We used a 2-arm parallel randomized pre-post design. Participants in a web-based survey were randomly allocated in a 1:1 ratio to the video-based intervention or the active control arm consisting of a text-based intervention. The study was designed to detect a significant pre-post increase in COVID-19 vaccine intentions in the video intervention group and the superiority of the video intervention compared to the text intervention in eliciting pro-COVID-19 vaccine intentions. We used the Consolidated Standards of Reporting Trials (CONSORT) statement to report the results [32].

Participants from all Canadian provinces or territories who met following eligibility criteria were enrolled in the study: (1) not vaccinated for COVID-19, (2) age range of 20-39 years, (3) Canadian resident, and (4) willing to complete the survey in either English or French. To ensure a balanced participation in the study and informed by the Canadian Census data, we used quota sampling for the primary language spoken at home (80% Anglophones, 20% Francophones); biological sex (50% males, 50% females), annual total income before taxes of all members of the household before the pandemic (50% more than CA $75,000 [US $58,563.80], 50% less than CA $75,000), and population density (80% urban, 20% rural). During data collection (July 30-September 13, 2021), the daily incidence of COVID-19 was rising, signaling the emergence of the fourth pandemic wave in Canada that reached its peak mid-September, when about 4300 new daily cases were reported nationwide. In this period, about one-third of daily cases were reported in Canadians aged 20-39 years and the estimated daily COVID-19 incidence in this age group reached 1500 (35% of total daily cases) [33]. In Canada, our target population became eligible for COVID-19 vaccination in April-May 2021, although provincial rollout varied widely. Therefore, as of April 17, 2021, the national cumulative percentage of individuals aged 20-39 years who received at least 1 COVID-19 vaccine dose was only about 9%. Vaccine uptake increased sharply in the upcoming months and the cumulative percentage of individuals in this age group who received at least one dose reached about 62% by June 5th, 2021. During data collection, the estimated national vaccine coverage (at least 1 dose) in individuals aged 20-39 years increased from about 72% at the start to 78% [34]. In this period that corresponded with the beginning of the academic year, extensive public health interventions (eg, messages distributed through media) aiming at increasing vaccine uptake were ongoing and vaccination mandates were beginning to be implemented in some jurisdictions (eg, Quebec).

Data collection was carried out by Dynata, an international online market research company with experience in programing surveys and collecting data for universities and companies in various fields (eg, public health, politics). Dynata used a combination of recruitment methods (eg, its own website, direct emails, ads on social media) to recruit participants. At the beginning of the survey, we checked whether participants’ electronic device (the survey could be completed on a smartphone, computer, or tablet) had adequate video and sound capabilities to complete the survey. After providing electronic consent, participants deemed eligible to participate were randomly allocated to 1 of the 16 strata based on the 4 quota sampling criteria (ie, primary language, biological sex, income, and population density; see Multimedia Appendix 1 for details). Within each stratum, a random concept picker approach was used to ensure a 1:1 allocation. Correspondingly, the first participant of a pair was randomly allocated to the intervention or the control arm and the second participant to the opposite arm. If a participant did not finish the survey (incomplete data), that place in the pair was allocated to the next participant. Thus, the quota in each stratum was filled in pairs and ensured a balanced group allocation throughout the data collection period.

After randomization, participants completed the remaining baseline sociodemographic questionnaire and provided their intentions to receive a COVID-19 vaccine. Then, they participated in the intervention (watched a short video eliciting altruism motives) or read a text related to general hygiene and preventive measures (active control group). All participants were prompted that attention check questions would follow. Those who did not correctly identify the names of the video characters were offered the possibility to watch the video a second time. Those who decided to watch the video again but still answered incorrectly were terminated. The video could be paused but not skipped or muted. Participants could not continue the survey until the video had been played entirely. In the active control arm, the sequence of information sections was randomized (to control for bias attributable to presentation order) and participants could neither skip sections nor progress to the next section until 10 seconds had elapsed to encourage careful reading. After each section, participants answered an attention check question asking them to identify a measure that was not mentioned in the section they had just read. Participants who answered all 3 attention check questions incorrectly were terminated.

Immediately after completing the intervention, we reassessed their intentions to receive a COVID-19 vaccine. Subsequently, participants answered additional questions (offered after the second assessment of vaccine intentions to avoid response bias), which included flu vaccination status, health care professional status, smoking history, and measures of altruism, empathy, and psychological distress. Only participants who provided complete survey data were retained in the final database. Participants were compensated by Dynata according to the reward system in which points are earned that can be later redeemed for company rewards (eg, Amazon, Starbucks).

