The COVID-19 pandemic is among the greatest challenges of today’s time. As of mid October 2021, more than 236 million people have already been infected with SARS-CoV-2 with 4.8 million deaths worldwide (WHO 2021). Several vaccines have been approved (Zimmer et al. 2021) and after initial supply shortfalls, vaccination rates accelerated as of spring 2021. In high-income countries, vaccination rates with at least one dose varied between 60% and 80% by mid-October 2021 (Ritchie et al. 2021). In developing countries where supply with vaccines is scarce, vaccination rates are much lower, for example 52% in India and under 10% in most parts of Africa (Ritchie et al. 2021). It has been estimated that considering now dominating vaccines, vaccination rates of 85–95% are necessary to protect from a severe increase of infections (Weber et al. 2021).

Next to availability of vaccines, the intention to get vaccinated against COVID-19 is regarded as a key variable for predicting actual vaccination uptake; it has been shown in a meta-analysis that health-related intentions are causally linked to the respective health-related behaviours (Webb and Sheeran 2006). A high availability of vaccination doses is a necessary but not sufficient prerequisite of actual vaccination uptake. If intentions are too low in the general population or in specific subgroups, the success of a COVID-19 vaccination campaign is seriously threatened.

In this paper, we are interested in COVID-19 vaccination intentions as a function of gender. Research on other vaccines showed gender differences in vaccination status and intentions favouring men (Bish et al. 2011) which may transfer to the COVID-19 vaccine. Some early surveys also reported lower COVID-19 vaccination intentions among women (Galanis et al. 2020; Lin et al. 2021; Robinson et al. 2020). Lower vaccination intentions among women could be problematic for various reasons. Next to exposing themselves to the danger of a COVID-19 infection, women have a central role in ensuring the health of their children. Additionally, women are more likely to be health and social care workers who are at high risk of contracting and passing on COVID-19.

Before vaccines against COVID-19 were approved and vaccination programs started, worldwide surveys were undertaken to assess individuals’ intentions to get vaccinated against COVID-19 in the general population and among samples of health care workers (HCWs). The main goal of our study was to review and analyse the results of these surveys, investigating whether there are systematic gender differences in the intention to get vaccinated against COVID-19.


Search strategy

The initial search was conducted on 19.11.2020 in PubMed, Web of Science and PsycInfo. We used the search terms (vaccination OR vaccine OR vaccinated) AND (corona OR coronavirus OR SARS-CoV-2 OR COVID-19) in combination with ‘refusal’, ‘hesitancy’, ‘hesitance’, ‘hesitation’, ‘acceptance’, ‘willingness’, ‘motivation’, ‘confidence’, ‘uptake’, ‘intention’, ‘attitude’, ‘emotion’, ‘opinion’, ‘trust’, ‘doubts’, ‘cognition’, ‘rejection’, ‘disapproval’, ‘belief’. This search identified 649 articles on PubMed, 192 on Web of Science and 17 on PsycInfo. We filtered results for the year 2020 to 2021 (because the global COVID-19 outbreak happened in 2020) and in PubMed for languages English or German. We used a method described by Bramer et al. (2016) to identify duplicates. The final number of articles for screening was 682.

We identified 26 papers reporting gender specific data on vaccination intentions in any sort. We manually conducted a forward and backward citation search of those 26 papers. In this way, we identified a further 18 papers that reported gender data on vaccination intentions. Data extraction can be seen in Fig. 1.

Fig. 1
figure 1

PRISMA flow diagram showing the study selection process

After the initial search, we regularly checked data bases for new publications. Inclusions based on these searches can be seen in Fig. 1. The last search was conducted the 7th of January 2021 on all three databases. Subsequently we wrote to the authors (N = 50) who did not report the data needed for meta-analytic calculations in their articles. Lastly, we had to exclude 19 papers for various reasons.Footnote 1

Eligibility and exclusion criteria

The search results were screened for inclusion following these eligibility criteria: Primarily adult population, reported outcome: intention/willingness to get vaccinated against COVID-19 for men and women separately or gender differences statistically tested, available in English or German. Cross-sectional and longitudinal studies were included. Studies reporting an interventional/experimental design were excluded from our analysis (see Fig. 1 for study selection).

