Patterns and predictors of COVID-19 vaccination among young adults at 44 US sites: secondary analysis of a randomized, controlled, open-label trial, March – December 2021

Abstract

Background:

The introduction of vaccines during the COVID-19 pandemic provided an opportunity to slow transmission of SARS-CoV-2, but initial uptake of COVID-19 vaccination was slow. We analyzed data from a randomized clinical trial of the mRNA-1273 vaccine (NCT04811664) to describe the patterns of uptake of COVID-19 vaccines among young adults.

Methods:

The CoVPN 3006 trial randomized adults ages 18–29 from 44 sites in the United States to receive 1) immediate mRNA-1273 vaccination from the study site, or 2) standard of care, including the option to seek vaccination at any time in the future. Randomization occurred between March and November 2021, and an observational arm of adults who declined vaccination was enrolled beginning June 2021. Among participants in the standard of care (SoC) or Vaccine Declined arms, we estimated demographic, behavioral, and health history correlates of vaccination, and the four-month cumulative incidence of COVID-19 vaccination using inverse probability weighted Kaplan-Meier estimators.

Results:

Among 728 SoC and 470 Vaccine Declined participants, 79% and 16% received COVID-19 vaccination, respectively. SoC and Vaccine Declined participants were more likely to seek and receive vaccination if they reported COVID-19 preventive behaviors, including wearing masks, physically distancing, and avoiding large gatherings. We identified strong predictors of vaccination in the Vaccine Declined arm, including attending class in person (adjusted risk ratio [aRR]: 0.47, 95% confidence interval [CI] 0.21, 1.03), having a COVID-19 relevant medical condition (aRR: 1.95, 95% CI: 0.89, 4.26), and avoiding large gatherings (aRR: 2.24, 95% CI: 1.18, 4.25), though low vaccination rates in this arm led to imprecise estimates.

Conclusions:

Individuals who initially decline vaccination can be convinced to vaccinate, particularly if they are already practicing other forms of COVID-19 prevention. Continued outreach and education from the scientific community can combat low vaccine confidence.

INTRODUCTION

When the US Food and Drug Administration (FDA) issued emergency use authorizations (EUA) for mRNA COVID-19 vaccines in December 2020 [1], elderly and chronically ill populations were prioritized for vaccination. By May 2021, following increases in mRNA vaccine production and distribution and EUA of the Ad26.COV2.S [2] vaccine in February 2021, the White House authorized and promoted COVID-19 vaccination for all adults ages 18 years and older [3]. Since then, COVID-19 vaccines have received full FDA approval, have been authorized for use in children as young as 6 months of age, and have been redesigned to prevent infections with novel variants and sub-variants of the SARS-CoV-2 virus [4].

As of October 2023, 81% of eligible US citizens [5] have received at least one dose of a COVID-19 vaccine. However, COVID-19 vaccine uptake in the first year following EUA was initially slow among the general population of adults. So-called “deliberative factors” [6–9] might have caused adults to delay vaccination after it was universally recommended, including beliefs that the limited vaccine supply should be prioritized for vulnerable populations and concerns about the safety and effectiveness of the vaccines due to their relatively short time in testing, development, and on market. In addition, distrust of the scientific community, government agencies, and pharmaceutical companies influenced vaccine confidence among many populations [10]. Lower vaccination rates were especially pronounced among Black individuals [11, 12], who voiced historically justified fears of unethical research practices targeting Black communities, and young adults [8, 9, 12, 13], who may consider themselves at lower risk for severe complications from COVID-19 compared to older adults.

Now that COVID-19 is considered an ongoing global health concern, the emergency declaration has been withdrawn, and recommendations for routine booster vaccination are currently being considered [14], patterns of use and predictors of routine or seasonal COVID-19 vaccination remain to be seen. However, the lessons learned from the first year of COVID-19 vaccination under emergency circumstances can be leveraged to better prepare for future rollout of novel vaccines for emergency and routine use, and to identify and address potential barriers to vaccination at the start of vaccination campaigns to counter vaccine hesitancy. To help understand factors associated with vaccine uptake among young adults in the US, we analyzed data collected from a randomized, controlled, open-label trial of young adults who received the two-dose mRNA-1273 vaccination series immediately or at the end of a four-month follow-up period, beginning March 2021 [15]. The study protocol evolved along with COVID-19 vaccine authorization and national recommendations for adult vaccination through December 2021, including 1) encouraging participants who were not yet vaccinated to seek COVID-19 vaccination outside of the study in accordance with national guidance for adults; 2) enrolling young adults who reported not wanting the COVID-19 vaccine to serve as a comparator for the two groups who were randomized to receive COVID-19 vaccination through the study. In this secondary analysis of data from the parent trial, we report COVID-19 vaccination patterns in the first year of COVID-19 vaccine availability in the context of changing COVID-19 epidemiology and recommendations for adult vaccination, comparing individuals who self-identified as willing or unwilling to receive COVID-19 vaccination. These findings will support a better understanding of the correlates of novel vaccine uptake in preparation for future responses to outbreaks or emerging infections.