To calculate the required sample size for the within-participant change in vaccine hesitancy (ie, pre- to postintervention), we used survey data that showed that in January/February 2021, approximately 40% of Canadians aged 20-39 years were hesitant toward a COVID-19 vaccine (ie, don’t know yet or would refuse vaccination) [55,56]. Estimating a 5% decrease in hesitancy in the intervention group and a correlation of about 0.4 between paired observations, the intervention group required a sample size of 907 pairs for detecting a 5% change in marginal proportions at a power of 80% and 2-sided significance of 5% [57]. To detect a 5% superiority of the video intervention in increasing vaccine intentions compared to the active control group at a power of 80%, we estimated a required sample per group of about 1300 participants. Considering a 1:1 allocation, the total sample required for this study was approximately 2600 participants (2×1300=2600).

The study was approved by the Research Ethics Board of the Integrated Health and Social Services University Network for West-Central Montreal (CIUSSS West-Central Montreal; Project ID #2021-2732).

Of 14,298 participants in the target age group who accessed the invitation to participate, 11,853 (82.9%) were assessed for eligibility, of whom 2097 (17.7%) were eligible (n=9578, 80.8%, were excluded because they were already vaccinated against COVID-19; n=174, 1.5%, did not meet other inclusion criteria; and n=4, 0.03%, dropped out) and were randomly allocated to the study arms: 1654 (78.9%) completed the postintervention assessment, and 1373 (65.5%; ie, 686, 50%, and 687, 50%, in the video and text intervention arms, respectively) were included in the analyses. See Figure 1.

Data collection took place from July 30 to September 13, 2021. At about 5 weeks into data collection, daily recruitment significantly declined. The main barrier was the relative low proportion (about 22%) of eligible participants (ie, unvaccinated in the age group of 20-39 years). We conducted preliminary analyses using a total sample of 1346 participants (673, 50%, per group) and found that the number of observations ensured 80% power to detect a 5% pre-post change in vaccine intentions. Preliminary analyses showed a difference of about 2% as opposed to the expected between-group difference of 5% in vaccine intentions that we had anticipated. To reach a similar level of power would have required about 5500 participants per group (ie, an increase of 4200 from our initial sample calculations) to detect a statistically significant superiority of the video intervention. Reaching the new sample size target would not have been feasible due to time and budget considerations, and we decided to stop data collection.

The sample consisted of slightly more females (n=740, 53.9%), the mean age was 30.7 years, the majority used English as the primary language at home (n=1122, 81.7%), most reported a total gross household income in the year preceding the pandemic of less than CA $75,000 (US $58563.80, n=848, 61.8%), and most resided in an urban area (n=1067, 77.7%). None of the sociodemographic characteristics differed significantly between the study groups (see Table 1 and Multimedia Appendix 3 for additional subgroup analyses). In the video group, 86 (12.5%) intended to receive the vaccine, 292 (42.6%) were decided against vaccination, 234 (34.1%) were undecided, and 74 (10.8%) had not thought about receiving the COVID-19 vaccine (ie, unengaged). Participants allocated to the active control group (text intervention) reported similar vaccine intentions, and the difference between groups was not statistically significant: χ²₃=1.62, P=.65; see Table 2.

In the video group, 43 (6.3%) participants changed from nonintenders at baseline (ie, unengaged, undecided, or decided not) to vaccine intenders (ie, decided to) postintervention and 10 (1.5%) participants changed from vaccine intenders at baseline to nonintenders postintervention. The McNemar test was significant (χ²₁=20.55, P<.001). In the active control (text) group, 24 (3.5%) participants changed from nonintenders at baseline to vaccine intenders postintervention and 5 (0.7%) participants changed from vaccine intenders at baseline to nonintenders postintervention. Unexpectedly, the McNemar test was also significant (χ²₁=12.45, P<.001).

In the video group, we found a statistically significant change from decided not at baseline to undecided postintervention (n=28, 4.1%; P=.02, OR 2.8, Cohen g=.24), from undecided to decided to (n=29, 4.2%; P<.001, OR 5.8, Cohen g=.35), and from unengaged to decided to (n=10, 1.5%; P=.03, OR 10, Cohen g=.41). In total, in the video group, 67 significant changes toward increased vaccination intentions were observed (see Figure 2 for a visual representation of PAPM stage transitions from baseline to postintervention in the video group and Tables S1 and S2 in Multimedia Appendix 4).

In the text group, we found a statistically significant change from unengaged at baseline to decided not postintervention (denoting a change toward decreased vaccine intentions; n=14, 2%; P=.02, OR 7, Cohen g=.38) and from undecided to decided to (n=16, 2.3%; P=.01, OR 8, Cohen g=.39). In other words, in the text group, 14 (2%) participants moved toward decreased intentions and 16 (2.3%) participants moved toward increased vaccination intentions (see Figure 3 for a visual representation of PAPM stage transitions from baseline to postintervention in the text group and Tables S1 and S3 in Multimedia Appendix 4). We found a significant difference between those who changed toward increased vaccine intentions in the video group (n=67, 9.77%) compared to the text group (n=16, 2.33%): χ²₁=33.43, P<.001.