Data analysis

Apart from looking at studies descriptively, we conducted meta-analytic calculations of averaged odds ratios. For this calculation, we first computed odds ratios using the exact frequency statistics reported in the papers or provided by the authors upon request. We made two different types of calculation: We compared ‘yes’-answers (including ‘definitely yes’- and ‘probably yes’-answers and similar answer options) with the remaining ‘non-yes’ answer categories which could also include ‘do not know’- or ‘not sure’-answers (in Table 1: YR, frequencies for yes vs. rest categories reported). We additionally compared ‘yes’-answers (including ‘definitely yes’- and ‘probably yes’-answers and similar answer options) with ‘no’-answers (including ‘definitely’ or ‘probably no’-answers and similar answer options, in Table 1: YN, frequencies for yes vs. no categories reported). For the meta-analytical calculations, we included 46 studies which provided the necessary data for the ‘yes’- versus ‘non-yes’ answers and used the metafor package in R described by Viechtbauer (2010) to compute mean gender effects of summarized odds ratios and confidence intervals. We also conducted a meta-analysis that was based on a smaller number of 40 studies which provided the necessary data for the ‘yes’- versus ‘no’ -answers. The results of this additional meta-analysis can be found in the Supplemental Material. We used a random effects model for our meta-analysis due to heterogeneity in our samples that (widely) differed in dimensions such as residence, age, and profession. As a result, we cannot assume that the effect estimates vary only because of chance differences from sampling participants, in which case a fixed-effect model is indicated (Riley et al. 2011). Restricted maximum likelihood estimation was used to fit a random-effects model to the data respectively as estimator to compute the heterogeneity τ2. In the meta-analysis, we used moderator analyses to determine if the effects of the studies differed depending on quality appraisal, month of assessment or being a healthcare worker or not (variable HCW). If recruitment took place over several months, the first month was coded. HCWs were chosen as a subgroup because it was the only group addressed by several studies. Knapp and Hartung adjustment was used to lower type I error rates (IntHout et al. 2014) and can be seen as a good replacement of the standard method (Jackson et al. 2017). Representativeness of samples was not used as a separate moderator because it was included in the quality rating. Owing to the rapidly evolving situation, it was not surprising that many papers were available as preprints which is unlikely to reflect the quality of research. For quality appraisal, we used the suitable aspects of already established tools as there were no comprehensive quality assessment tools that fitted for the survey studies. More detailed information can be found in the Supplementary Material. We adhered to PRISMA guidelines in the preparation and realization of our review.

Table 1 Description of studies (n = 60) included in review


Description of the studies

Sample sizes, sampling techniques, countries and month of assessment, publication type, item wording for the variable of interest, as well as quality ratings and reported gender differences can be seen in Table 1. Sample sizes ranged from 128 (Grech et al. 2020a) to 32,361 (Paul et al. 2020) participants with a total of 195,974 people across all 60 studies. The vast majority (70%) of studies, namely 42, had sample sizes of over 1,000 participants. Most papers (n = 35) were peer reviewed, but a substantial number were preprints (n = 24), and one was a report of scientific surveys made accessible online (Perlis et al. 2020). Surveys took place in 40 different countries. Most papers included samples from the USA (n = 22), UK (n = 13), Italy (n = 5), France (n = 5) and Australia (n = 4). Twenty-three studies took place in Europe exclusively.

Wording of the vaccination intention item was similar across the surveys. Most items asked about ‘likelihood’, ‘intention’ or ‘willingness’ to vaccinate or ‘acceptance’ of a COVID-19 vaccine. However, response categories varied from two (‘yes’, ‘no’) to five or more categories and one with 11 categories (Sherman et al. 2020). Many studies (n = 24) explicitly included a ‘not sure’/‘undecided’/‘maybe’-response category. We included studies that were conducted from February 2020 (Papagiannis et al. 2020) to November 2020 (Barry et al. 2020). Most studies (n = 15) took place in April 2020 (including those studies lasting more than one month), few studies were conducted in February and November (n = 2 each). Three studies did not report the time or period of recruitment (Pogue et al. 2020; Al-Mohaithef and Padhi 2020; Lucia et al. 2020). Results of a quality appraisal for the 60 studies are reported in the Supplementary Material.

