Alexandre de FigueiredopersName, Clarissa SimaspersName, Emilie KarafillakispersName, Pauline PatersonpersName, Heidi J. LarsonpersName (2020date). “Mapping global trends in vaccine confidence and investigating barriers to vaccine uptake: a large-scale retrospective temporal modelling study.” The Lancettitle·venue.
Object of study — VACCINE CONFIDENCEterm as a globally comparable quantity - 290 surveysterm·mth, 149 countries, 284,381 individuals, pooled into one time series
Research question — whether public confidence in the importance, safety and effectiveness of vaccines can be tracked across countries and over time, and which component predicts uptake
Contribution — estimates of vaccine confidenceterm for 149 countries between 2015date and 2019date, and the finding that confidence in IMPORTANCE - not safety or effectiveness - carries the strongest association with uptake
phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | confidence in vaccines rising and falling by country on a timescale of years, and doing so unevenly across its three components — “Between November, 2015, and December, 2019, we estimate that confidence in the importance, safety, and effectiveness of vaccines fell in Afghanistan, Indonesia, Pakistan, the Philippines, and South Korea. We found significant increases in respondents strongly disagreeing that vaccines are safe between 2015 and 2019 in six countries: Afghanistan, Azerbaijan, Indonesia, Nigeria, Pakistan, and Serbia.” | |
| claim-type | existence — “Between November, 2015, and December, 2019, we estimate that confidence in the importance, safety, and effectiveness of vaccines fell in Afghanistan, Indonesia, Pakistan, the Philippines, and South Korea. We found significant increases in respondents strongly disagreeing that vaccines are safe between 2015 and 2019 in six countries: Afghanistan, Azerbaijan, Indonesia, Nigeria, Pakistan, and Serbia.” | |
| method — evidence | — “In this large-scale retrospective data-driven analysis, we examined global trends in vaccine confidence using data from 290 surveys done between September, 2015, and December, 2019, across 149 countries, and including 284 381 individuals.” | |
| object of study — evidence | — “In this large-scale retrospective data-driven analysis, we examined global trends in vaccine confidence using data from 290 surveys done between September, 2015, and December, 2019, across 149 countries, and including 284 381 individuals.” | |
| research question — evidence | — “Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidence.” | |
| contribution — evidence | — “Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptake compared with other determinants considered.” | |
| mechanism | the study exists because THE THING IT MEASURES HAS NO ADMINISTRATIVE RECORD: `Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidence`. Coverage is counted; confidence has to be surveyed — “Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidence.” | mechanisms[0] |
| mechanism | and the component that predicts uptake is not the one the public debate is about: `Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptake` — “Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptake compared with other determinants considered.” | mechanisms[1] |
| extends | the Vaccine Confidence Project's own decade of surveys, focus groups and media analytics, of which this is the pooled quantitative product — “These findings have resulted in the development of a Vaccine Confidence Index (VCI) survey tool (first implemented in 2015 14 ) to measure individual perceptions on the safety, importance, effectiveness, and religious compatibility of vaccines. The VCI questionnaire has the primary focus of measuring confidence across multiple countries while being minimal, thus allowing ready integration into existing global surveys.” | claim_relations[0] |
| supports | WHO's naming of vaccine hesitancy as one of the top ten threats to global health in 2019 - the framing the study is written into — “In this context, WHO has named vaccine hesitancy as one of the top ten threats to global health in 2019 alongside climate change.” | claim_relations[1] |
| sense | vaccine confidence - HERE, TRUST IN THE IMPORTANCE, SAFETY AND EFFECTIVENESS OF VACCINES, plus religious compatibility. Four components, and the VCI is built to measure exactly those four `while being MINIMAL, thus allowing ready integration into existing global surveys`. THE INSTRUMENT'S SHORTNESS IS A DESIGN GOAL, not a compromise — “These findings have resulted in the development of a Vaccine Confidence Index (VCI) survey tool (first implemented in 2015 14 ) to measure individual perceptions on the safety, importance, effectiveness, and religious compatibility of vaccines. The VCI questionnaire has the primary focus of measuring confidence across multiple countries while being minimal, thus allowing ready integration into existing global surveys.” | senses[0] |
| sense | FIVE INSTRUMENTS FOR TWO CONSTRUCTS, IN ONE SENTENCE, JOINED BY `OR`: the VCI, the Parent Attitudes About Childhood Vaccines Survey, the Vaccination Confidence Scale, the 5-C scale, and THE SAGE VACCINE HESITANCY SCALE - which this corpus holds as larson2015. `used to measure confidence or hesitancy` — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | senses[1] |
| sense | THE THREE Cs HAVE BECOME FIVE AND ONE OF THE ORIGINALS IS GONE. macdonald2015's model was CONFIDENCE, COMPLACENCY, CONVENIENCE. The 5-C scale named here is `confidence, complacency, constraints, calculation, and collective responsibility`. Convenience is not in it, and nothing in this paper marks the substitution — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | senses[2] |
| sense | `conversely` - the word doing the work in `Factors modulating vaccine confidence and, conversely, hesitancy`. It treats the two as opposite ends of one axis, which is precisely what macdonald2015 argued they are not: confidence is one of three categories of hesitancy determinant, not its inverse — “Factors modulating vaccine confidence and, conversely, hesitancy include (but are not limited to) trust in health-care systems, providers, governance, information, and perceptions of vaccine importance, safety, and efficacy.” | senses[3] |
| control | harmonising 290 surveys with different response options onto their shared extreme categories — failed — “Having a common metric of confidence and a baseline for comparison is crucial to understanding these changing trends over time” — rules out nothing that has been demonstrated - the authors call it 'A KEY ASSUMPTION', say it 'needs testing for validation', and name its cost: it POOLS the middle and MASKS POTENTIALLY KEY INFORMATION — “First, as not all surveys used have consistent responses, we have made a key assumption that, presented with different options between the extreme categories of “strongly agree” and “strongly disagree” (which are consistent across all surveys), respondents with the strongest sentiment will fall into one of these extreme groups regardless of additional categories. While this approach probably allows meaningful comparison across surveys—although it needs testing for validation—it pools vaccination beliefs among those without the strongest beliefs, masking potentially key information.” | controls[0] |
| control | whether the association with uptake belongs to confidence as a whole or to one of its components — held — “Among a multiplicity of factors influencing vaccine decisions, 17 key drivers of public confidence in vaccines were identified as trust in the importance, safety, and effectiveness of vaccines, along with compatibility of vaccination with religious beliefs.” — rules out the undifferentiated reading - IMPORTANCE carries the association, 'rather than' safety or effectiveness, which are the components the public debate is about — “Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptake compared with other determinants considered.” | controls[1] |
