This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on http://publichealth.jmir.org, as well as this copyright and license information must be included.
There is a large presence of provaccination and antivaccination content on the Internet. The Internet has been identified as an important source for parents to seek and share vaccine information. There are concerns that parental fears or hesitancy on childhood immunizations are increasing due to the popularity of social media and exposure to online antivaccination sentiment. No other studies have investigated the association between seeking vaccine information online and Canadian parents’ perception of risk on childhood immunization.
We aimed to investigate the potential association between seeking vaccine information on the Internet and Canadian parents’ perception of risk on childhood immunization in order to quantify the perceived association and increase our understanding on the impact of the Internet to help guide public health interventions.
We analyzed this association in two population samples: a self-selecting Web-based sample of Canadian parents recruited through Facebook (n=966) and a population-based sample of parents recruited by random digit dialing (RDD; n=951). The outcome was parental perception of vaccine safety on a seven-point ordinal scale from “not safe” to “extremely safe.” An ordinal regression model was used to investigate if Internet information seeking on childhood vaccination predicted parental perception of vaccine safety.
After adjusting for income level, Internet reliability, age of parent, and region, the odds of perceiving vaccines as less safe rather than more safe were 1.6 times higher (95% CI 1.3-2.1) for parents who used the Internet to search for vaccination information compared to parents who did not search the Internet in the Web-based sample, and 2.0 times higher (95% CI 1.6-2.5) in the population-based RDD sample.
The results suggest the Internet is significantly associated with Canadian parents’ negative perception of vaccine risk. Governmental and scientific sectors should consider the development and implementation of Web-based vaccine interventions to promote confidence in immunization.
A decrease in public confidence in the safety of vaccines and subsequent lower vaccine uptake has been described as an “impending crisis” in the developed world [
The popularity of social media has been identified by the public health community as one of the reasons for the increase in parental fears about childhood vaccines because the Internet is an important vehicle for individuals seeking health information and support, and sharing health knowledge, opinions, and experiences [
The Health Belief Model is widely applied to determine what factors influence individuals when making vaccination decisions. In terms of immunization, the decision to vaccinate is balanced by the perceived risk of contracting a vaccine-preventable disease and the perceived risk of vaccine adverse events.
Conceptual model on the association between using the Internet to search for information on vaccinations and parental perception on safety of vaccinations.
Due to the abundance and availability of antivaccination sentiment online and the relatively low prevalence of vaccine-preventable disease in the population, it is suggested that individuals may perceive a greater risk of suffering from vaccination side effects than from contracting a vaccine-preventable disease [
We examined the potential association between seeking vaccine information on the Internet and Canadian parents’ perception of risk on childhood immunization data on two different data sources: primary data collected via Web-based survey and secondary data collected via population-based random digit dialing (RDD). We used two independent data sources with the same variables to test the association in two Canadian parent populations recruited at different times and via different methods. Both the Web-based and RDD survey contained questions on respondent demographics and knowledge, awareness, attitudes, and behaviors related to immunization. Identical questions to the RDD survey were used to measure the exposure, main outcome, and confounders in the primary data collection via Web-based survey, with the question format slightly altered for Web-based delivery.
We collected the Web-based survey data via targeted advertisement recruitment on Canada’s most popular social media platform, Facebook. French and English advertisements invited Canadian parents to click on the advertisement and participate in a Web-based survey on childhood immunization with a chance to win an iPad mini. Based on sample size calculations to detect an odds ratio of 1.5 and available budget, we aimed to recruit 800 participants. An odds ratio of 1.5 was used for two reasons: (1) to ensure sufficient sample size should the exposure be mildly but statistically associated with the outcome [
The population-based RDD data are secondary data deidentified and extracted from a survey collected by a reputable research company, EKOS Research Associates, contracted by PHAC. Experts in immunization and epidemiology at PHAC worked with the research company in the development and testing of the questionnaire. The objective was to collect descriptive data on Canadian parents’ knowledge, awareness, attitudes, and behaviors related to immunization to inform policy makers. The secondary data were collected via telephone survey on a population-based RDD sample of Canadian parents during a period of 3 weeks in March 2011. Respondent inclusion criteria were (1) 18 years of age or older, (2) parents of at least one child younger than 18 years, (3) resident of Canada, and (4) able to respond to questions in English or French. The research company compiled a summary report available online [
Researchers calculated the response rate based on the empirical method (completed + ineligible) / (unresolved + ineligible + nonresponding eligible + completed + nonresponding unknown) and reported a rate of 23·43% (7898/33,698) resulting in a sample size of 1745 Canadian parents [
Ethical approval was obtained from the University of Toronto’s Office of Research Ethics (REF#29309).
