The literature search returned 5447 studies and following removal of duplicates, 4739 studies were considered for title and abstract screening. Following title and abstract screening, 55 full-text articles were assessed for eligibility. Of these, 23 studies met the inclusion criteria and were deemed eligible for the review43–65. Reasons for exclusion at full-text screening included a lack of specific focus on the COVID-19 vaccine and the absence of extractable raw data (Fig. 1). Cohen’s Kappa was 0.7 at title and abstract screening and 0.8 at full-text screening.
Summary of Included Studies
All studies had a cross-sectional study design: 22 used online surveys43–46, 48–65 and one study used a mixed-methods approach, including both an online survey and semi-structured interviews47. Eighteen surveys were conducted in the first half of 2020 (January-June) 43–46,48−53,55–58,62−65 and five were conducted in the latter half of 2020 (July-December) 45,54,59−61. Fifteen countries were represented in the review, the commonest being the USA (n = 4)50,54,58,59 and the UK (n = 4)47,56,61,63. In total, 41,403 participants were sampled across all surveys, with sample sizes ranging from 52554 to 567744 participants. The majority of participants were aged between 25–50 years old, and all surveys reported a higher proportion of female participants with one exception54. Ethnicity data were only reported in nine studies47,50,54, 56–59,61,64, with Black, Asian and Minority Ethnic (BAME) representation ranging from 3.6%64 to 36.7%50 (Additional File 1).
Survey questions used to assess vaccination intentions could largely be categorised into two; 18 studies used neutral questions such as ‘Will you get the coronavirus vaccine when available?’ 43,46–50,52−55,57,58,60−64; five studies used persuasive language that may have potentially influenced self-reported vaccine acceptance, for example ‘If a new vaccine for COVID-19 was released that was proven to be safe and effective, I would get vaccinated immediately’44,45,56,59,65. Fifteen studies recorded responses using a Likert-scale, adopting variations of the terms ‘Strongly Agree to Disagree’43–45,47,49,53,55,58,59,61−65, seven studies utilised a simple ‘Yes’, ‘No’ and/or ‘Maybe’ response scale46,48, 50–52,56,57,60 and one study used a best-fit statement response54 (Additional File 2).
Quality Assessment and Risk of Bias
Of the 23 studies included in the review, 17 studies used piloted, trialled or previously published survey instruments (Additional File 3) 43,44,47,48, 50–61,64,65. Only 11 studies used an adequate sampling frame to achieve a representative sample45,46,50,51, 54–56,58,59,61, 62 and 10 studies were deemed to use an adequate selection process47,50,51, 54–57,59,61,62. Of the studies that used adequate sampling frames, eight used existing online research panels46,50,54,56,58,59,61,62 (two most common being Qualtrics, n = 254,56 and the AmeriSpeak panel, n = 250,59). Sixteen studies did not categorise non-responder rates43,45,47–49,51,52,54,56,58,60−64 and 11 non-responder rates raised concerns over non-response bias44–46, 49,53,55,59,60,62,63,65. Non-responder bias could not be determined for six studies due to lack of adequate information43,48,52,56,61,62.
Five studies were removed from the meta-analysis due to the use of persuasive questions to assess vaccine intentions44,45,56,59,65. From the 18 studies included in the meta-analysis, the pooled proportion of survey participants willing to receive the COVID-19 vaccine was 73.3% (n = 18, 95%CI 64.2–81.5%, I2 = 99.7%, p = 0.00 Fig. 2) 43,46–50,52−55,57,58,60−64. Only two studies included in the meta-analysis reported a higher proportion of participants unwilling to receive the vaccine (71.3% in Sallam et al.60 and 51.9% in Mouchtouri et al).55
Across the 10 studies that included four response categories (variations of ‘strongly agree’, ‘agree’, ‘disagree’ and ‘strongly disagree’), individuals were more confident in accepting the vaccine than rejecting the vaccine43–45, 47,49,53,55,58,63,64. A mean proportion of 51.3% participants were definitely willing, compared to only 30.7% participants possibly willing to receive the COVID-19 vaccine. Contrastingly, a higher proportion of participants reported improbable rather than very improbable intentions to receive the COVID-19 vaccine, a mean proportion of 6.0% and 4.9% respectively.
Across the 15 studies that included a hesitant response choice (variations of ‘maybe’), participants were more likely to be vaccine hesitant than either improbable/very improbable, with a mean proportion of 22.2% and 9.4% respectively 43–45, 49,50, 55–59,61, 63–65.
Factors Associated with Vaccine Intentions
A lower perceived individual risk and perceived severity of COVID-19, lower levels of worry regarding the pandemic and lower perceived likelihood of becoming infected with COVID-19 were all found to be major variables reducing vaccine acceptance in all eight studies investigating these factors46,49,53,57,58,62,64,65. One survey reported that personal fear about COVID-19 meant the individual was almost 2.5 times significantly more likely to accept the vaccine (Odds Ratio (OR) 2.5, 95%CI 2.0–3.0, p < 0.001) compared to individuals with no fear49. Additionally, positive attitudes towards past influenza vaccines significantly increased the likelihood of COVID-19 vaccine acceptance46,50,61. Higher levels of perceived vaccine harm, concerns about side-effects and vaccine efficacy significantly contributed to a reduced vaccine acceptance in four out of four studies53,58,61,64. One survey reported a significant increase in the likelihood of vaccine acceptances if individuals perceived the vaccine to reduce the risk of COVID-19 infection (OR 3.1, 95%CI 2.1 to 4.8, p < 0.001)53.