| control | whether religion-specific effects could be estimated — failed — “Finally, owing to low case counts of respondents who have not had their children vaccinated ( appendix 2 ) and the varying religious groups across countries, religious groups were recoded into the largest and minority groups to extract results from our regression analysis.” — rules out nothing - low case counts forced recoding 'into the largest and minority groups', so the finding is about majority status and not about any religion — “Finally, owing to low case counts of respondents who have not had their children vaccinated ( appendix 2 ) and the varying religious groups across countries, religious groups were recoded into the largest and minority groups to extract results from our regression analysis.” | controls[2] |
| control | recoding every intermediate Likert response into a single middle category so only 'strongly agree' and 'strongly disagree' are compared across surveys — held — “This reclassification assumes that respondents who report the most confident (“strongly agree”) and least confident (“strongly disagree”) beliefs are unlikely to change their response regardless of whether they are presented with additional Likert categories such as “tend to agree”, “somewhat agree”, or “neither agree nor disagree” (or a combination thereof; appendix 1 p 4 ).” — rules out apparent trend differences that are really differences in how many response options a given survey offered — “Thus, to allow meaningful comparisons in trends between surveys, individual survey responses falling between “strongly agree” and “strongly disagree” were recoded into the new category “neither strongly agree nor strongly disagree”, with the “strongly agree” and “strongly disagree” responses remaining unchanged.” | controls[3] |
| control | recoding 'do not know' answers about a child's vaccination to 'no' — held — “If respondent has children: “…have any of your children ever received a vaccine that was supposed to prevent them from getting childhood diseases…?” — rules out inflation of reported uptake by counting uncertainty as vaccination — “If respondent has children: “…have any of your children ever received a vaccine that was supposed to prevent them from getting childhood diseases…?” | controls[4] |
| control | recoding religion WITHIN each country into largest, minority and refused rather than by denomination — held — “For each country, religion is recoded into the most frequently reported religion in a given country (largest), all other religious affiliations (minority), or refusal to answer (refused) Other Education Years in education grouped into <9 years, 9–15 years, and ≥16 years <9 years Science education: “Have you…learned about science at [school]?” — rules out treating the same denomination as majority in one country and minority in another - the variable becomes social position rather than creed — “For each country, religion is recoded into the most frequently reported religion in a given country (largest), all other religious affiliations (minority), or refusal to answer (refused) Other Education Years in education grouped into <9 years, 9–15 years, and ≥16 years <9 years Science education: “Have you…learned about science at [school]?” | controls[5] |
| weight | the Vaccine Confidence Project's own decade of surveys, focus groups and media analytics, of which this is the pooled quantitative product — 0.8 — “These findings have resulted in the development of a Vaccine Confidence Index (VCI) survey tool (first implemented in 2015 14 ) to measure individual perceptions on the safety, importance, effectiveness, and religious compatibility of vaccines. The VCI questionnaire has the primary focus of measuring confidence across multiple countries while being minimal, thus allowing ready integration into existing global surveys.” | claim_relations[0] |
| weight | WHO's naming of vaccine hesitancy as one of the top ten threats to global health in 2019 - the framing the study is written into — 0.7 — “In this context, WHO has named vaccine hesitancy as one of the top ten threats to global health in 2019 alongside climate change.” | claim_relations[1] |
| audit | vaccine uptake —operationalised-as→ parental recall — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the study 'relies on' parental recall being accurate; operationalised-as is loose. — “Second, the WGM survey data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recall being accurate, and patterns of recall error not varying substantially across countries.” | concept_edges[0] |
| outcome | control — failed — “First, as not all surveys used have consistent responses, we have made a key assumption that, presented with different options between the extreme categories of “strongly agree” and “strongly disagree” (which are consistent across all surveys), respondents with the strongest sentiment will fall into one of these extreme groups regardless of additional categories. While this approach probably allows meaningful comparison across surveys—although it needs testing for validation—it pools vaccination beliefs among those without the strongest beliefs, masking potentially key information.” — “Having a common metric of confidence and a baseline for comparison is crucial to understanding these changing trends over time” | controls[0] |
| outcome | control — held — “Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptake compared with other determinants considered.” — “Among a multiplicity of factors influencing vaccine decisions, 17 key drivers of public confidence in vaccines were identified as trust in the importance, safety, and effectiveness of vaccines, along with compatibility of vaccination with religious beliefs.” | controls[1] |
| outcome | control — failed — “Finally, owing to low case counts of respondents who have not had their children vaccinated ( appendix 2 ) and the varying religious groups across countries, religious groups were recoded into the largest and minority groups to extract results from our regression analysis.” | controls[2] |
| outcome | control — held — “Thus, to allow meaningful comparisons in trends between surveys, individual survey responses falling between “strongly agree” and “strongly disagree” were recoded into the new category “neither strongly agree nor strongly disagree”, with the “strongly agree” and “strongly disagree” responses remaining unchanged.” — “This reclassification assumes that respondents who report the most confident (“strongly agree”) and least confident (“strongly disagree”) beliefs are unlikely to change their response regardless of whether they are presented with additional Likert categories such as “tend to agree”, “somewhat agree”, or “neither agree nor disagree” (or a combination thereof; appendix 1 p 4 ).” | controls[3] |
| outcome | control — held — “If respondent has children: “…have any of your children ever received a vaccine that was supposed to prevent them from getting childhood diseases…?” | controls[4] |
| outcome | control — held — “For each country, religion is recoded into the most frequently reported religion in a given country (largest), all other religious affiliations (minority), or refusal to answer (refused) Other Education Years in education grouped into <9 years, 9–15 years, and ≥16 years <9 years Science education: “Have you…learned about science at [school]?” | controls[5] |
| limitation | THE COMPARABILITY OF 290 DIFFERENTLY-WORDED SURVEYS RESTS ON AN ASSUMPTION THE AUTHORS SAY IS UNTESTED: `we have made A KEY ASSUMPTION that, presented with different options between the extreme categories ... respondents with the strongest sentiment will fall into one of these extreme groups regardless of additional categories ... although it needs testing for validation - it POOLS vaccination beliefs among those without the strongest beliefs, masking potentially key information` — “First, as not all surveys used have consistent responses, we have made a key assumption that, presented with different options between the extreme categories of “strongly agree” and “strongly disagree” (which are consistent across all surveys), respondents with the strongest sentiment will fall into one of these extreme groups regardless of additional categories. While this approach probably allows meaningful comparison across surveys—although it needs testing for validation—it pools vaccination beliefs among those without the strongest beliefs, masking potentially key information.” | limitations[0] |