We classified respondents who sought out information on childhood vaccines and reported the Internet as one of their top three sources used for information on vaccines as “used the Internet” and those who do not seek out information on childhood vaccines or do not report the Internet as one of their top three sources as “did not use the Internet.”
We measured respondent perception on vaccine safety as an ordinal variable from 1 to 7: 1=not at all safe, 4=moderately safe, and 7=extremely safe.
We hypothesized parental education level and income, parental age and sex, age of youngest child, number of children, place or residence, and the relative importance of the Internet as a source of information relative to the importance of family, friends, and/or a health care professional as potential confounders. We measured education level according to four levels: high school or less, trade or vocational school, some university, and bachelors/graduate degree/professional certification. We measured household income level in Can $10,000 increments ranging from less than Can $30,000 to Can $120,000 and we categorized the variable into four levels (less than Can $30,000, Can $30,000-$59,999, Can $60,000-$99,000, and more than Can $100,000) in order for sufficient sample size in each category and to make comparisons among intermediary groups from lowest to highest income. We measured parental age as continuous (years) in the Web-based survey and it was measured as a categorical variable in the RDD survey (younger than 30 years, 30-34 years, 35-39 years, 40-44 years, and 45 years and older). The age of youngest child was measured as continuous (years) in both surveys, and the number of children was measured as categorical (1, 2, 3, 4, 5, and 6 or more) in the Web-based survey and as continuous in the RDD survey. We classified the perceived reliability of the Internet relative to family, friends, or health care professionals as (1) “reported as most reliable and trustworthy source on vaccines” to (4) “not reported in respondent’s top three choices as a reliable source of information on vaccines.” We categorized place of residence into six regions due to low numbers and to reflect the regions reported in the RDD data: British Columbia, Alberta, Saskatchewan or Manitoba, Ontario, Québec, and Atlantic provinces or Territories.
We excluded participants with missing data from the analyses as sufficient power remained and differences were not detected on the primary independent and dependent variables [
Both samples had similar education and income level distributions with almost half of the respondents following the education distribution of Canadian adults by completing some level of higher education [
The majority of Web-based respondents (68.77%, 751/1092) were younger than 35 years, female (92.61%, 1003/1083), and reported two or fewer children (81.49%, 894/1097). In the population-based RDD survey, the majority of respondents (62.29%, 674/1082) were 40 years or older and the mean age of their youngest child was 8.3 (SE 0.1) years. In addition, 41.02% (711/1745) were male and the median number of children per respondent was 2 (IQR 1.0).
For both data sources, approximately one-quarter of the respondents reported the Internet to be a reliable source for information on vaccines or vaccination, and approximately 40% (39.10%, 427/1092 vs 41.57%, 716/1729) reported using the Internet to search for information on vaccines. In terms of perception on safety of childhood immunizations, 26.77% (292/1091) of the Web-based survey respondents and 18.74% (324/1729) of the RDD survey respondents reported childhood immunizations as not at all safe to moderately safe (
Complete data were available for 966 respondents. The variables sex of parent and age of youngest child were removed from the multivariable analysis due to nonsignificance in bivariate analyses. Multicollinearity was not present and no interaction terms were retained due to nonsignificance of the likelihood ratio test between the model with all possible covariates and two-way interaction terms and the reduced model without interaction terms. Thus, ordinal logistic regression was performed with the following full model: Internet use, education level, income level, age of parent, age of youngest child, region, and reliability of the Internet. Nonsignificant variables (education level, number of children, and income level) were tested for potential confounding with only income level being retained in the model due to a significant change (26%) in the predictor’s estimate compared to the full model excluding education level and number of children. Originally excluded variables (sex of parent and age of youngest child) were individually re-entered into the model and were not found to be significant confounders. The covariates income level, Internet reliability, age of parent, and regions of residence remained in the final model (
Complete data were available for 951 RDD respondents. The variables sex of parent and income level were removed from the multivariable analysis due to nonsignificance in bivariate analyses. Multicollinearity was not present and all interactions terms were removed from the model. No interaction terms were retained due to nonsignificance of the likelihood ratio test between the model with all possible covariates and two-way interaction terms and the reduced model without interaction terms. Thus, ordinal logistic regression was performed with the following full model: education level, age group of parent, age of youngest child, number of kids, region, and reliability of the Internet. Nonsignificant variables in the full model (education level, number of children, age of youngest child, and age group of parent) were tested for potential confounding with only age group of parent being retained in the model due to a significant change (21.7%) in the predictor’s estimate compared to the full model excluding education level, number of children, and age of youngest child. All originally excluded variables (sex of parent and income level) were individually re-entered into the reduced model to check for confounding, and income level was then retained in the final model due to a significant change (16%) of the predictor’s estimate (
Characteristics of both study samples for continuous and categorical variables.