Gender differences in vaccination intentions

Thirty-six studies report significant gender differences in vaccination intentions in their result section for the whole sample. Male gender was associated with a greater likelihood of intending to accept a COVID-19 vaccine in 35 studies (58%). Only one study (Lazarus et al. 2020), reported men to be less likely to intend to accept of the vaccination compared with women. In five studies (Butter et al. 2020; Davis et al. 2020; McAndrew and Allington 2020; Salali and Uysal 2020; Khubchandani et al. 2021) results were not clear because significant gender differences could be found only in some subgroups and analyses but not in others. Most studies recruited only from the general adult population (n = 41). Twelve looked exclusively at health care workers and/or health care students (Grech et al. 2020a; Papagiannis et al. 2020; Barry et al. 2020; Lucia et al. 2020; Gadoth et al. 2020; Grech and Gauci 2020; Grech et al. 2020b; Nzaji et al. 2020; Kose et al. 2020; Kwok et al. 2020; Unroe et al. 2020; Wang et al. 2020a). Of those, eight reported significant gender differences (66.7%) as can be seen in Table 1. Four studies purposefully oversampled HCWs or key workers (Butter et al. 2020; Detoc et al. 2020; Dror et al. 2020; Grüner and Krüger 2020) to compare their intentions with the general population. Of those, only Butter et al. (2020) analysed gender differences separately for the two groups. They reported a significant association of COVID-19 vaccine hesitancy and being female only for key workers (mainly individuals employed in positions in health care, education and childcare or positions crucial for providing food, necessities and utilities).

Meta-analytic results

Forty-six papers (77%) included frequency statistics for the calculation of averaged odds ratios (ORs) or they were provided by the authors upon request. This is noted in Table 1 in the column Frequencies. Roozenbeek et al. (2020) provided us with data from more countries and months than in the original paper which is why we have a larger sample for our own calculations then they did in their paper.Footnote 2 For Sethi et al. (2020), we computed the frequencies for the ‘yes’-category from the frequencies of the other categories given in the paper. Loomba et al. (2020) conducted their study in the USA and UK but only data for the UK was available for meta-analytic computations. Daly and Robinson (2020) had frequencies for their assessment in April and October. We used data for April after verifying that follow up data for October did not make a big difference for our calculations. We conducted meta-analytic computation of the available data.

Data were available for 141,550 female and male participants, excluding people not identifying as male or female or with missing data. Of those papers not providing frequency statistics (n = 14), seven papers (50.0%) reported significant gender effects in their results section in favour of men and two papers each found significant effects for one of two subgroups. For the papers with reported or provided frequency statistics this percentage was higher with 60.9% of the papers (n = 28) reporting significant gender effects in favour of men. Mean quality rating of the papers without frequencies was M = 7.92 (SD = 2.3) with six papers (42.9%) with a rating of nine and higher (up to 12 which was the maximum). Mean quality rating of the papers with frequency statistics provided was M = 8.02 (SD = 2.19) with 18 papers (39%) with a rating of nine and higher and three papers with a rating of 12. Mean quality ratings did not differ, U = 294.00, Z = −0.495, p = 0.62.

Not all of the 46 papers had frequencies broken up into every answer category but summarized over several categories so that absolute ‘no’-answers could not be obtained for all of them. We therefore compared ‘yes’-answers with the rest-categories, that is, all but the ‘yes’-categories, including ‘no’ and ‘not sure’-answers. Wang et al. (2020b) was the only study that contained ‘delay of vaccination’ in the rest-category. The averaged OR was 1.41, 95% CI [1.28, 1.55] with higher odds for men than for women. This effect was significant, z = 7.10, p < .0001. The heterogeneity among the studies was substantial with I2 = 93.87%, Q = 542.83, p < 0.0001. The lowest OR was 0.49, 95% CI [0.40, 0.58] and the highest OR was 2.88, 95% CI [1.74, 4.77]. Figure 2 displays the corresponding forest plot.