| limitation | the outcome is COARSER THAN THE QUESTION: uptake means only `whether a parent has had any of their children vaccinated against at least one childhood disease`, so no vaccine-specific or child-specific determinant can be investigated - and it relies on PARENTAL RECALL being accurate and its errors constant across countries — “Second, the WGM survey data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recall being accurate, and patterns of recall error not varying substantially across countries.” | limitations[1] |
| limitation | and religion, one of the four named drivers, was `recoded into the largest and minority groups` because of low case counts - so the religious finding is about majority-versus-minority status, not about any religion — “Finally, owing to low case counts of respondents who have not had their children vaccinated ( appendix 2 ) and the varying religious groups across countries, religious groups were recoded into the largest and minority groups to extract results from our regression analysis.” | limitations[2] |
| limitation | THE TERM THE PROJECT'S OWN DEFINITION PAPER REJECTED IS NOW THE HEADLINE CONSTRUCT, AND THE TWO ARE JOINED BY `OR`: five instruments are described as `used to measure confidence or hesitancy`. macdonald2015 - same research group - refused `confidence` as the name because it covers `ONLY ONE CATEGORY of factors`, and larson2015 has a numbered rule that refusal is not hesitancy — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | limitations[3] |
| limitation | the Likert options genuinely differed between surveys and only the two extreme categories were consistent across all of them — “Likert responses presented to respondents differed between surveys ( appendix 1 p 4 ); however, the outermost categories (“strongly agree” and “strongly disagree”) remained consistent across surveys.” | limitations[4] |
| limitation | the harmonisation rests on an assumption the paper states but does not test - that respondents at the extremes would not have answered differently under a finer scale — “This reclassification assumes that respondents who report the most confident (“strongly agree”) and least confident (“strongly disagree”) beliefs are unlikely to change their response regardless of whether they are presented with additional Likert categories such as “tend to agree”, “somewhat agree”, or “neither agree nor disagree” (or a combination thereof; appendix 1 p 4 ).” | limitations[5] |
| limitation | only three confidence statements could be used because the religious-compatibility item was not asked in the survey supplying half the data — “As data on perceptions about the religious compatibility of vaccines was not posed to respondents in the Wellcome Global Monitor, we only consider three statements on vaccine confidence. † For this question, vaccine-preventable diseases were given as examples to respondents, which varied by country (see appendix 1 pp 28–29 ).” | limitations[6] |
| limitation | three survey modes were mixed - online, telephone and face to face - across the same pooled estimate — “Online, telephone, and face-to-face survey methodologies were used (see appendix 1 pp 3–4 for further details on survey methodologies).” | limitations[7] |
| limitation | country coverage is deeply uneven in time: one country has six timepoints while 67 have only one — “One country was surveyed over six different timepoints (Philippines), 13 countries were surveyed over four timepoints, 28 over three timepoints, 40 over two timepoints, and 67 once.” | limitations[8] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | content_tags.findings[0] |
| finding | — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | content_tags.findings[1] |
| finding | — “First, as not all surveys used have consistent responses, we have made a key assumption that, presented with different options between the extreme categories of “strongly agree” and “strongly disagree” (which are consistent across all surveys), respondents with the strongest sentiment will fall into one of these extreme groups regardless of additional categories. While this approach probably allows meaningful comparison across surveys—although it needs testing for validation—it pools vaccination beliefs among those without the strongest beliefs, masking potentially key information.” | content_tags.findings[2] |
| finding | — “Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptake compared with other determinants considered.” | content_tags.findings[3] |
| finding | — “Between November, 2015, and December, 2019, we estimate that confidence in the importance, safety, and effectiveness of vaccines fell in Afghanistan, Indonesia, Pakistan, the Philippines, and South Korea. We found significant increases in respondents strongly disagreeing that vaccines are safe between 2015 and 2019 in six countries: Afghanistan, Azerbaijan, Indonesia, Nigeria, Pakistan, and Serbia.” | content_tags.findings[4] |
| finding | — “In this large-scale retrospective data-driven analysis, we examined global trends in vaccine confidence using data from 290 surveys done between September, 2015, and December, 2019, across 149 countries, and including 284 381 individuals.” | content_tags.findings[5] |
| finding | — “Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidence.” | content_tags.findings[6] |
| finding | — “Second, the WGM survey data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recall being accurate, and patterns of recall error not varying substantially across countries.” | content_tags.findings[7] |
| finding | — “Finally, owing to low case counts of respondents who have not had their children vaccinated ( appendix 2 ) and the varying religious groups across countries, religious groups were recoded into the largest and minority groups to extract results from our regression analysis.” | content_tags.findings[8] |
| finding | no comparable global estimates of vaccine confidence existed before this, which is why the relationship between coverage, demographics and confidence could not be quantified — “To date, no comparable global estimates and monitoring of vaccine confidence are available, prohibiting a quantitative understanding of the relationship between vaccine coverage, socioeconomic demographics, and confidence.” | content_tags.findings[9] |
| finding | the study's stated implication is that confidence needs REGULAR monitoring to catch changes early enough to act on — “A key implication of this analysis is the importance of regular monitoring of vaccine confidence levels to detect trends and changes that suggest the need for interventions to sustain confidence and pre-empt negative impacts on vaccination uptake.” | content_tags.findings[10] |
| quantity | surveys pooled, across 149 countries and 284,381 individuals — “In this large-scale retrospective data-driven analysis, we examined global trends in vaccine confidence using data from 290 surveys done between September, 2015, and December, 2019, across 149 countries, and including 284 381 individuals.” | content_tags.quantities[0] |
| quantity | individuals whose responses were harmonised onto shared extreme categories — “In this large-scale retrospective data-driven analysis, we examined global trends in vaccine confidence using data from 290 surveys done between September, 2015, and December, 2019, across 149 countries, and including 284 381 individuals.” | content_tags.quantities[1] |
| quantity | the pooled sample: individuals, countries and nationally representative surveys — “Between September, 2015, and December, 2019, reported vaccine confidence levels were collected from 284 381 individuals aged 18 years or older across 149 countries as part of 290 nationally representative surveys.” | content_tags.quantities[2] |
| quantity | how many countries were surveyed at how many timepoints — “One country was surveyed over six different timepoints (Philippines), 13 countries were surveyed over four timepoints, 28 over three timepoints, 40 over two timepoints, and 67 once.” | content_tags.quantities[3] |