METHODS

Design of the parent trial

The COVID-19 Prevention Network (CoVPN) protocol 3006 (ClinicalTrials.gov #NCT04811664) was a randomized, controlled, open-label trial to assess the efficacy of the mRNA-1273 vaccine in preventing 1) primary SARS-CoV-2 infection, and 2) secondary transmission among young adults and their close contacts through daily nasal swabbing and symptom reporting over four months [15]. Because mRNA-1273 had already demonstrated safety and efficacy in clinical trials and it was not ethical to withhold vaccination from a control group, enrolled participants were initially randomized 1:1 to receive two doses of mRNA-1273 vaccine immediately on enrollment (Immediate Vaccination arm) or on conclusion of the four-month observation period (Delayed Vaccination arm), in accordance with the staged rollout strategy that prioritized older adults and individuals with certain chronic diseases. Randomization was stratified by study site and type of residence reported by participants to ensure balance between study arms within these strata. In March 2021, CoVPN 3006 began enrolling adults ages 18–26 years who were current undergraduate or graduate students at any of 27 university sites with no history of a known SARS-CoV-2 infection or COVID-19 vaccination. Additional academic and clinical sites joined the study through September 2021, resulting in 32 total study sites in 24 states (Figure S1).

Following the universal recommendation in May 2021 for all adults to receive COVID-19 vaccination, the CoVPN 3006 protocol was modified on May 20, 2021 in three ways. First, the inclusion criteria were expanded to include adults through 29 years of age and individuals who were not enrolled in academic institutions but lived in communities surrounding the original study sites. Next, the delayed Vaccination arm was renamed the Standard of Care (SoC) arm, and participants in this arm were encouraged to seek COVID-19 vaccination from a pharmacy or health care facility outside of the study (“off-study vaccination”). Finally, an observational group of individuals who were not interested in COVID-19 vaccination (Vaccine Declined group) was enrolled for comparison with vaccinated or vaccine-accepting participants. Vaccine Declined participants were also encouraged to seek vaccination at non-study sites but were not offered any vaccinations from the supply allocated for study participants. All participants from the SoC and Vaccine Declined groups were asked to report their off-study vaccinations via mobile app and verbal confirmation with study staff, including the vaccine manufacturer and date(s) of vaccination. As a result of rapid uptake of COVID-19 vaccination in the community and declining study enrollment, the study was closed early on December 3, 2021. At study end, all participants in the SoC and Vaccine Declined groups who had not yet received vaccination were offered vaccination at that time by their respective study sites.

At enrollment we collected self-reported data on demographic characteristics, living arrangements, health status, and SARS-CoV-2 risk and preventive behaviors. Each week thereafter, participants self-reported SARS-CoV-2 risk and preventive behaviors, potential SARS-CoV-2 exposures, and COVID-19 symptoms through a mobile app. Participants self-collected nasal swabs daily and returned them to study sites in person or by mail for molecular testing for SARS-CoV-2 infection; these results are reported elsewhere [15]. While the main trial publication reports on participants who were SARS-CoV-2 seronegative at baseline, our study includes all participants regardless of baseline serostatus.

Measures of interest for the secondary analysis

Primary outcome

As participants in the Immediate Vaccination arm were vaccinated on the day of enrollment, the primary outcome of our secondary analysis was time to off-study COVID-19 vaccination among participants in the SoC and Vaccine Declined groups only, defined as the number of days from enrollment to the self-reported date of receipt of a first dose of any COVID-19 vaccine received outside of the study. Participants were censored at the earliest occurrence of 1) four months of follow-up; 2) study closeout on December 3, 2021; 3) the date of receiving the first dose of study COVID-19 vaccine; or 4) loss to follow-up. Vaccinations from the study supply were offered to SoC participants who completed follow-up as part of the study protocol, and to Vaccine Declined participants at the end of the study to minimize vaccine waste; only the former were counted as on-study vaccination. Vaccination cards were requested to confirm self-reported vaccination dates, but were not obtained from all participants.

Primary exposures

The primary exposures of interest were captured at enrollment, including self-reported race (Black/African American, Asian, Multiple races, or White), self-reported ethnicity (Hispanic/Latinx or not Hispanic/Latinx), in-person class attendance (yes/no; restricted to students), and having one or more medical condition known to increase the risk or severity of COVID-19 infection (yes/no) [16]. Conditions considered in this group were: high blood pressure, heart conditions, autoimmune disease, immunodeficiency, HIV/AIDS, cancer, seasonal allergies, lung conditions, current smoker, blood disorder, neurological conditions, diabetes, chronic kidney disease, chronic liver disease, and being overweight or obese (body mass index >25). We also examined four self-reported COVID-19 prevention behaviors in the two weeks prior to enrollment: always/mostly wearing a mask inside around other people (yes/no), always/mostly physically distancing (yes/no), avoiding gathering in groups ≥10 people (yes/no), and encountering people outside your home who always/mostly wore a mask (yes/no). Additional variables of interest included age (in years), gender, enrollment date, in-person work status (in person vs. remote), and student status (yes/no).