Fig. 2.
figure 2

Forest plot of the odds of men reporting the intention to get vaccinated against COVID-19 compared to the reference group of women in unspecified samples (above) and Health Care Workers Samples (HCW, below). Results are expressed as odds ratio (OR) and 95% confidence intervals

Moderator analyses with study quality, first month of assessment and a HCW status as moderators revealed a significant moderation effect, F = 5.22, p = 0.004 (Ritchie et al. 2021; Grüner and Krüger 2020). Model results showed that only the factor HCW was a significant moderator for the observed study effects, t = 3.51, p = 0.001 (quality: t = −0.36, p = 0.720; month: t = 0.03, p = 0.975). The amount of heterogeneity R2 accounted for was 25.26%. Subgroup analysis revealed a significant subgroup difference for the yes vs. rest analysis with QM = 23.65, p = 0.00. Heterogeneity in the HCW subsample was lower than in the other subgroup but substantial in both (see Fig. 2). Averaged odd ratios for the subgroup of HCW was OR 1.79, 95% CI [1.61, 2.36] vs. OR 1.31, 95% CI [1.19, 1.44] for the unspecific sample.


In our systematic review we investigated gender differences in COVID-19 vaccination intentions. In our meta-analysis of averaged odds ratios across all the studies that provided us with the necessary frequency data (n = 46) we found an overall significant gender difference with males being on average 41% more likely to report that they intended to receive a vaccine (rather than being unwilling or undecided) compared with women. Quality ratings of the studies or first month of assessment did not have a significant impact on study effects. Subgroup analyses in response to our moderator analyses revealed that gender effects were even higher among health care workers (HCWs) compared with unspecific samples. However, this result must be interpreted cautiously because in HCW samples gender proportions were highly unbalanced and the number of studies with HCW samples was comparatively small.

Our finding that men showed on average a higher COVID-19 vaccination intention supports initial trends indicating systematic gender differences in reviews of COVID-19 vaccination intention (Galanis et al. 2020; Lin et al. 2021; Robinson et al. 2020). They are also in line with research on other vaccinations. For example, a study of vaccination coverage among adolescents found that females had a lower likelihood of being fully vaccinated compared with men (Sakou et al. 2011). Men have also been found to have higher vaccination rates than women in the case of influenza and pandemic influenza vaccinations (Bish et al. 2011; Pulcini et al. 2013; Jiménez-García et al. 2010).

Vaccination intentions and actual vaccination uptake

In our efforts to compare COVID-19 vaccination intentions with the uptake of COVID-19 vaccinations, the majority of data has not yet been broken down by gender. In the COVID-19 Sex-Disaggregated Data Tracker (Global Health 50/50 2021), data only refer to the proportion of men/women in a country among all vaccinated people. This is skewed given that in some countries small numbers of people have been offered the vaccine to date.

There is much less data on the proportion of men and women who have accepted an offer to be vaccinated. In Germany, a representative survey conducted in August 2021 with 4,144 adults, showed that 79% of men and 73% of women reported that they have received a first vaccination dose (Huebner and Wagner 2021). In Austria, as of October 10, among most age groups (55 to over 84 years old), more men than women received a first dose of the COVID-19 vaccine (e.g. 98% of men vs. 90% of women among those aged over 84) (Bundesministerium Soziales Gesundheit Pflege und Konsumentenschutz Österreich 2021). Only in two age groups, namely between 15 and 24 years and 45 to 54 years, slightly more women had been vaccinated by mid-October. In the UK, overall 90.1% of females compared to 87.7% of males have been vaccinated with at least one dose since the vaccinations started (National Health Service 2021).

Evidence about vaccination uptake among HCWs in the UK and USA support our findings about female HCWs being more hesitant to get vaccinated. In the SIREN study in the UK on 29,378 hospital personnel, male HCWs were significantly more likely to be vaccinated than female HCWs, namely 90.8% of men vs. 88.1% of women (Hall et al. 2021). Among members of the Athens Medical Associations, more men (86.4%) than women (83.8%) were vaccinated. This difference failed to reach significance though. In the USA by July 2021, in a representative sample of 1,591 HCWs, female HCWs were less likely to be vaccinated, with 69% of female HCWs compared with 79% of male HCWs being vaccinated (Lazer et al. 2021).