| variable | religious beliefs — “Among a multiplicity of factors influencing vaccine decisions, 17 key drivers of public confidence in vaccines were identified as trust in the importance, safety, and effectiveness of vaccines, along with compatibility of vaccination with religious beliefs.” | content_tags.variables[0] |
| definition | vaccine confidence — “Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidence.” | content_tags.definitions[0] |
| definition | vaccine uptake — “Second, the WGM survey data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recall being accurate, and patterns of recall error not varying substantially across countries.” | content_tags.definitions[1] |
| definition | vaccine confidence — “Vaccine confidence was measured through three survey statements relating to individual perceptions on the importance, safety, and effectiveness of vaccines ( table ).” | content_tags.definitions[2] |
| concept | vaccine confidence — not defined in this paper's own words; trust in the importance, safety and effectiveness of vaccines, plus compatibility with religious beliefs — “Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidence.” | concepts[0] |
| concept | Vaccine Confidence Index — not defined in this paper's own words; a deliberately minimal survey tool measuring perceptions of safety, importance, effectiveness and religious compatibility, built to slot into existing global surveys — “These findings have resulted in the development of a Vaccine Confidence Index (VCI) survey tool (first implemented in 2015 14 ) to measure individual perceptions on the safety, importance, effectiveness, and religious compatibility of vaccines. The VCI questionnaire has the primary focus of measuring confidence across multiple countries while being minimal, thus allowing ready integration into existing global surveys.” | concepts[1] |
| concept | SAGE Vaccine Hesitancy Scale — not defined in this paper's own words; one of the five instruments named here - and the one this corpus holds, as larson2015 — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | concepts[2] |
| concept | 5-C scale — not defined in this paper's own words; confidence, complacency, constraints, calculation and collective responsibility - macdonald2015's three Cs grown to five, with CONVENIENCE dropped — “The VCI survey is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Survey, which measures vaccine hesitancy among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scale (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scale, which has been deployed across multiple countries.” | concepts[3] |
| concept | vaccine hesitancy — not defined in this paper's own words; named by WHO as one of the top ten threats to global health in 2019, and treated here as the converse of confidence — “In this context, WHO has named vaccine hesitancy as one of the top ten threats to global health in 2019 alongside climate change.” | concepts[4] |
| concept | parental recall — not defined in this paper's own words; the sole source of the uptake variable - and the paper's own caveat is that it must be accurate AND that its error must not vary across countries — “Second, the WGM survey data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recall being accurate, and patterns of recall error not varying substantially across countries.” | concepts[5] |
| concept | vaccine uptake — not defined in this paper's own words; operationally, whether a parent has had ANY of their children vaccinated against AT LEAST ONE childhood disease - by parental recall — “Second, the WGM survey data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recall being accurate, and patterns of recall error not varying substantially across countries.” | concepts[6] |
| concept | source of trust — not defined in this paper's own words; WHO the respondent trusts for health advice - doctor, family, government, traditional healer - measured as a determinant alongside income and education — “Source of trust “[Who] do you trust most to give you medical or health advice?” | concepts[7] |
| concept | information seeking — not defined in this paper's own words; a joint variable requiring BOTH having sought health information recently AND wanting to know more - high only if yes to both — “A joint information-seeking behaviour variable is defined with responses “high” (if “yes” answered to both statements) and “low” (otherwise) Low Demographics and socioeconomic status Sex Male or female Female Age Integer-valued age, scaled to have a mean of 0 and unit SD No baseline Income quintile Quintiles: Q1 (lowest income) to Q5 (highest income) Q1 Religion: “Could you tell me what your religion is?” | concepts[8] |
| concept | Wellcome Global Monitor — not defined in this paper's own words; the survey supplying half the data and the only one carrying trust, information-seeking and uptake items - which is why those analyses rest on a subset — “Of the 290 surveys, 144 were collected as part of the 2018 Wellcome Global Monitor (WGM; appendix 2)” | concepts[9] |
Full text — tagged
The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.
Abstract. Summary Background There is growing evidence of vaccine delays or refusals due to a lack of trust in the importance, safety, or effectiveness of vaccines, alongside persisting access issues. Although immunisation coverage is reported administratively across the world, no similarly robust monitoring system exists for vaccine confidenceterm. In this study, vaccine confidenceterm was mapped across 149 countries between 2015date and 2019date. Methods In this large-scale retrospective data-driven analysis, we examined global trends in vaccine confidenceterm using data from 290 surveysterm·mth done between September, 2015date, and December, 2019date, across 149 countries, and including 284 381 individuals. We used a Bayesian multinomial logit Gaussian process model to produce estimates of public perceptions towards the safety, importance, and effectiveness of vaccines. Associations between vaccine uptaketerm and a large range of putative drivers of uptake, including vaccine confidenceterm, socioeconomic status, and sources of trust, were determined using univariate Bayesian logistic regressions. Gibbs sampling was used for Bayesian model inference, with 95% Bayesian highest posterior density intervals used to capture uncertainty. Findings Between November, 2015date, and December, 2019date, we estimate that confidence in the importance, safety, and effectiveness of vaccines fell in Afghanistan, Indonesia, Pakistan, the Philippines, and South Korea. We found significant increases in respondents strongly disagreeing that vaccines are safe between 2015date and 2019date in six countries: Afghanistan, Azerbaijan, Indonesia, Nigeria, Pakistan, and Serbia. We find signs that confidence has improved between 2018 and 2019date in some EU member states, including Finland, France, Ireland, and Italy, with recent losses detected in Poland. Confidence in the importance of vaccines (rather than in their safety or effectiveness) had the strongest univariate association with vaccine uptaketerm compared with other determinants considered. When a link was found between individuals' religious beliefsterm·var and uptake, findings indicated that minority religious groups tended to have lower probabilities of uptake. Interpretation To our knowledge, this is the largest study of global vaccine confidenceterm to date, allowing for cross-country comparisons and changes over time. Our findings highlight the importance of regular monitoring to detect emerging trends to prompt interventions to build and sustain vaccine confidenceterm. Funding European Commission, Wellcome, and Engineering and Physical Sciences Research Council. Introduction Alongside persisting issues of access to health services, waning vaccine confidenceterm has taken a toll on immunisation programmes across the globe, 1 , 2 , 3 , 4 contributing to stagnating or decreasing immunisation rates and consequent surges in vaccine-preventable diseases such as measles. 