Statistical analysis

Because the different selection processes that gave rise to the two study arms led to major differences in sociodemographic profiles and were plausibly related to future vaccine uptake, we stratified all analyses by study group (SoC versus Vaccine Declined). Participant baseline characteristics and receipt of COVID-19 vaccination were summarized using descriptive statistics. We estimated the association between several hypothesized predictors of vaccination (specified above) and the time to COVID-19 vaccination. Weighted Kaplan-Meier estimators with robust standard errors were used to estimate the four-month cumulative incidence of off-study COVID-19 vaccination and the corresponding risk ratios (RRs). We used the delta method to calculate the standard error of the log of the cumulative incidence ratio and generated 95% confidence intervals (CIs) for each hypothesized predictor. Inverse probability of treatment weights (IPTWs) were used to adjust for measured baseline confounders identified from directed acyclic graphs (DAGs). For each predictor of interest, we created a separate DAG illustrating the unique relationship between said predictor, subsequent vaccination, and factors associated with them to identify potential confounders. Potential confounders were identified a priori. Each predictor of vaccination was assessed in a separate model with a unique covariate adjustment set based on its unique DAG: 1) ethnicity was adjusted for race; 2) preventive behaviors were adjusted for age, enrollment date, ethnicity, gender, student status, in-person work, medical condition, and race; 3) attending class in-person was adjusted for enrollment date; 4) COVID-19 relevant medical condition was adjusted for age, ethnicity, gender, and race. The model for race was not adjusted for any other variables. Multiple imputation by chained equations (MICE) was used to impute missing covariate data (n=112; missing data on medical conditions [n=58], race [n=37], ethnicity [n=5], gender [n=2], age [n=1], COVID-19 prevention behaviors [n=7], or date of first COVID-19 vaccine [n=2]), IPTWs were generated for each of 15 imputed datasets, weighted Kaplan-Meier curves were estimated for each dataset, and the results were pooled using standard ‘Rubin’s rules’ to generate the final weighted estimates [17]. All statistical tests were 2-sided with an alpha of 0.05, and all analyses were conducted using R version 4.3.1 (R Foundation for Statistical Computing). No adjustments were made for multiple comparisons.

RESULTS

Demographic characteristics

CoVPN 3006 enrolled 1,923 participants into the three study arms, of which 1,198 (62%) were either in the SoC (n=728) or Vaccine Declined (N=470) group and eligible for this analysis (Figure 1). Enrollment of Immediate Vaccination and SoC participants began in March 2021, and enrollment of Vaccine Declined participants began in June 2021; enrollment for all study arms ended in November 2021 (Figure 2).

Figure 1.

Study flow diagram of participants in the CoVPN 3006 Study

Figure 2.

Participant enrollment in the CoVPN 3006 Study in 2021, stratified by study arm (n=1,198)

Participants were a median of 21 years of age (interquartile range [IQR]: 20,23 in the SoC group and 22 years of age (IQR: 20,26) in the Vaccine Declined group (Table 1). Half of SoC participants (50%) and Vaccine Declined participants (54%) identified as women; participants in both study groups were primarily White (SoC: 62%; Vaccine Declined: 65%), non-Hispanic (SoC: 79%; Vaccine Declined: 74%), and lived in off-campus housing (SoC: 80%; Vaccine Declined: 88%) (Table 1). Participants in both study groups shared communal household spaces with a median of two other people (IQR: 1,3). Most participants (SoC: 66%; Vaccine Declined: 74%) attended a job, volunteered, or studied on campus, and most participants (SoC: 68%; Vaccine Declined: 79%) ate food inside of a restaurant, bar, or cafeteria in the two weeks prior to enrollment. Participants in the SoC group were more likely to report masking indoors (SoC: 70%; Vaccine Declined: 31%), physically distancing themselves from other people (SoC: 65%; Vaccine Declined: 35%) most or all the time, avoiding gatherings of 10 or more people (SoC: 41%; Vaccine Declined: 21%) and that others in their social circles wore masks most or all the time (SoC: 28%; Vaccine Declined: 8%). Participants were generally free of chronic conditions associated with COVID-19 infection or severity, though seasonal allergies (SoC: 35%, Vaccine Declined: 39%), smoking or vaping (SoC: 22%; Vaccine Declined: 26%), and overweight or obesity (SoC: 31%, Vaccine Declined: 43%) were commonly reported (Table 1).

Table 1.

Baseline characteristics of participants in the CoVPN 3006 Study by COVID-19 study arm and vaccination status, 2021 (N=1,198)