Population-based RDD survey (n=1745)a | |||
Age of parent (years), mean (SE) | 32.24 (6.69) | — | |
Age of youngest child (years), mean (SE) | 2.50 (3.78) | 8.31 (0.14) | |
Number of children, mean (SE) | — | 1.84 (0.02) | |
<30 | 395 (36.17) | 57 (5.07) | |
30-34 | 356 (32.60) | 129 (11.97) | |
35-39 | 189 (17.31) | 222 (19.66) | |
40-44 | 96 (8.79) | 244 (22.61) | |
≥45 | 56 (5.13) | 430 (40.69) | |
Missing, n | 5 | 663 | |
1 | 492 (44.85) | — | |
2 | 402 (36.65) | — | |
3 | 147 (13.40) | — | |
4 | 44 (4.01) | — | |
5 | 5 (0.46) | — | |
≥6 | 7 (0.64) | — | |
Male | 80 (7.39) | 711 (41.02) | |
Female | 1003 (92.61) | 1034 (58.98) | |
Missing, n | 14 | — | |
Education level, n (%) | |||
High school or less | 172 (16.09) | 358 (20.41) | |
Trade or vocational | 286 (26.75) | 514 (29.63) | |
Some university | 110 (10.29) | 144 (7.75) | |
Bachelor’s or graduate degree or professional certification | 501 (46.87) | 722 (42.19) | |
Missing, n | 28 | 7 | |
<$30,000 | 85 (8.6) | 157 (9.89) | |
$30,000-$59,999 | 170 (17.3) | 385 (24.52) | |
$60,000-$99,999 | 351 (35.6) | 498 (32.09) | |
≥$100,000 | 379 (38.5) | 519 (33.50) | |
Missing, n | 112 | 186 | |
British Colombia | 160 (14.61) | 175 (10.41) | |
Alberta | 259 (23.65) | 200 (10.17) | |
Saskatchewan and Manitoba | 137 (12.51) | 197 (6.50) | |
Ontario | 336 (30.68) | 486 (38.19) | |
Québec | 120 (10.96) | 427 (24.26) | |
Atlantic/Territories | 83 (7.58) | 260 (10.47) | |
Used the Internet | 427 (39.10) | 716 (41.57) | |
Did not use the Internet | 665 (60.90) | 1013 (58.43) | |
Missing, n | 5 | 16 | |
1 (Not at all safe) | 49 (4.49) | 43 (2.49) | |
2 | 48 (4.40) | 24 (1.39) | |
3 | 64 (5.87) | 50 (2.89) | |
4 (Moderately safe) | 131 (12.01) | 207 (11.97) | |
5 | 134 (12.28) | 275 (15.90) | |
6 | 338 (30.98) | 500 (28.92) | |
7 (Extremely safe) | 327 (29.97) | 630 (36.44) | |
Missing, n | 6 | 16 | |
Most reliable | 64 (5.88) | 149 (8.80) | |
Second most reliable | 97 (8.91) | 282 (16.82) | |
Third most reliable | 123 (11.29) | 30 (1.78) | |
Not in top three choices | 805 (73.92) | 1247 (72.60) | |
Missing, n | 8 | 37 |
aRDD: random digit dialing. Percentages for the population-based RDD survey are weighted.
Perception of risk of childhood immunizations in parents who used the Internet to search for information on immunizations (Web-based survey data: n=1086; RDD data: n=1713).
Adjusted cumulative odds ratios of proportional odds logistic regression analysis for the association between parental Internet use to search for information on immunizations and parental perception on safety of childhood immunizations.