5 , 6 In this context, WHO has named vaccine hesitancyterm as one of the top ten threats to global health in 2019date alongside climate change. 7 Founded in 2010, the Vaccine Confidenceterm Project (VCP) was established to develop systematic approaches to monitoring public confidence in vaccines and to inform policy makers and stakeholders of the changing trends and determinants of vaccine confidenceterm across the globe. Over the past decade, the VCP has comprehensively explored the landscape of confidence issues and experiences in managing confidence crises around the world. 8 , 9 , 10 , 11 The VCP has conducted numerous surveysterm·mth, focus groups, in-depth qualitative research, and large-scale digital media analytics, 12 , 13 , 14 as well as convened expert roundtables and workshops to understand context-specific attitudes to vaccines among the general public, 14 , 15 health-care professionals and providers, 15 and pregnant women. 16 The VCP continues to research the roots, trends, and impacts of vaccine confidenceterm issues at national and supranational levels to inform policy and trust-building activities and mitigate the need for crisis management in immunisation programmes. Among a multiplicity of factors influencing vaccine decisions, 17 key drivers of public confidence in vaccines were identified as trust in the importance, safety, and effectiveness of vaccines, along with compatibility of vaccination with religious beliefsterm·var. 18 These findings have resulted in the development of a Vaccine Confidence Indexterm (VCI) surveyterm·mth tool (first implemented in 2015date 14 ) to measure individual perceptions on the safety, importance, effectiveness, and religious compatibility of vaccines. The VCI questionnaire has the primary focus of measuring confidence across multiple countries while being minimal, thus allowing ready integration into existing global surveysterm·mth. The VCI surveyterm·mth is one of a diverse set of metrics and indices used to measure confidence or hesitancy such as the Parent Attitudes About Childhood Vaccines Surveyterm·mth, which measures vaccine hesitancyterm among parents; 19 the Vaccination Confidence Scale, which measures confidence in adolescent vaccination; 20 the 5-C scaleterm (confidence, complacency, constraints, calculation, and collective responsibility), which identifies psychological barriers of vaccination behaviour; 21 and the SAGE Vaccine Hesitancy Scaleterm, which has been deployed across multiple countries. 22 , 23 , 24 , 25 , 26 Research in context Evidence before this study We have previously done three systematic reviews identifying the key determinants of vaccine hesitancyterm to inform questionnaire design around vaccine confidenceterm. Vaccine refusals and delays are contributing to an increasing number of vaccine-preventable disease outbreaks (eg, measles and polio) globally. For this reason, vaccine hesitancyterm was named by WHO as one of the top ten threats to global health in 2019date. Factors modulating vaccine confidenceterm and, conversely, hesitancy include (but are not limited to) trust in health-care systems, providers, governance, information, and perceptions of vaccine importance, safety, and efficacy. The Vaccine Confidenceterm Project (VCP) was founded 10 years ago to establish a systematic approach to monitoring public confidence in vaccines. In 2015date, the VCP developed the Vaccine Confidence Indexterm (VCI) surveyterm·mth: a tool used to monitor spatiotemporal trends in vaccine confidenceterm at national and global levels. Added value of this study We used new and existing VCI surveysterm·mth comprising nearly 300 000 individual responses from 149 countries around the world, allowing unprecedented insights into global time-varying trends in vaccine confidenceterm. Datasets previously collected in collaboration with ORB International (Gallup International), the European Commission, and the Philippines Surveyterm·mth and Research Centre were used alongside novel data from the Sahel and worldwide data collected up to the end of 2019date. We used Bayesian tools to estimate national-level vaccine confidenceterm trends and to investigate the link in each country between vaccine uptaketerm and socioeconomic and non-socioeconomic (eg, confidence in vaccine, trust in health system) determinants. To our knowledge, this is the largest study of global vaccine confidenceterm to date, using a common metric to allow cross-country comparisons of vaccine sentiment globally. Implications of all the available evidence This study provides novel insights into worldwide variations in vaccine confidenceterm and presents the country-dependent factors that modulate vaccine decisions. These include perceptions of vaccine safety, efficacy, and importance; socioeconomic and demographic determinants; and individually reported levels of trust. The study findings are discussed in light of past and ongoing vaccine confidenceterm issues in different settings. A key implication of this analysis is the importance of regular monitoring of vaccine confidenceterm levels to detect trends and changes that suggest the need for interventions to sustain confidence and pre-empt negative impacts on vaccination uptake. In this large-scale retrospective study, we explore global trends in vaccine confidenceterm between 2015date and 2019date. We combine previously published data from nearly a quarter of a million individual surveyterm·mth responses with 50 000 additional interviews from 2019date. To date, no comparable global estimates and monitoring of vaccine confidenceterm are available, prohibiting a quantitative understanding of the relationship between vaccine coverage, socioeconomic demographics, and confidence. This analysis aims to provide multiyear global-level estimates of vaccine confidenceterm for 149 countries worldwide, exploring trends in confidence and the global determinants of uptake including socioeconomic determinants and sources of trust. Methods Data sources Between September, 2015date, and December, 2019date, reported vaccine confidenceterm levels were collected from 284 381 individuals aged 18 years or older across 149 countries as part of 290 nationally representative surveysterm·mth. One country was surveyed over six different timepoints (Philippines), 13 countries were surveyed over four timepoints, 28 over three timepoints, 40 over two timepoints, and 67 once. We grouped countries and territories by WHO regional classification. We classified the territories of Hong Kong, Northern Cyprus, and the occupied Palestinian territory into the Western Pacific, European, and Eastern Mediterranean regions, respectively. Surveyterm·mth collection has been conducted through collaboration with ORB International (Gallup International), the European Commission, 15 the Philippines Surveyterm·mth and Research Center, 27 and Wellcome. 28 Vaccine confidenceterm was measured through three surveyterm·mth statements relating to individual perceptions on the importance, safety, and effectiveness of vaccines ( table ). Online, telephone, and face-to-face surveyterm·mth methodologies were used (see appendix 1 pp 3–4 for further details on surveyterm·mth methodologies). Responses to the three statements were answered on Likert scales ranging from “strongly disagree” to “strongly agree”. Likert responses presented to respondents differed between surveysterm·mth ( appendix 1 p 4 ); however, the outermost categories (“strongly agree” and “strongly disagree”) remained consistent across surveysterm·mth. Thus, to allow meaningful comparisons in trends between surveysterm·mth, individual surveyterm·mth responses falling between “strongly agree” and “strongly disagree” were recoded into the new category “neither strongly agree nor strongly disagree”, with the “strongly agree” and “strongly disagree” responses remaining unchanged. This reclassification assumes that respondents who report the most confident (“strongly agree”) and least confident (“strongly disagree”) beliefs are unlikely to change their response regardless of whether they are presented with additional Likert categories such as “tend to agree”, “somewhat