Many of the studies included in this review asked individuals about their intentions to get the vaccine before a vaccine was available. It is well established that intentions do not always materialise into behaviour (Sheeran and Webb 2016). Usually, people are more likely to state they intend to do something and subsequently fail than the other way around. For example, in the field of physical activity, people often intend to exercise but do not always successfully translate this intention (inclined abstainers) into behaviour (Rhodes and de Bruijn 2013). In contrast, COVID-19 vaccine uptake in the UK is currently higher than anticipated, e.g. 64% of UK adults intended to get the vaccine when surveyed in September/October 2020 (Paul et al. 2020), while over 78.5% of people have received the first dose of the vaccine one year later (Government UK 2021b).

Individual vs. policy factors

The fact that vaccination uptake was in most cases higher than indicated by early surveys may be attributable to a number of factors. Thinking of the intention-behaviour gap, a certain proportion of women (but also men) who had expressed low intentions, turned out to be ‘disinclined actors’, i.e. people who originally did not intend but nevertheless acted on something (Sheeran 2001). Information campaigns and the implementation of roll out may have addressed individual modifiable barriers that underpinned vaccine hesitancy at the time the surveys were conducted. In this respect, it can be considered a success that early data helped policy makers increase uptake and diminish gender disparities. In part, the on average lower intentions stated by women found in our meta-analysis may have been overcome by several factors (especially in high-income countries). Rising infections and the associated increased mortality, positive experiences with the COVID-19 vaccination by millions of people and very high initial uptake among high-risk groups positively influencing perceived norms to accept the vaccination may have contributed to this. Another important influence would have been policies around vaccination passports and increased personal freedom.

Regarding current vaccination mandates and policies, no country has a federal vaccination mandate. However, in the USA, many institutions, including universities, hospitals and big companies such as Walmart, require a vaccination by their employees or are about to install a vaccination mandate (Hals 2021). Additionally, the new Biden–Harris Administration will demand vaccination requirements for staff within all Medicare and Medicaid-certified facilities (Centers for Medicare and Medicaid Services 2021). In the UK, by 11 November 2021, care home staff must be fully vaccinated (Government UK 2021a). In Austria, similar to the USA, certain institutions in certain regions are allowed to and do require a COVID-19 vaccination, especially for new staff members (Tempfer 2021). In Germany, legislation to make COVID-19 vaccinations mandatory for health care workers has recently been passend and will come into effect in March 2022 (Bundesgesundheitsministerium 2021). A vaccine mandate for the general German population is also being discussed. These differences in actual vaccination policies and next to that different ways of promoting the vaccines and communicating vaccination information may play an important role in convincing vaccine non-intenders to get vaccinated.


Some limitations have to be addressed. We were not able to compare vaccination intentions in men and women among subgroups, for example, age groups or education levels. Therefore, we cannot rule out that our findings may be more or less pronounced in certain subgroups. We used a dichotomous format in our analyses; therefore, we were not able to see if women were maybe more hesitant but not strongly rejecting of the vaccine. Additional analyses distinguishing between people answering ‘yes’ or ‘no’ to vaccine acceptance with certainty and those being ‘unsure’ revealed that a greater proportion of women reported being ‘unsure’. Respondents being unsure might have been more easily convinced to take the vaccine once the campaigns started in comparison with those having had a strong negative opinion. In addition, meta-analytic calculations have not been adjusted for potential confounders such as country or study design. Accordingly, comparability between studies included in the meta-analysis is limited which is reflected in a rather high heterogeneity score.

Our search was conducted from November 2020 to January 2021 and therefore our findings do not incorporate if vaccination intentions changed from then on.

Implications for policy and practice

Even if a large fraction of people – at least in high-income countries – is vaccinated already, the number of those who are not remains high. In the USA by the end of October 2021, 44% of the population were not fully vaccinated, in the UK 33% and in Germany 34% (Ritchie et al. 2021). In most countries where availability of vaccines is sufficient, people who had a high intention should be vaccinated by now. It might be necessary to focus on policy measures rather than individual psychological factors to reach the last share of unvaccinated people. It would be interesting to compare the policy measures in their effectiveness to convince still unvaccinated people in different countries. In the field of HCW, the next step is to find out how to convince hesitant women (and men) to get the vaccine.