBT due to the variability of racial and ethnic groupings between countries and limited reported evidence.

Conclusion
This review demonstrates real world effectiveness of COVID-19 vaccines among a working age population, particularly against severe illness.We find evidence of waning VE against any SARS-CoV-2 infection within 6 months and reduced effectiveness against the delta and omicron VOCs, supporting the decision to provide further booster vaccinations for highly exposed workforces (including HCWs) and ongoing development of new vaccine candidates.Understanding the effectiveness of vaccines in the working age population is useful going forward in terms of maintaining workforce health and economic productivity, and this allows us to focus resource and advice.Further work is needed to identify risk factors for VBT and protection against long COVID.

Fig. 1 .
Fig. 1.(a) Search Strategy and (b) PRISMA Diagram.Panel a indicates the terms used to search Ovid MEDLINE and Embase and their combinations.PRISMA diagram shows number of studies removed at each step in the study selection process.PRISMA=Preferred Reporting In Systematic Reviews and Meta Analyses.

Fig. 2 .
Fig. 2. VE against (a) any SARS-CoV-2 Infection, (b) Symptomatic SARS-CoV-2, and (c) Hospitalisation from SARS-CoV-2 Infection over time.Time from vaccination is measured in months and calculated from the end of the primary vaccination course.Each estimate is plotted at the end of the period reported, e.g. a study reporting a VE between 3 and 6 months from vaccination will be plotted at 6 months.Studies that feature in more than one panel have the same symbol used throughout.VE=Vaccine Effectiveness, SARS-CoV-2 = Severe Acute Respiratory Distress Syndrome Coronavirus 2.

Fig. 3 .
Fig. 3. Meta-analysis of VE Estimates against (a) any SARS-CoV-2 Infection, (b) Symptomatic Infection SARS-CoV-2, and (c) Hospitalising Infection SARS-CoV-2.VE estimates of included studies are plotted as a square with horizontal bar representing 95 % CI.A study's point estimate size is proportional to the weight of that study in the meta-analysis presented.The summary estimate is plotted as a diamond.The time after vaccination, vaccine mechanism assessed, and point estimate with 95 % CI is included.Some studies included estimates for multiple vaccines and/or multiple time points, and these are all presented where included (including different vaccines with the same mechanism where multiple are presented in a paper).Hall et al. [69] and El Adam et al. [66] present VE for different intervals between V1 and V2; for Hall et al. [69] 'Short' indicates a dose interval < 6 weeks and 'Long' an interval > 6 weeks, for El Adam et al. [66] 'Short', 'Mid', and 'Long' indicate dose intervals of 3-5 weeks, 6 weeks, and ≥ 6 weeks, respectively.*Lan et al. [53] and Gaio et al. [64] pool estimates from mRNA and adenoviral vector vaccines, therefore no mechanism is listed.VE=Vaccine Effectiveness, SARS-CoV-2 = Severe Acute Respiratory Distress Syndrome Coronavirus 2, 95 % CI=95 % Confidence Interval, V1 = first vaccine dose, V2 = second vaccine dose.

up After Vaccination Time Period of Vaccine Assessment y % Female Age
[33]n et al conducted their study from March 2020 to May 2021, but their VE estimate was calculated from data collected between March 2021 and May 2021.*calculated from data presented in paper.SD=Standard deviation, NR=not reported.Age is presented as median (IQR) unless otherwise stated.agewas39(IQR35-45)(Table1).Studies defined VBT as a positive test for SARS-CoV-2 either ≥ 7 days or ≥ 14 days following completion of a primary vaccination course.An infection was most commonly defined by a positive PCR test, although Kale et al.[31]and Moncunil et al.[32]used PCR or anti-N antibody seroconversion and Poukka et al.[33]used "laboratory confirmed" with no further explanation.Included studies covered the period from December 2020 to May 2022, thereby providing information about VE against infection by wildtype, alpha, delta, and omicron BA.1 VOCs. median