Postintervention, in the video group, 119 (17.3%) intended to receive the vaccine, 277 (40.4%) were decided against vaccination, 236 (34.4%) were undecided, and 54 (7.9%) reported being unengaged. In the text group, 106 (15.4%) intended to receive the vaccine, 285 (41.5%) were decided against vaccination, 249 (36.2%) were undecided, and 47 (6.8%) reported being unengaged. The between-group difference in vaccine intentions was not significant: χ²₃=1.70, P=.64.

Results of additional subgroup analyses did not significantly differ from per protocol analyses (see Multimedia Appendix 5). The only difference consisted in the loss of statistical significance of the transition from unengaged to decided to in the video group (see Table S2 in Multimedia Appendix 5) that could be explained by 2 additional participants who transitioned from decided to at baseline to unengaged postintervention. Since one cannot change from decided to get the vaccine to unengaged, this was an artifact introduced by careless responding.

Results of exploratory analyses provided a signal that the video intervention was superior to the text intervention as the between-group difference in vaccine intentions was significant when using preintervention intentions in the text group and postintervention intentions in the video group (χ²₁=5.90, P=.02) and not significant when using preintervention intentions in the video group and postintervention intentions in the text group (χ²₁=2.39, P=.12); see Tables S10 and S11 in Multimedia Appendix 5. The same results were obtained using samples comprising 1654 (all completers of the second vaccination intention assessment; see Tables S5 and S6 in Multimedia Appendix 5) and 2089 (intention-to-treat) participants (see Tables S8 and S9 in Multimedia Appendix 5).

To the best of our knowledge, this is the first population-based study that examined the effect of a video-based intervention eliciting prosocial (altruistic) motives on intentions to receive the COVID-19 vaccine in younger Canadian adults. We used a pre-post and randomized control trial (RCT) study design and recruited a national sample of unvaccinated 20-39-year-old Canadians who participated in a web-based survey between July and September 2021 in the context of the fourth COVID-19 pandemic wave. Our study had 2 specific objectives: (1) to estimate pre- to postintervention vaccine intention changes in participants who were randomly allocated to the video intervention or the text-based intervention that provided non-vaccine-related preventive health measures and (2) to estimate between-group vaccine intentions postintervention.

First, we found that the video intervention was effective in changing vaccine intentions and that 4.8% more participants intended to receive the vaccine postintervention. The size of the effect is consistent with results of the experimental study conducted by Li et al [28], who studied 3952 participants (median age range 31-40 years) from 8 countries (China, France, Japan, United Kingdom, United States, Israel, Brazil, and South Africa) who participated in an internet survey in 2013 before the start of the flu season at the time. They reported a 6% absolute increase in intentions to receive the influenza vaccine in participants who were exposed to prosocial (altruism) messages [28]. Understanding the evolving context in which our study was conducted could explain the modest (4.8%) increase in vaccine intentions. At the time of data collection (July 30-September 13, 2021), about 3 months had elapsed since adults 20-39 years old became eligible to receive the COVID-19 vaccine in Canada. Three-quarters of them had received at least 1 dose [34]. In surveys conducted before the start of vaccination, approximately 40% of our target population was vaccine hesitant compared to 87% who reported vaccine hesitancy in our analyzed sample who are more resistant to vaccination. Therefore, it is possible that had this study been conducted 2 months earlier, our results would have shown a higher increase in pre- to postintervention vaccine intentions. Surprisingly, vaccine intentions also significantly increased in the group that received information about nonvaccine preventive measures in text format, although the effect was smaller than that in the video group, as only 2.7% reported higher vaccine intentions postintervention. Because we used a vaccine-neutral intervention in the active control group, it is possible that the increase represents social desirability. Since we did not measure social desirability, it is possible this bias was also present in the video intervention group as the video depicted vaccination as a social benefit.