Population-based RDDa survey (n=951) |
|||||
Use of the Internet | 1.61 (1.25-2.09) | 1.99 (1.55-2.54) | |||
Did not use the Internet | 1.00 Reference | 1.00 Reference | |||
<$30,000 | 1.42 (0.91-2.21) | 1.60 (1.03-2.48) | |||
$30,000 to $59,999 | 1.67 (1.20-2.33) | 1.19 (0.86-1.63) | |||
$60,000 to $99,999 | 1.23 (0.94-1.62) | 1.10 (0.82-1.47) | |||
≥$100,000 | 1.00 Reference | 1.00 Reference | |||
Most reliable | 4.77 (2.88-7.91) | 2.18 (1.41-3.36) | |||
Second most reliable | 3.96 (2.58-6.07) | 1.12 (0.81-1.57) | |||
Third most reliable | 1.12 (0.78-1.62) | 1.66 (0.61-4.50) | |||
Not in top three choices | 1.00 Reference | 1.00 Reference | |||
Age of parent (continuous) | 0.98 (0.96-0.99) | — | |||
<30 | — | 1.71 (0.98-2.98) | |||
30-34 | — | 0.99 (0.68-1.45) | |||
35-39 | — | 1.20 (0.87-1.67) | |||
40-44 | — | 1.16 (0.86-1.57) | |||
≥45 | — | 1.0 Reference | |||
British Colombia | 0.93 (0.65-1.33) | 1.63 (1.04-2.57) | |||
Alberta | 0.77 (0.56-1.06) | 1.38 (0.89-2.15) | |||
Saskatchewan and Manitoba | 0.64 (0.43-0.95) | 1.71 (1.13-2.59) | |||
Ontario | 1.00 Reference | 1.00 Reference | |||
Québec | 1.89 (1.27-2.83) | 1.26 (0.89-1.78) | |||
Atlantic/Territories | 1.00 (0.63-1.60) | 1.08 (0.75-1.56) |
aRDD: random digit dialing.
Although the Internet has been reported as an important influence on parental perception of risk on childhood immunizations, to our knowledge no study has quantified the association between seeking vaccine information on the Internet and perception on safety of childhood immunizations among Canadian parents. The analyses on both datasets resulted in the same conclusion with similar effect sizes not significantly different from one another. The findings from both data sources confirm the assumed relationship between looking for vaccine information on the Internet and perception of risk on vaccine safety, with both samples revealing higher odds of perceiving vaccines as “not safe” in parents who used the Internet to search for information on vaccines compared to parents who did not use the Internet for vaccine information. These results are consistent with a before-and-after Internet experiment study conducted in Germany where participants exposed to short searches on vaccine critical websites reported an increase perceived risk of vaccinating [
This study utilized two different data sources on Canadian parents, sampled at different times. The RDD data were collected in March 2011 and the Web-based data were collected between December 2013 and January 2014, thus the results represent a specific period in time. To our knowledge, there have been no significant policy changes from 2011 to 2014 and although several measles outbreaks have occurred since early 2011, both populations would have been exposed to the media coverage. Respondents were also asked about factors influencing vaccination decisions and there was no significant difference in time-related contextual influences reported. Furthermore, we received similar results in both samples, thus the bias introduced by time-varying contextual influences is likely nondifferential.
Due to incomplete and unreliable data, our study could not account for the reliability of the websites parents searched or in the type of communications they were exposed to on the Internet. For example, many Web-based respondents reported using search engines and clicking on the websites from their search results, as opposed to identifying specific websites or types of websites. According to the summary report by the research company who conducted the RDD survey, “Google search engine” was the primary website reported to be used by almost half of the respondents who searched for vaccination information online, followed by various government websites and other websites such as medical sites (eg, WebMD), online chat rooms, wikis, etc [
As more people abandon landline telephones, the validity of traditional population telephone surveys is compromised with low response rates and potentially nonrepresentative samples. Representativeness and validity concerns are also relevant for Web-based surveys as research relies on the collection of self-reported data by self-selected online participants [
Current initiatives aiming to reach and influence parents’ decision to vaccinate have not adequately abated the influence of the online antivaccination movement. Health agencies currently have an online presence; however, they have been slow to fully adopt the true nature of social media platforms and communication remains mostly by top-down dissemination of information [
measles-mumps-rubella
Public Health Agency of Canada
random digit dialing
The authors wish to thank the Public Health Agency of Canada (PHAC) for providing the sample RDD data. We would also like to thank the Canadian parents who participated in this study. Funding was provided by Public Health Ontario, Toronto, ON, Canada, and the Dalla Lana School of Public Health, Toronto, ON, Canada.
None declared.