	SoC arm N=728 (60.8%)			Vaccine Declined arm N=470 (39.2%)
	Unvaccinated N=155 (21.3%)	Vaccinated off-study N=422 (58.0%)	Vaccinated on-study N=151 (20.7%)	Unvaccinated N=396 (84.3%)	Vaccinated off-study N=74 (15.7%)
DEMOGRAPHIC INFORMATION:	N (%)	N (%)	N (%)	N (%)	N (%)
Age
Median, IQR	21 (20,23)	21 (20,23)	21 (20,22)	22 (20,26)	24 (21,26)
Gender
Man	70 (45.2)	194 (46.0)	79 (52.3)	183 (46.2)	23 (31.1)
Woman	80 (51.6)	218 (51.7)	66 (43.7)	208 (52.5)	45 (60.8)
Transgender or non-	5 (3.2)	10 (2.4)	6 (4.0)	3 (0.8)	4 (5.4)
binary
Other/not reported	0 (0)	0 (0)	0 (0)	2 (0.5)	2 (2.7)
Ethnicity
Hispanic/Latinx	46 (29.7)	72 (17.1)	29 (19.2)	93 (23.5)	23 (31.1)
Not Hispanic/Latinx	107 (69.0)	348 (82.5)	122 (80.8)	C1298 (75.3)	51 (68.9)
Not reported	2 (1.3)	2 (0)	0 (0)	5 (1.3)	0 (0)
Race
Black/African American	17 (11.0)	42 (10.0)	10 (6.6)	60 (15.2)	16 (21.6)
Asian	16 (10.3)	72 (17.1)	21 (13.9)	12 (3.0)	2 (2.7)
White	98 (63.2)	255 (60.4)	99 (65.6)	263 (66.4)	42 (56.8)
Multiple races	11 (7.1)	28 (6.6)	13 (8.6)	31 (7.8)	9 (12.2)
Other/not reported	13 (8.4)	25 (5.9)	8 (5.3)	30 (7.6)	5 (6.8)
Residence
Apartment or Condo	58 (37.4)	215 (50.9)	69 (45.7)	185 (46.7)	37 (50.0)
Stand-alone house	60 (38.7)	121 (28.7)	56 (37.1)	161 (40.7)	30 (40.5)
Dormitory or campus housing	21 (13.5)	64 (15.2)	17 (11.3)	27 (6.8)	5 (6.8)
Fraternity or sorority house	4 (2.6)	9 (2.1)	1 (0.7)	1 (0.3)	0 (0)
Other/not reported	12 (7.7)	13 (3.1)	8 (5.3)	22 (5.6)	2 (2.7)
COVID-19 RISK ASSESSMENT:
Number of people in shared bedroom
Median, IQR	0 (0,1)	0 (0,1)	0 (0,1)	0 (0,1)	1 (0,1)
Number of people in communal spaces
Median, IQR	2 (1,4)	2 (1,3)	2 (1,3)	2 (1,3)	2 (1,3)
Attending a job, volunteering, or studying on campus
Yes	109 (70.3)	274 (64.9)	97 (64.2)	289 (73.0)	57 (77.0)
No	46 (29.7)	146 (34.6)	53 (35.1)	104 (26.3)	16 (21.6)
Not reported	0 (0)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Playing in-person team sports
Yes	11 (7.1)	40 (9.5)	5 (3.3)	23 (5.8)	8 (10.8)
No	144 (92.9)	380 (90.0)	145 (96.0)	370 (93.4)	65 (87.8)
Not reported	0 (0)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Past 2 weeks: sat inside a bar, restaurant, or cafeteria
Yes	107 (69.0)	286 (67.8)	104 (68.9)	317 (80.1)	55 (74.3)
No	48 (31.0)	134 (31.8)	46 (30.5)	76 (19.2)	18 (24.3)
Not reported	0 (0)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
# days drinking in past week
Median, IQR	6 (4,7)	6 (5,7)	6 (5,7)	6 (5,7)	6 (5,7)
Past week: maximum # drinks/day
1–3	74 (47.7)	168 (39.8)	57 (37.7)	155 (39.1)	34 (45.9)
4–6	16 (10.3)	40 (9.5)	16 (10.6)	50 (12.6)	5 (6.8)
7+	2 (1.3)	10 (2.4)	2 (1.3)	7 (1.8)	0 (0)
Did not drink	63 (40.6)	202 (47.9)	75 (49.7)	181 (45.7)	34 (45.9)
Not reported	0 (0)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Current students only (N=907)
Fraternity/sorority member
Yes	21 (16.4)	57 (15.1)	18 (12.6)	15 (6.7)	1 (2.8)
No	107 (83.6)	318 (84.4)	134 (86.7)	206 (92.4)	35 (97.2)
Not reported	0 (0)	4 (1.1)	1 (0.7)	2 (0.9)	0 (0)
Attending in-person class
Yes	52 (40.6)	141 (37.3)	51 (35.7)	168 (75.3)	26 (72.2)
No	75 (58.6)	232 (61.5)	90 (62.9)	38 (17.0)	10 (27.8)
Not reported	1 (0.8)	4 (1.1)	2 (1.4)	17 (7.6)	0 (0)
Activities In the past 2 weeks…
Contact with suspected/confirmed COVID-19 case
Yes	5 (3.2)	13 (3.1)	1 (0.7)	23 (5.8)	4 (5.4)
No	150 (96.8)	407 (96.4)	149 (98.7)	370 (93.4)	69 (93.2)
Not reported	0 (0)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Always/mostly wore a mask inside around others
Yes	99 (63.9)	307 (72.7)	104 (68.9)	114 (28.8)	30 (40.5)
No	55 (35.5)	113 (26.8)	46 (30.5)	279 (70.5)	43 (58.1)
Not reported	1 (0.6)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Always/mostly physical distanced
Yes	93 (60.0)	288 (68.2)	91 (60.3)	130 (32.8)	33 (44.6)
No	61 (39.4)	132 (31.3)	59 (39.1)	263 (66.4)	40 (54.1)
Not reported	1 (0.6)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Avoided gathering in groups ≥10
Yes	58 (37.4)	182 (43.1)	61 (40.4)	77 (19.4)	26 (35.1)
No	96 (61.9)	238 (56.4)	89 (58.9)	316 (79.8)	47 (63.5)
Not reported	1 (0.6)	2 (0.5)	1 (0.7)	3 (0.8)	1 (1.4)
Others always/mostly wore a mask
Yes	37 (23.9)	122 (28.9)	45 (29.8)	33 (8.3)	6 (8.1)
No	117 (75.5)	298 (70.6)	105 (69.5)	360 (90.9)	67 (90.5)
Not reported	1 (0.6)	2 (0.5)	1 (0.5)	3 (0.8)	1 (1.4)
COVID-RELEVANT MEDICAL HISTORY:
BMI (median, IQR)	23.5 (21.2,28.0)	23.3 (21.2,26.5)	22.9 (20.6,26.6)	23.8 (21.4,28.1)	26.8 (23.0,31.2)
Cardiovascular Health
High blood pressure	2 (1.3)	2 (0.5)	0 (0)	5 (1.3)	3 (4.1)
Heart conditions	0 (0)	0 (0)	0 (0)	5 (1.3)	0 (0)
Compromised immunity
Autoimmune disease	3 (1.9)	3 (0.7)	4 (2.6)	6 (1.5)	1 (1.4)
Immunodeficiency	0 (0)	1 (0.2)	0 (0)	4 (1.0)	0 (0)
HIV/AIDS	0 (0)	0 (0)	0 (0)	4 (1.0)	0 (0)
Cancer	0 (0)	0 (0)	0 (0)
Respiratory conditions
Seasonal allergies	58 (37.4)	142 (33.6)	53 (35.1)	149 (37.6)	32 (43.2)
Lung conditions	11 (7.1)	20 (4.7)	7 (4.6)	18 (4.5)	3 (4.1)
Current smoker (including vaping)	42 (27.1)	85 (20.1)	30 (19.9)	108 (27.3)	13 (17.6)
Other conditions
Blood disorder	1 (0.6)	2 (0.5)	1 (0.7)	2 (0.5)	0 (0)
Neurologic conditions	1 (0.6)	0 (0)	0 (0)	5 (1.3)	1 (1.4)
Diabetes	1 (0.6)	0 (0)	0 (0)	5 (1.3)	0 (0)
Chronic kidney disease	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)
Chronic liver disease	0 (0)	0 (0)	0 (0)	1 (0.3)	0 (0)
Overweight/obese (BMI ≥25)	54 (34.8)	125 (29.6)	50 (33.1)	163 (41.2)	41 (55.4)
Currently pregnant	0 (0)	0 (0)	0 (0)	3 (0.8)	1 (1.4)
Any of the above medical conditions	103 (66.5)	256 (60.7)	93 (61.6)	282 (71.2)	60 (81.1)