Using the theoretical PAPM to inform the measurement of vaccine intentions, we found a more nuanced understanding of pre- to postintervention change in vaccine intentions. Our results show that significantly more participants who watched the video changed toward a more advanced vaccine decision stage than participants in the text group. In both groups, we found that individuals who had not thought about receiving the vaccine (unengaged) and those who were undecided were more likely to change their intentions to decided to vaccinate compared to those who reported being decided not at baseline, and this effect was more pronounced in the video group. This pattern of decision-making changes aligns with our previous findings from a longitudinal study evaluating human papillomavirus (HPV) vaccine intention change over a 9-month period in parents of 9-16-year-old boys and girls [22]. In that study, we demonstrated that parents who were unengaged or undecided at baseline were more likely to increase their HPV vaccine acceptability over time and deemed “flexible hesitant” (ie, changed to decided to vaccinate or vaccinated their child). This was in contradistinction with parents who were initially in the decided not stage and remained decided not over time, whom we deemed as “rigid hesitant” [22]. Therefore, investigating vaccine hesitancy as a binary outcome does not convey the nuances of movement in vaccine intention stages. For individuals who are “flexible hesitant,” viewing messages that highlight altruism may provide the necessary “push” to move toward adoption stages of accepting the vaccine. This could reflect behavioral nudging, in which promoting the positive impacts of a behavior without changing incentives or forbidding negative options can have a substantial impact on the behavior [61,62]. A recent systematic review by Reñosa et al [63] found that nudging messages that invoked emotional affect, such as storytelling and dramatic narratives, can improve vaccine confidence and uptake. In addition, Wood and Schulman (2021) [64] suggested that apathy toward vaccination, a characteristic that might contribute to someone being unengaged, could be addressed with peripheral, emotional messaging to motivate behavior change. Interestingly, in the video group, significantly more people moved from decided not to undecided, suggesting that the evocation of concern for others (altruism) may prompt even “rigid” hesitant individuals to reflect and rethink their decision.

Although pre-post analyses showed that the video intervention was effective in increasing vaccine intentions, between-group analyses did not confirm our hypothesis that watching the video would result in statistically significant higher intentions compared to reading non-vaccine-related information. Two factors may have contributed to this outcome: (1) The unexpected 2.7% increase in vaccine intentions in the active control group that reduced the hypothesized 5% between-group difference, and (2) the higher-than-expected vaccine hesitancy in our sample (which comprised ~40% “rigid hesitant” compared to ~10% found in 2 population-based studies conducted by our team that investigated HPV vaccine hesitancy [21,65]) that could have attenuated the effect of the video on vaccine intentions because “rigid hesitant” are less amenable to changes in intentions.

Although achieving statistical significance for the between-group difference would have sent a strong signal related to the efficacy of the video intervention, we believe that our study can inform future research using interventions that elicit prosocial motives to increase COVID-19 vaccine intentions. For example, interventions could be adapted to include other forms of prosocial motivations, such as collectivism (the practice of prioritizing a group over individuals within the group) [66]. Previous research has shown that collectivism is associated with COVID-19 vaccine acceptance [12,67], while individualism (ie, emphasis on the autonomous individual) is associated with COVID-19 vaccine hesitancy [68]. Therefore, to override feelings of personal invulnerability to COVID-19 in countries that are more individualistic than collectivistic (eg, Canada, the United States), messages that promote community well-being, highlight shared goals, and induce feelings of interdependence should be used to encourage COVID-19 vaccination [69]. Importantly, the design of our intervention aligns well with the recommendations for animated, video-based health communication interventions published by Adam et al [70] in 2021. Our intervention used a narrative approach, was well adapted to the Canadian cultural context as it was available in English and French, used characters of different ages and ethnic backgrounds, used appealing colors that ensured an optimal contrast independent of the size of the screen, included the voice of a narrator with experience in media communications, and had a length aligned with the recommend optimal length of around 2.5 minutes [70].

The main limitations derive from the premature termination of the study dictated by barriers in participant recruitment and by lower-than-anticipated COVID-19 vaccine hesitancy in the population of interest. As the target sample size was not reached, the sampling quotas used to match Canadian Census data deviated from the planned quotas and we included 3.9% more females, 2.3% more participants residing in rural areas, 5.2% less Francophones, and 11.8% less participants with annual total income before taxes of all members of the household before the pandemic of CA $75,000 (US $58,563.80). Although between-group differences were not significant, these differences in sociodemographics could impede the generalizability of the results to the Canadian population. The high proportion of participants who were in the decided not to vaccinate stage could have diminished our ability to prove the superiority of the video intervention in increasing vaccine intentions. Additionally, the use of an active control group could have diminished our capacity to prove the statistical superiority of the video intervention, perhaps due to social desirability. Finally, follow-up 3-6 months later would have allowed us to evaluate the translation of increased vaccine intentions into actual vaccine uptake.

Using a web-survey and a pre-post and RCT study design, we showed that a brief video eliciting prosocial (altruism) motives increased COVID-19 vaccine intentions of Canadians aged 20-39 years, especially among those who were less engaged in the decision to vaccinate or were undecided. As web streaming is highly popular among younger adults, using short videos is an efficient modality to disseminate public health messages. The effect of the new intervention on increasing intentions was modest, but delivering messages that elicit prosocial motives to vaccinate to a large population could increase vaccine intentions in a significant number of individuals and assist in reaching vaccination targets and curbing the effect of the pandemic. As vaccine hesitancy is complex, it is likely that a multifaceted messaging approach that includes the benefits of vaccination for the community would be beneficial, especially in societies where individual values prevail over collective values. Our intervention could be adapted to align with the latest COVID-19 immunization recommendations (eg, boosters) or to increase vaccine intentions for other preventable diseases.