Males were significantly more willing to receive the COVID-19 vaccine than females in all seven studies investigating this variable46,50,53,58,60,62,64. One survey reported that males were almost twice as likely as females to receive the COVID-19 vaccine (OR 1.9, 95%CI 1.5–2.3, p < 0.001)49. A subgroup analysis by gender across the seven studies reporting gender proportions revealed a similar trend; the pooled proportion willing to vaccinate for males was 71.9% (95%CI 59.4–83.0%) and 58.0% (95%CI 37.1–77.4%) for females, but this was not statistically significant (p = 0.247, Fig. 3)46,50,53,58,60,62,64. Similarly, females were consistently recorded as more likely to be vaccine hesitant than their male counterparts45,46,56 with an Australian survey recording females as almost twice as likely to be vaccine hesitant than males (Relative Risk Ratio (RRR) = 2.0, 95%CI 1.5 to 2.6, p < 0.001)46.
BAME individuals reported lower vaccination intentions than White individuals in all four studies that assessed acceptance by ethnicity47,50,56,58. Specifically, individuals of Black ethnicity were reported to be less accepting than White ethnic individuals in both studies investigating specific ethnicities50,58 and less accepting than both Hispanic and White ethnic individuals in one survey58. One study reported Black individuals to be up to 6.4 times more likely to be either hesitant or resistant (RRR 6.4, 95%CI 3.2 to 13.0, no p-value reported) than their White counterparts50.
Individuals with a lower household income were significantly less willing to receive the vaccine in three out of four studies47,54,58. One study reported that lower income households were over two times more likely to reject the vaccine than higher income households (OR 2.1, 95%CI 1.3–3.3, p < 0.001)47. However, one survey appeared to contradict this trend, suggesting that individuals in the lowest income band were significantly more likely to express vaccine hesitancy than rejection compared to individuals in higher income brackets56.
In all three studies investigating education, lower education was associated with lower vaccine acceptance45,50,60. In one study, the risk of individuals with no high school diploma rejecting and/or hesitating over the vaccine was almost eight times higher than those with a diploma or higher (RRR 7.8, 95%CI 3.1 to 19.6, no p-value reported)50.
Four out of seven studies reported younger individuals to be less vaccine willing49,54,61,62. However, there were substantial variations in the age groupings used by included studies. Two studies reported individuals aged < 30 years49 and < 35 years old62 to be the least willing age group to receive the vaccine (OR 1.5, 95%CI 1.3–1.9, p < 0.001 and OR 1.2, 95%CI 1.1–1.5, p < 0.001 respectively). One study opposed this trend, reporting individuals aged 35–44 years as most likely to reject (OR 3.3, 95%CI 1.2–9.5, p < 0.05)56. Another conflicting study drew conclusions from proportions alone, suggesting individuals aged < 35 years old were more vaccine willing than those in older age groups44.
Time of Survey
Across the 11 studies adopting a large sample size (n ≥ 1,000), the proportion reporting vaccine acceptance reduced significantly over time46,47,49,52,53,55,58,60,61,62,64. The nine surveys conducted between March-June had a pooled mean proportion of 76.8% survey participants reporting vaccine acceptance (n = 9, 95%CI 68.5–84.1%, p = 0.0)46,47,49,52,53,55,58,62,64 compared to 39.1% survey participants reporting vaccine acceptance (n = 2, 95%CI 39.1–40.5%, p = 0.0) across the two studies conducted between July-December (Fig. 4)60,61. Smaller studies (n < 1,000) were more likely to increase the heterogeneity of the results, so following the example of Robinson et al.24, the authors chose to restrict the subgroup analysis to larger studies which may have had more robust estimates of vaccine acceptance.
A subgroup analysis assessing study methodology used for recruitment reveals that survey participants recruited via probability sampling50,54,55,60,62 were significantly less willing to receive the COVID-19 vaccine than survey participants recruited via non-probability sampling43,46–49,52,53,57,58,61,63−65 (p=0.029, Fig.5), 55.6% (95%CI 34.0–76.1%) compared to 79.3% (95%CI 73.0–85.1%) respectively.
Reasons for Vaccine Hesitancy
Concern over vaccine safety was the most common reason reported for both vaccine hesitancy and rejection cited in all six studies investigating vaccine reasoning47,50,53,54,58,63. Three studies explicitly stated that fears of potential side-effects were the main cause for concern53,54,58. Other reasons include concern over vaccine efficacy53,58, speed of vaccine production and lack of evidence47,62, a lack of trust in both scientific and governmental bodies65, and general anti-vaccination attitudes47,62. For all studies investigating reasons for vaccine willingness, the main justification for vaccine acceptance was for the protection of both the individual and others47,53,63.