Differences in vaccination by demographic and behavioral factors

Seventy-nine percent (n=573) of SoC participants received COVID-19 vaccination; 21% (n=151) received vaccination on-study and 58% (n=422) received vaccination off-study. Among SoC participants, median time to off-study vaccination was 23 days (IQR: 8, 65). By contrast, 16% (n=74) Vaccine Declined participants received off-study vaccination (Table 1) at a median 61 days (IQR: 39, 98) after enrollment.

Racial and ethnic disparities in the incidence of off-study vaccination were observed in both the SoC group (Asian [72/109, 66%], Black [42/69, 61%], White [255/452, 56%], Hispanic [72/147, 49%], non-Hispanic participants [348/577, 60%]) and the Vaccine Declined group (Asian [2/14, 14%], Black [16/76, 21%], White [42/263, 16%], Hispanic [23/116, 20%], non-Hispanic [51/349, 15%]). SoC participants who received off-study vaccination rather than waiting to receive on-study vaccination were more likely to be women (52% vs. 44%), to identify as Black (10% vs. 7%) or Asian (17% vs. 14%). (Table 1). Participants receiving vaccines off-study were more likely to have had recent contact with a COVID-19 case (3% vs 1%) and more likely to report masking (73% vs. 69%) and physical distancing (68% vs. 60%) than on-study vaccinees.

Within each study group, the frequency of off-study vaccination did not appear to differ by in-person classroom attendance, performing work, volunteering, or schoolwork on campus, or. eating indoors at a bar, restaurant, or cafeteria. Among participants who played team sports, off-study vaccination was more common than remaining unvaccinated (SoC: 9% vs. 7%; Vaccine Declined: 11% vs. 7%) or (among SoC participants) receiving on-study vaccination after four months (9% vs. 3%) (Table 1). Vaccine Declined participants who received off-study vaccination were more likely to reported mostly or always wore a mask indoors (41% vs. 29%), physically distancing (45% vs. 33%), or avoiding large gatherings (35% vs. 19%). Current smokers in both study groups were more likely to remain unvaccinated than to receive off-study vaccination (SoC: 27% vs 20%; Vaccine Declined: 27% vs. 18%), and overweight or obese participants in the Vaccine Declined group were more likely to receive off-study vaccination than to remain unvaccinated (55% vs. 41%).