agree”, or “neither agree nor disagree” (or a combination thereof; appendix 1 p 4 ). Table Data used throughout the study Responses Baseline for univariate Bayesian regressions Vaccine confidenceterm * “I think vaccines are safe” Likert scale recoded to “strongly disagree”, “strongly agree”, and “neither strongly agree nor strongly disagree” Not strongly agree “I think vaccines are important for children to have” As above As above “I think vaccines are effective” As above As above Vaccine uptaketerm † If respondent has children: “…have any of your children ever received a vaccine that was supposed to prevent them from getting childhood diseases…?” Yes, no, do not know; “do not know” responses are recoded to “no” No Source of trustterm “[Who] do you trust most to give you medical or health advice?” Family and friends (social circle), a doctor or nurse, other sources (famous people, traditional healers, or none) A doctor or nurse “[How much] do you trust medical and health advice from the government…?” A lot, some, not much, not at all; responses are recoded to “high” (a lot) and “low” (others) Low “[How much] do you trust medical and health advice from medical workers, such as doctors and nurses…?” A lot, some, not much, not at all; responses are recoded to “high” (a lot) and “low” (others) Low “How much do you trust…traditional healers…?” A lot, some, not much, not at all; responses are recoded to “high” (a lot) and “low” (others) Low Information seekingterm “Have you…tried to get any information about medicine, disease, or health in the past 30 days?” and “Would you…like to know more about medicine, disease, or health?” A joint information-seeking behaviour variable is defined with responses “high” (if “yes” answered to both statements) and “low” (otherwise) Low Demographics and socioeconomic status Sex Male or female Female Age Integer-valued age, scaled to have a mean of 0 and unit SD No baseline Income quintile Quintiles: Q1 (lowest income) to Q5 (highest income) Q1 Religion: “Could you tell me what your religion is?” For each country, religion is recoded into the most frequently reported religion in a given country (largest), all other religious affiliations (minority), or refusal to answer (refused) Other Education Years in education grouped into <9 years, 9–15 years, and ≥16 years <9 years Science education: “Have you…learned about science at [school]?” Primary, secondary, university, no, do not know; “primary” and “do not know” responses are recoded to “low”, with all others to “high” Low * As data on perceptions about the religious compatibility of vaccines was not posed to respondents in the Wellcome Global Monitorterm·mth, we only consider three statements on vaccine confidenceterm. † For this question, vaccine-preventable diseases were given as examples to respondents, which varied by country (see appendix 1 pp 28–29 ). Of the 290 surveysterm·mth, 144 were collected as part of the 2018 Wellcome Global Monitorterm·mth (WGM; appendix 2) . 28 In addition to probing individuals' perceptions on vaccine confidenceterm across the globe, the WGM also surveysterm·mth individuals on a range of factors including sources of trust, information-seeking behaviours, and whether respondents with children report having vaccinated at least one child against any routine immunisation programme (if the respondent reports having at least one child). To explore barriers to vaccine uptaketerm, we extracted data from the WGM surveysterm·mth on demographics (sex, age, and religious beliefsterm·var) and socioeconomic status (income and education, including science education), in addition to sources of trust and information-seeking behaviours. These variables are summarised in the table . Missing data by country are listed in appendix 1 (pp 25–26) . On average, each surveyterm·mth contained approximately 1000 individuals. The surveysterm·mth were weighted by sex and age according to national distributions, with equal sex representation in most surveysterm·mth ( appendix 2 ). Model-based estimates of vaccine confidenceterm The proportion of respondents falling into each of the three response categories (“strongly agree”, “strongly disagree”, and “neither strongly agree nor strongly disagree”) for each confidence statement and in each country at any given timepoint was modelled as multinomial logit Gaussian process model. 29 Model inference was done using Gibbs sampling and 10 000 samples (or draws) were obtained from the posterior predictive distributions, from which mean estimates were calculated (see appendix 1 pp 5–7 for full model details). Model performance was assessed using out-of-sample validation using five-fold cross-validation. Out-of-sample metrics indicated good model fit with a mean error of 10 −17 , mean absolute error of 10·10, and a root mean square error of 15·29. Our model was used to estimate vaccine confidenceterm across all surveyed countries at two timepoints (November, 2015date, and November, 2018) when many surveysterm·mth were conducted. To evaluate more recent temporal trends, changes in confidence estimates under our model are provided between November, 2015date, and December, 2019date. More recent changes in vaccine safety perceptions are evaluated for the EU, where a higher frequency of surveysterm·mth has been conducted, on average, compared with the rest of the world. A difference of posterior distributions of the probability of strongly agreeing (and strongly disagreeing) is calculated between each country's two most recent datapoints from 2018 onwards spaced at least 12 months apart (see appendix 2 for further details). A change in confidence is recorded if the 95% highest posterior density (HPD) interval of this difference distribution excludes zero. The same methodology is used to identify countries worldwide with an increase in the proportions of respondents strongly disagreeing that vaccines are safe, important, or effective between November, 2015date and December, 2019date. Vaccine uptaketerm determinants Bayesian logistic regressions were used to investigate the link in each country between vaccine uptaketerm and confidence, source of trustterm, information-seeking behaviour, and demographics and socioeconomic status ( table ). 30 This analysis serves to highlight consistent trends in uptake determinants across the globe and provides a platform for deeper analysis examining the most important factors of uptake in each country or developing predictive models of vaccine uptaketerm. For the vaccine uptaketerm analysis, we included respondents from the WGM dataset who reported having heard of vaccinations and having had children. The strength of the relationship between the percentage of respondents in each country strongly agreeing that vaccines are important, safe, and effective and the percentage of respondents reporting having had their children vaccinated was assessed using Pearson's correlation coefficient, where percentages p were first transformed onto the real line using −log(100/ p – 1). All models were implemented in JAGS 31 using R (version 3.6.1). 95% HPD intervals were used to represent confidence in parameter estimates. The 95% HPD interval is the smallest interval of the posterior distribution that contains 95% of the probability mass. Raw surveyterm·mth data for all 290 surveysterm·mth can be found in appendix 2 , alongside model-based vaccine confidenceterm estimates from November, 2015date, to January, 2020. Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Model-based estimates of the percentage of respondents strongly agreeing that vaccines are safe, important, and effective in 2015date and 2018 are shown in figure 1 . Estimates for all countries with associated uncertainties are shown in the appendices, alongside time-series plots showing temporal trends in vaccine confidenceterm ( appendix 1 pp 8–23 ; appendix 2 ). Figure 1 Global trends in perceptions towards the safety of vaccines in November, 2015date, and November, 2018 Figure shows model-based estimates of the percentage of respondents strongly agreeing that vaccines are safe (panels A, B), important for children to have (panels C, D), and effective (panels E, F) in November, 2015date, and November, 2018. No data were available for countries in grey. Argentina (89·4%, 95% HPD interval 87·7–91·3), Liberia (86·1%, 67·1–97·7), and Bangladesh (86·1%, 83·7–88·1) had the highest estimated percentage of respondents strongly agreeing that vaccines are safe in late 2015date, whereas Japan (8·9%, 7·4–10·6), France (8·9%, 7·2–10·5), and Mongolia (8·1%, 6·4–9·8) had the lowest ( figure 1A ). Ethiopia (96·3%, 95% HPD interval 95·2–97·3), Argentina (95·7%, 94·5–97·0), and Bangladesh (95·1%, 93·8–96·4) had the highest estimated percentage of respondents strongly agreeing that vaccines are important in 2015date, whereas Turkey (22·1%, 19·5–24·7), Morocco (15·8%, 13·7–18·4), and Georgia (2·7%, 1·6–3·8) had the lowest ( figure 1C ). Ethiopia (86·6%, 84·5–88·7), Argentina (86·3%, 84·3–88·4), and Mauritania (81·9%, 64·4–97·2) had the highest estimated percentage of respondents strongly agreeing that vaccines are effective in late 2015date, whereas Japan (14·7%, 12·5–16·7), Mongolia (13·0%, 11·0–15·2), and Morocco (10·3%, 8·6–12·2) had the lowest ( figure 1E ). Between November, 2015date, and December, 2019date, we estimate that vaccine confidenceterm fell for all three elements of confidence in Indonesia, the Philippines, Pakistan, and South Korea, and for two elements in Afghanistan and Vietnam ( figure 2 ). The Philippines, which ranked in the top ten countries worldwide in late 2015date for vaccine confidenceterm for all three elements, ranked no higher than 70th in 2019date, having experienced a large fall in the percentage of respondents strongly agreeing that vaccines are safe (absolute difference 23·1%, 95% HPD interval 6·3–40·0), important (23·0%, 7·4–37·4), and effective (23·7%, 8·5– 39·7; figure 2 ). Indonesia also witnessed large drops in confidence over this time in all three elements: vaccine safety (absolute difference 13·8%, 9·0–18·9), importance (14·6%, 9·9–19·5), and effectiveness (12·2%, 7·5–17·7; figure 2 ). Vaccine confidenceterm increased between 2015date and 2019date across all three elements for France, India, Mexico, Poland, Romania, and Thailand ( figure 2 ). Figure 2 Distributions in absolute confidence changes between November, 2015date, and December, 2019date Distributions of model-based estimates in the absolute differences in the proportions of respondents strongly agreeing that vaccines are safe, important, and effective. Positive values denote an increase in confidence between 2015date and 2019date. Owing to increased uncertainty around estimates further away from surveyterm·mth dates, some significant changes in confidence over the study period are not captured by this figure. Time-series trends in vaccine confidenceterm until the end of 2019date are shown for the EU (including the UK) and the Philippines ( figure 3 ). Recent losses in the percentage of respondents agreeing that vaccines are safe (between 2018 and late 2019date) are detected in Poland, with increases detected in Finland, France, Ireland, Italy, and the UK ( figure 3A ; appendix 2 ). Vaccine confidenceterm plummeted between 2015date to 2018 in the Philippines (over well documented fears around the Dengvaxia vaccine in 2017 27 , 32 , 33 ); however, since the start of 2018, confidence in the importance of vaccines has made substantial gains, with less substantial increases in vaccine safety and effectiveness perceptions ( figure 3B ). The loss of vaccine confidenceterm in the Philippines triggered by fears over Dengvaxia appears to also have affected uptake of routine vaccines recommended by the national immunisation programme ( figure 3C ). 27 This pattern is not limited to the Philippines: there were larger increases in the percentage of respondents perceiving vaccines to be important than safe or effective across the majority of countries in the European region for which there was an improvement in recorded importance confidence ( figure 2 ). Figure 3 Trends in the perceived safety of vaccines in the EU and the Philippines (A) Time series of estimated percentages of respondents in EU countries strongly agreeing, strongly disagreeing, or neither strongly agreeing nor strongly disagreeing that vaccines are safe. Lines are means and shaded regions are 95% HPD intervals. Circles show the observed percentage of respondents from raw data ( appendix 2 ). Time series for all countries for all three confidence statements are shown in appendix 1 (pp 12–23) . (B) Time series of surveyterm·mth responses across all three surveyterm·mth questions for the Philippines. (C) WHO-UNICEF national immunisation estimates for routine vaccination programmes in the Philippines against tuberculosis (BCG), diphtheria-pertussis-tetanus (DPT1), measles (MCV1), and polio (Pol3). HPD=highest posterior density. For ten countries worldwide, our model estimated a higher percentage of respondents strongly disagreeing that vaccines are safe, important, or effective in December, 2019date, compared with November, 2015date ( appendix 1 p 24 ): Afghanistan, Azerbaijan, Bosnia and Herzegovina, Georgia, Indonesia, Japan, Malaysia, Nigeria, Pakistan, and Serbia. Of note, Afghanistan, Azerbaijan, Indonesia, Nigeria, Pakistan, and Serbia all had increased concerns about vaccine safety ( appendix 1 p 24 ) Significant associations between uptake and vaccine confidenceterm, sources of trust, information-seeking behaviour, and demographics and socioeconomic status are shown in figure 4 , with consistent trends shown across the globe. Overall, the determinants most consistently associated with improved uptake were high confidence in vaccines (66 countries); trusting health-care workers more than family, friends, or other non-medical sources for medical and health advice (43 countries); higher levels of science education (35 countries); sex, with women more likely than men to report any child having at least one vaccine in 41 countries and men more likely than women in just one country (Chad); age (younger age groups associated with increased chances of uptake in 43 countries); and high information-seeking behaviour (18 countries). Income and religion were less widely associated with uptake; however, when a link was found between religion and uptake, it is the minority religious groups (or those refusing to provide their religious beliefterm·var) who were associated with lower probability of uptake ( figure 4 ). Figure 4 Univariate determinants of vaccine uptaketerm within the Wellcome Global Monitorterm·mth dataset Global trends in univariate associations between vaccine uptaketerm and confidence in vaccines, demographics and socioeconomic status, sources of trust, and information-seeking behaviours. Each point represents a significant association (95% HPD interval excludes zero) between a variable and uptake in a given country. Boxplots show the median log risk ratio and IQR. All variables (except age, which is continuous) are categorical and baseline groups are specified by each category (eg, high vs low denotes low as the baseline group; see the table for definitions). HPD=highest posterior density. Countries with higher percentages of respondents strongly agreeing that vaccines are safe, important, and effective had higher percentages of respondents reporting that they have had their children vaccinated (WGM data only; figure 5 ). These effect sizes are small but significant, with a Pearson's correlation of 0·28 (95% CI 0·12–0·42) between the percentage strongly agreeing that vaccines are safe and the percentage of respondents reporting vaccinating their children across all countries, 0·45 (0·31–0·57) for vaccine importance and uptake, and 0·28 (0·12–0·42) for vaccine effectiveness and uptake. Figure 5 Association between national level vaccine uptaketerm and vaccine confidenceterm as reported in the Wellcome Global Monitorterm·mth The solid line represents the mean regression with 70% and 95% HPD intervals shaded in dark and light grey, respectively. Datapoints for (lowest confidence) and the Philippines (case study) are shown, alongside randomly selected countries. HPD=highest posterior density. Discussion To our knowledge, this is the largest study of vaccine confidenceterm, offering important insights into the global state of vaccine confidenceterm. Vaccine confidenceterm was estimated and mapped for 149 countries from 2015date to 2019date, revealing key spatial and temporal trends. Although confidence remained low across Europe compared with other continents, there are signs that vaccine confidenceterm is increasing for much of Europe, including France, where vaccine confidenceterm has been persistently low since 2015date. Confidence increased overall in Poland between 2015date and 2019date, but a recent loss in confidence was seen between late 2018 and late 2019date, reflecting the growing impact of a highly organised local anti-vaccine movement. 