Four-month cumulative incidence of vaccination

Participants who were vaccinated off-study were most likely to receive two doses of BNT-162b2 vaccine (n=201, 44.7%) or two doses of mRNA-1273 vaccine (n=173, 38.4%) (Table 2). Table 3 shows the crude and weighted four-month cumulative incidence of off-study COVID-19 vaccination stratified by study group, which demonstrate the extent to which associations differed by study arm. The factors associated with vaccination were modified by study group. While not associated with off-study vaccination in the SoC group, identifying as multiple races (RR: 2.68, 95% CI: 1.29, 5.58), Black (RR: 1.93, 95% CI: 0.74, 5.01); Asian (RR: 2.13, 95% CI: 0.49, 9.19), avoiding large gatherings (RR: 2.24, 95% CI: 1.18, 4.25), and presence of a COVID-relevant medical condition (RR: 1.95, 95% CI: 0.89, 4.26) were imprecise predictors of off-study vaccination in the Vaccine Declined group (Table 3). Attending class in-person was not associated with off-study vaccination in the SoC group (RR: 1.02, 95% CI: 0.89, 1.18), but was associated with reduced incidence of vaccination in the Vaccine Declined group (RR: 0.47, 95% CI: 0.21, 1.03). Hispanic ethnicity was associated with reduced incidence of off-study vaccination in the SoC group (RR: 0.74, 95% CI: 0.57, 0.97) and increased incidence in the Vaccine Declined group (RR: 1.78, 95% CI: 0.92, 3.44). Crude and adjusted analyses within study arm generally yielded similar results, though the results for two behavioral predictors of vaccination (physically distancing and reporting that others in their social circles frequently wore masks) were attenuated on adjustment for covariates. While confidence intervals frequently overlapped, trends in cumulative incidence of off-study vaccination by self-reported race, ethnicity, and physical distancing behavior are more clearly observed in the Standard of Care group compared to the Vaccine Declined group (Figure 3a–b).

Table 2.

COVID-19 vaccines received outside of the CoVPN 3006 study (N= 450)

	SOC Arm (N=383)		Vaccine Declined Arm (N=67)		Overall (N=450)
	N	%	N	%	N	%
BNT162b2 – two doses	183	47.8	18	26.9	201	44.7
BNT162b2 – one dose	19	5.0	7	10.4	26	5.8
mRNA-1273 – two doses	150	39.2	23	34.3	173	38.4
mRNA-1273 – one dose	14	3.7	8	11.9	22	4.9
Ad26.COV2.S	14	3.7	9	13.4	23	5.1
Missing manufacturer	3	0.8	2	3.0	5	1.1
Total	383	100.0	67	100.0	450	100.0

Table 3.

Baseline predictors of off-study COVID-19 vaccination in the CoVPN 3006 study, stratified by study arm (n=1,198)

	Crude analysis		Adjusted analysis2,3
	SOC (N=728) RR4 (95% CI)	Vaccine Declined (N=470) RR (95% CI)	SOC (N=728) RR (95% CI)	Vaccine Declined (N=470) RR (95% CI)
Race5 – multiple	0.89 (0.67,1.18)	2.68 (1.29,5.58)	0.89 (0.67,1.18)	2.68 (1.29,5.58)
Race – Black	1.12 (0.66,1.90)	1.93 (0.74,5.01)	1.12 (0.66,1.90)	1.93 (0.74,5.01)
Race – Asian	1.15 (0.65,2.04)	2.13 (0.49,9.19)	1.15 (0.65,2.04)	2.13 (0.49,9.19)
Hispanic/Latino	0.80 (0.65,0.99)	1.51 (0.82,2.78)	0.74 (0.57,0.97)	1.78 (0.92,3.44)
In-person class (students only, N=801)	0.96 (0.84,1.11)	0.43 (0.21,0.91)	1.02 (0.89,1.18)	0.47 (0.21,1.03)
Any medical condition6	0.92 (0.80,1.06)	2.18 (0.99,4.78)	0.95 (0.82,1.10)	1.95 (0.89,4.26)
Behavior in the past 2 weeks…
Wore a mask	1.19 (1.01,1.40)	1.62 (0.91,2.87)	1.11 (0.94,1.34)	1.22 (0.63,2.36)
Physically distanced	1.22 (1.05,1.43)	1.25 (0.69,2.26)	1.14 (0.98,1.33)	0.81 (0.40,1.64)
Avoided large gatherings	1.14 (1.00,1.30)	2.96 (1.70,5.18)	1.08 (0.94,1.24)	2.24 (1.18,4.25)
Others wore a mask	1.10 (0.95,1.26)	1.41 (0.58,3.43)	1.03 (0.86,1.24)	0.91 (0.27,3.11)

¹Vaccination defined as receiving a COVID-19 vaccine outside of the study; participants were censored after 4 months of observation time, on December 3, 2021, or if they received an on-study vaccine, whichever occurred first

²

The following minimally sufficient adjustment covariate sets were identified from directed acyclic graphs:

1. Race: no adjustment necessary;
2. Ethnicity: race;
3. Preventative behaviors: age, study enrollment date, ethnicity, gender, student status, in-person work, medical condition, race;
4. In-person class: study enrollment date;
5. medical condition: age, ethnicity, gender, race

³In multivariable analyses adjusting for gender, transgender and gender non-conforming individuals were not included due to positivity concerns (N=25/1,086 individuals)

⁴Four-month cumulative incidence ratios (i.e., risk ratios (RRs)) and 95% confidence intervals (CIs) from Kaplan-Meier estimators with robust standard errors

⁵Using White race as a comparison; “Other” race was not examined as a category in the time-to-event analyses given the small sample size and limited inferences that can be drawn from this homogenous group

⁶Those specified in Table 2: high blood pressure, heart conditions, autoimmune disease, immunodeficiency, HIV/AIDS, cancer, seasonal allergies, lung conditions, current smoker, blood disorder, neurological conditions, diabetes, chronic kidney disease, chronic liver disease, overweight/obesity (BMI≥25)

Figure 3a-b.

Four-month adjusted¹ cumulative incidence of off-study COVID-19 vaccination² by self-reported race, ethnicity, and physical distancing behavior³, in the a) Standard of Care arm and b) Vaccine Declined arm

¹Race models were not weighted to adjust for any covariates given that race was used as a proxy for racism, a fundamental cause of health inequities. Ethnicity models adjusted for race. Physical distancing model was adjusted for age, study enrollment date, ethnicity, gender, student status, in-person work, medical condition, race.

²Vaccination defined as receiving a COVID-19 vaccine outside of the study; participants were censored after 4 months of observation time, on December 3, 2021, or if they received an on-study vaccine, whichever occurred first.

³Self-reported racial categories included: Black, Asian, Multiracial, or White. “Other” was excluded from statistical analyses due to small cells. Physical distancing was defined as always/mostly physically distancing (yes/no).

DISCUSSION

In a secondary analysis of data from a randomized trial of mRNA-1273 vaccination beginning March 2021, we had the opportunity to compare the relative contributions of sociodemographic and behavioral predictors to COVID-19 vaccination behaviors in two populations that differed with respect to their level of vaccine confidence. We found that over half of participants who expressed interest in COVID-19 vaccination and willingness to delay vaccination (SoC participants) sought and received COVID-19 vaccination off-study before the end of their four-month follow-up period. Considering the SoC participants who were offered and accepted on-study vaccination upon study closure, nearly 80% of the SoC arm were vaccinated. By contrast, only 16% of participants who were initially uninterested in COVID-19 vaccination (Vaccine Declined participants) received it. We did not identify salient predictors of off-study vaccination among SoC participants. Because they enrolled in the randomized component of the study in which they agreed to be vaccinated, either immediately or in four months’ time, this group on the whole was willing to receive vaccination independent of their social or demographic characteristics. By contrast, certain demographic and behavioral characteristics were associated with vaccination among Vaccine Declined participants. The stark differences in vaccination behaviors and correlates between these two study arms suggest that a) individuals who are interested in vaccination can be motivated to vaccinate at the earliest opportunity; b) challenges with vaccine confidence can be difficult to overcome in the context of an emerging infection and a novel vaccine, and c) individuals with reduced vaccine confidence who eventually receive vaccination might perceive a higher risk of being infected with SARS-CoV-2 than those who do not. These findings provide insight into decision-making for receiving a novel vaccine during a public health emergency and can inform vaccination rollout and communications strategies for future campaigns.

Because our target population included young adults, we were not surprised that many participants reported living, working, and socializing in-person with others. In addition, prior studies have found reduced vaccine confidence among young adults compared to older adults, suggesting generational differences in the interpretation or acceptance of vaccination recommendations. Because COVID-19 tends to be less severe in young, healthy adults, it is possible that a low perception of risk or severity contributed to reduced vaccine confidence in our sample. However, consistent with the fact that COVID-19 vaccine rollout occurred rapidly following the launch of this study, SoC participants sought vaccination outside of the study more often than Vaccine Declined participants. Our description of the study sample suggested that SoC participants were more cautious than Vaccine Declined participants in preventing COVID-19 through masking and physical distancing, which were the primary prevention methods prior to the availability of vaccines. Furthermore, SoC participants who received off-study vaccination showed higher levels of preventive behaviors than those who waited to receive on-study vaccination. In addition, while they reported COVID-19 preventive measures at lower levels than SoC participants, Vaccine Declined participants who received off-study vaccination were twice as likely to avoid large gatherings than those who did not in adjusted analysis. We hypothesize that a) SoC participants were generally more motivated to receive COVID-19 vaccination, and thus more likely to access vaccination once it became widely available, and b) Vaccine Declined participants who were more likely to practice preventive behaviors had greater COVID-19 risk perception and were more likely to eventually accept vaccination.

Though COVID-19 vaccination was low among Vaccine Declined participants, our findings suggest that initially hesitant individuals can change their minds to accept vaccination. The literature has shown that vaccination preferences can change over time or may be influenced by external factors [18–23], such as increasing availability of vaccines, full FDA approval of COVID-19 vaccines, or societal pressure to vaccinate. Increased vaccination uptake was likely also influenced by vaccination mandates by employers and educational institutions, which began in July 2021 and increased throughout the remainder of the year [24]. In addition, the onset of the autumn and winter holiday seasons might have influenced hesitant individuals to receive vaccination for the purposes of traveling and spending time indoors with friends and family. Finally, increasing availability of COVID-19 vaccine safety and effectiveness data might have convinced individuals who were skeptical about the speed at which vaccines were developed and authorized. Whatever the motivation, it is important to note that vaccine acceptability can increase over time and continued recommendations and education to unvaccinated individuals can instill vaccine confidence. It is possible that our continued engagement with Vaccine Declined participants over the course of the study encouraged some of them to seek COVID-19 vaccination.