34 Between 2015date and 2018, vaccine confidenceterm plummeted in the Philippines and Indonesia. In 2017, the vaccine manufacturer Sanofi announced that their newly introduced dengue vaccine Dengvaxia posed a risk to individuals who had not previously been exposed to the virus, prompting outrage and panic across the population where nearly 850 000 children had been given the new vaccine the previous year. As the VCP measured a baseline confidence value in 2015date, we were able to measure the change in confidence following the vaccine scare and found a significant drop in confidence in vaccine importance, safety, effectiveness. 27 The VCI surveyterm·mth tool has detected a rise in confidence across the country—although confidence is not back to 2015date levels—indicating a possible recovery and highlighting the value of the tool in assessing the effectiveness of national-level policy. Japan ranked among the countries with the lowest vaccine confidenceterm in the world: this might be linked to the human papillomavirus (HPV) vaccine safety scares that started in 2013, and following the decision by the Japanese Ministry of Health, Labour and Welfare in June, 2013, to suspend proactive recommendation of the HPV vaccine. 35 As a result of this vaccine safety scare, HPV vaccination coverage decreased from 68·4–74·0% in the 1994–98 birth cohort to 0·6% in the 2000 birth cohort. 36 The news of Japan suspending their proactive recommendation of the HPV vaccine has travelled globally through online media and social media networks, being applauded by anti-vaccination groups but not by the global scientific community. 4 The way in which the HPV vaccine scare was approached by health officials, as well as an ongoing outbreak of rubella in Japan, 37 , 38 indicate continuing issues with the Japanese vaccination programme that need resolving. 39 Indonesia witnessed a large drop in confidence between 2015date and 2019date, partly triggered by Muslim leaders questioning the safety of the measles, mumps, and rubella (MMR) vaccine, and ultimately issuing a fatwa —a religious ruling—claiming that the vaccine was haram and contained ingredients derived from pigs and thus not acceptable for Muslims. Local healers promoting natural alternatives to vaccines also contributed to the waning confidence in vaccines. 40 , 41 Results from our surveyterm·mth can inform the need for further research, to explore why certain countries might experience sudden increases or decreases in confidence. We have highlighted countries with marked decreases in percentages reporting that they strongly agree that vaccines are safe and countries with significant increases in those strongly disagreeing that vaccines are safe. These countries are candidates for more nuanced follow-up surveysterm·mth to understand the precise drivers of confidence and the link between confidence and uptake. In South Korea and Malaysia, online mobilisation against vaccines has been identified as a key barrier to vaccination. 42 , 43 The internet is a main source of vaccination information in Malaysia, where misinformation has been identified as influencing vaccine reluctance. 44 In South Korea, an online community named ANAKI (Korean abbreviation of “raising children without medication”) has been strongly advocating against childhood immunisation. 45 Future studies in both countries should further investigate this trend and propose mitigation strategies. In Georgia, unfounded vaccine safety concerns, amplified by the media, were found to profoundly affect a nationwide MMR vaccine campaign in 2008. 46 Our findings of low vaccine confidenceterm in Georgia could suggest that concerns about vaccine safety are again on the rise. The determinants of vaccine uptaketerm across the globe show strong consistency, with being male or having fewer years of education associated with decreased chances of uptake. Positive information-seeking behaviours and trusting health-care workers more than other sources such as one's social circle for medical and health advice were associated with increased chances of uptake. There are several study limitations to note. First, as not all surveysterm·mth used have consistent responses, we have made a key assumption that, presented with different options between the extreme categories of “strongly agree” and “strongly disagree” (which are consistent across all surveysterm·mth), respondents with the strongest sentiment will fall into one of these extreme groups regardless of additional categories. While this approach probably allows meaningful comparison across surveysterm·mth—although it needs testing for validation—it pools vaccination beliefs among those without the strongest beliefs, masking potentially key information. Second, the WGM surveyterm·mth data only permit an investigation of uptake defined as whether a parent has had any of their children vaccinated against at least one childhood disease. These uptake data are therefore not defined on a vaccine or child-by-child basis, precluding an investigation of determinants of vaccine-specific or child-specific uptake (which has been done recently in the literature 22 , 24 , 41 ). Moreover, we rely on parental recallterm being accurate, and patterns of recall error not varying substantially across countries. Finally, owing to low case counts of respondents who have not had their children vaccinated ( appendix 2 ) and the varying religious groups across countries, religious groups were recoded into the largest and minority groups to extract results from our regression analysis. In many settings, more nuanced regression findings are possible, and a comprehensive regression analysis could reveal more informative country-specific determinants of vaccine uptaketerm. Sentiments seeding doubt and distrust and the viral spread of misinformation are contributing to a landscape of uncertainty. Some actors have purposefully polarised vaccine debates, exploiting the doubting public and system weaknesses for political purposes, 47 , 48 while waning vaccine confidenceterm in other settings might be influenced by a wider environment of distrust in government and scientific elites. 14 The findings of declining confidence in Afghanistan, Azerbaijan, Pakistan, and Nigeria mirror trends in political instability and religious extremism in these settings. 49 Over the past few years in Pakistan and Nigeria, new waves of misinformation surrounding the polio vaccine have been circulating, 50 , 51 and have led to recent increases in poliovirus cases in both countries. 52 Further research should investigate the link between political polarisation, religious extremism, and populism and vaccination beliefs to better understand these complex ties. 48 Having a common metric of confidence and a baseline for comparison is crucial to understanding these changing trends over time, which can serve as an early warning system to prompt needed intervention to avert drops in vaccine confidenceterm and acceptance. In the context of new and emerging disease outbreaks, such as the COVID-19 pandemic, the VCI provides a valuable baseline of confidence levels to measure change in times of evolving disease threats and to help to identify where more trust building is needed to optimise uptake of new life-saving vaccines. Data sharing All raw data used as input to our multinomial Gaussian process model ( appendix 1 pp 5–6 ) can be found in appendix 2 .