In keeping with other studies and with national trends, we observed that Asian participants in the SoC arm had a higher cumulative incidence of vaccination than other racial groups, and Black and multiracial participants had the lowest cumulative incidence [13, 25]. Historic lack of trust in medical institutions has been documented as a barrier to COVID-19 vaccination in Black populations and may also play a role in our study, which sampled Black participants proportionally to the US population. The initial lack of full FDA approval and the speed at which the vaccines were developed and rolled out was a cause for concern among many individuals, and a need for more information on the safety and effectiveness of the vaccines, as well as more time to observe the effects of vaccination in their communities, was a frequently documented barrier to COVID-19 vaccine uptake [26–30]. Consistent with this history, we observed greater representation of Black (16% vs. 9%) and Hispanic (25% vs. 20%) participants in the Vaccine Declined group compared to the SoC group, respectively. However, we observed slightly higher vaccination incidence in Black and multiracial participants compared to White and Asian participants in the Vaccine Declined group, and in Hispanic compared to non-Hispanic participants, suggesting that there were more gains to be made among Black and Hispanic participants with lower vaccine confidence.

The major challenge to the execution of the study and the interpretation of our findings is the changing epidemiology of COVID-19 and availability of COVID-19 vaccination during the study period. At the start of the study, COVID-19 vaccination was prioritized for older and high-risk adults, and it was expected that vaccination would roll out more gradually for young adults. However, the EUA for Ad26.COV2.S vaccination among all adults was likely an attractive option for individuals who wanted fewer vaccine doses, a lower risk of side effects, and a traditional vaccine platform compared to mRNA vaccines [31, 32]. Furthermore, the expansion of the recommendation for vaccination among all adults in May 2021 increased demand for vaccination and reduced willingness to delay vaccination. Finally, the emergence of the Delta variant in the summer of 2021, which was associated with higher transmissibility and greater disease severity among young adults, might have increased risk perceptions and willingness to vaccinate. As such, we must interpret our findings in this context, and additionally limited by the absence of data regarding vaccination intentions over time. While we collected weekly data on self-reported COVID-19 risk and preventive behaviors, the high level of missingness in this data precluded an analysis of time-varying influences on vaccination, which might be more highly correlated with vaccination than baseline behaviors. One example is in-person school attendance, which likely changed over time because our study period encompassed the Spring 2021 semester, summer break, and the Fall 2021 semester. Additionally, evolving university policies on remote learning and vaccination mandates could certainly have influenced vaccination intentions and behaviors more strongly when they were introduced in the later months of the study period. It is possible that the rapid increase in availability of COVID-19 vaccination during the study period contributed to poor compliance with weekly surveys, as participants who received vaccination off-study might have lost motivation to continue participating in weekly study procedures. Twenty-one percent of SoC participants were categorized as unvaccinated by the end of the study, which appeared to be more affected by the fact many were censored due to loss to follow up before the event could be observed (101/155, 65%), than by outright refusal to vaccinate. However, those who were lost to follow-up may have sought vaccination after leaving the study, resulting in potential bias due to censoring. It is also possible that the added barrier of seeking vaccination off-study reduced their motivation to vaccinate. An additional limitation of our study is the lack of data from smaller demographic subgroups, namely participants who identified as “other” race or a non-binary gender. Future studies should be designed to specifically examine COVID-19 vaccination intentions and uptake among these minoritized populations. However, the large geographic scope of the study provided a large, diverse, and robust sample for analysis of COVID-19 vaccination in young adults. Importantly, the shift to endemicity and the Center for Disease Control and Prevention’s recommendation for annual COVID boosters presents a new set of challenges for vaccine uptake like those for seasonal influenza vaccination, including vaccine fatigue, low perceived risk for infection and severity of disease, and diminished trust in the healthcare system. Low vaccine confidence is an ongoing area of concern for infectious disease prevention, and our findings provide evidence that unvaccinated individuals can accept vaccination after a time. Thus, we can focus efforts on strategies to better communicate the benefits of vaccination and to make vaccination accessible and convenient. Furthermore, our findings can be applicable to the next novel outbreak pathogen, providing insights into who is most or least likely to seek vaccination and potential methods to encourage vaccination and other preventive behaviors.

While hesitancy or delay in receiving prophylactic vaccines can hinder control efforts during an outbreak, our findings show that COVID-19 vaccination does occur among individuals who initially decline it, and can even be accelerated among individuals who accept it. Furthermore, individuals who have reservations about receiving vaccinations can nonetheless actively engage in vaccination research, serving as comparators to assess vaccine efficacy. This engagement provides critical data on barriers and drivers of vaccination; and enhancing awareness of vaccines and vaccination recommendations that can encourage vaccination at a later time. Our study found that practicing COVID-19 prevention behaviors, which suggests an interest in protecting oneself from infection, protecting others, or both, was associated with higher incidence of COVID-19 vaccination. These findings suggest that sufficiently and accurately communicating the risk and severity of COVID-19, and emphasizing the importance of prevention, can likely influence vaccination uptake. Continued education and advocacy from health care providers and the scientific community might be one way to encourage vaccination among those with low vaccine confidence.