Predicting COVID-19 vaccination timing by integrating the theory of planned behavior and the diffusion of innovations: a cross-sectional survey in Macao, China
Original Article

Predicting COVID-19 vaccination timing by integrating the theory of planned behavior and the diffusion of innovations: a cross-sectional survey in Macao, China

Wei He1,2# ORCID logo, Jianwei Wu3,4#, Caleb Huanyong Chen3, Aaron Finley3, Hui Wang5, Hui Huang6, Chanwa Ng7, Tinyat Chui7, Jinghua Zhang1,3, Chitin Hon1,2,8

1Respiratory Disease AI Laboratory on Epidemic Intelligence and Medical Big Data Instrument Applications, Macau University of Science and Technology, Macao, China; 2Department of Engineering Science, Faculty of Innovation Engineering, Macau University of Science and Technology, Macao, China; 3School of Business, Macau University of Science and Technology, Macao, China; 4Nursing and Health Education Research Centre, Kiang Wu Nursing College of Macau, Macao, China; 5Medical-Surgical Department, Whittier Hospital Medical Center, Whittier, CA, USA; 6Faculty of Business, Hong Kong Polytechnic University (PolyU), Hong Kong, China; 7Pui Ching Middle School Macau, Macao, China; 8Guangzhou Laboratory, Guangzhou, China

Contributions: (I) Conception and design: W He, J Zhang, C Hon; (II) Administrative support: J Zhang; (III) Provision of study materials or patients: None; (IV) Collection and assembly of data: J Wu, W He; (V) Data analysis and interpretation: W He, J Wu, C Hon; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

#These authors contributed equally to this work.

Correspondence to: Jinghua Zhang, PhD. Respiratory Disease AI Laboratory on Epidemic Intelligence and Medical Big Data Instrument Applications, Macau University of Science and Technology, Macao, China; School of Business, Macau University of Science and Technology, Avenida Wai Long, Taipa, Macao 999078, China. Email: jhuzhang@must.edu.mo; Chitin Hon, PhD. Respiratory Disease AI Laboratory on Epidemic Intelligence and Medical Big Data Instrument Applications, Macau University of Science and Technology, Macao, China; Department of Engineering Science, Faculty of Innovation Engineering, Macau University of Science and Technology, Avenida Wai Long, Taipa, Macao 999078, China; Guangzhou Laboratory, Guangzhou 510120, China. Email: cthon@must.edu.mo.

Background: This study integrates the theory of planned behavior (TPB) and the diffusion of innovations (DOI) to investigate determinants of coronavirus disease 2019 (COVID-19) vaccination timing among tourism industry workers in Macao, China, a high-risk group during the pandemic.

Methods: A cross-sectional survey of 608 respondents was analyzed using hierarchical generalized linear model (GLM) to identify predictors of vaccination timing. The analysis was further complemented by K-means clustering to segment population groups based on attitudes, social norms, and behavioral control.

Results: Cluster analysis revealed diverse attitudes toward COVID-19 vaccination, suggesting the need for targeted public health interventions. The study identified several significant behavioral predictors of COVID-19 vaccination timing, including perceived behavioral control (PBC) (0.519, P<0.001), COVID-19 vaccine attitude (VA) (0.559, P<0.001), and past influenza vaccination history (1.268, P<0.001). The factor of conformity trait is not significant.

Conclusions: Integrating DOI and TPB underscores the interplay of cognitive, social, and innovation-related factors in vaccination timing. Within the DOI framework, individuals classified as innovators and early adopters typically exhibit favorable attitudes and strong subjective norms (SNs) toward vaccination. This underscores the utility of combining DOI and TPB to design targeted vaccination campaigns to capture cognitive, social, and innovation-related drivers of behavior.

Keywords: Coronavirus disease 2019 vaccination (COVID-19 vaccination); vaccination timing; theory of planned behavior (TPB); diffusion of innovations (DOI); vaccine hesitancy


Submitted Aug 31, 2024. Accepted for publication Mar 13, 2025. Published online May 27, 2025.

doi: 10.21037/jtd-24-1313


Highlight box

Key findings

• This study identifies key psychological and behavioral predictors of coronavirus disease 2019 (COVID-19) vaccination timing among tourism industry workers in Macao. Perceived behavioral control, COVID-19 vaccine attitude (VA), and history of influenza vaccination are significantly associated with earlier vaccination. The conformity trait shows no significant effect.

What is known and what is new?

• Theory of planned behavior (TPB) and the diffusion of innovations (DOI) are commonly used to explain vaccine uptake, typically treating adoption as a binary outcome.

• This study integrates TPB and DOI to examine vaccination timing, highlighting that early adopters show stronger attitudes, greater perceived control, and prior flu vaccination. It also identifies conformity trait as a delaying factor and a moderator of VA, offering new insights for timing-sensitive interventions.

What is the implication, and what should change now?

• Combining TPB and DOI provides a comprehensive framework to understand both compliance and timing of vaccination. Public health campaigns should promote early adoption by reinforcing positive attitudes and social influence, especially in high-risk occupational sectors. Past influenza vaccination records may also help identify potential early adopters during future pandemics.


Introduction

Vaccines are recognized as the safest, most effective, and most convenient innovation in healthcare for the global control and prevention of infectious diseases, marking a significant milestone in public health. However, skepticism and concerns have persisted since their inception, leading to vaccine hesitancy, particularly delays in coronavirus disease 2019 (COVID-19) vaccination worldwide. Promoting vaccine acceptance is crucial for effective emergency responses to future pandemics. Various factors influence vaccination decisions, including demographics, cognitive, and psychosocial elements, such as age, gender, education, insurance status, attitudes toward the vaccine, trust in public health information, perceived susceptibility to COVID-19, and perceptions of vaccine benefits and side effects. Behavioral theories, such as the theory of planned behavior (TPB) and the diffusion of innovations (DOI) Theory, have been utilized to explain intentions to vaccinate against COVID-19. Each theory addresses different aspects of behavior change, and their combination can enhance the comprehensiveness of public health strategies aimed at improving vaccination rates. Integrating the TPB and the DOI theory provides a multifaceted understanding of vaccination behaviors.

The TPB

The TPB (1) has been widely applied to explain vaccination behavior (2). Developed by Icek Ajzen in 1991, the TPB posits that an individual’s intention to engage in a behavior is influenced by three key components: attitudes (i.e., positive or negative evaluations of the behavior), subjective norms (SNs) (i.e., perceived social pressure to perform the behavior), and perceived behavioral control (PBC) (i.e., confidence in one’s ability to execute the behavior). These factors collectively shape behavioral intention, which is the strongest predictor of actual behavior (3). TPB has been extensively applied to health-related decision-making, including vaccination uptake, due to its capacity to disentangle psychosocial drivers of action (4,5). For instance, attitudes toward vaccines—shaped by perceived benefits and safety concerns—significantly predict immunization intent (6). SNs, such as recommendations from healthcare providers or family, further modulate willingness to vaccinate (7). PBC, for example vaccine accessibility or confidence in navigating the vaccination process, also plays a critical role (8). Moreover, extended TPB models that incorporate variables such as past behavior and personality have been shown to be significant predictors of vaccination behavior (9). During the COVID-19 pandemic, TPB has played an active role in explaining acceptance of COVID-19 vaccines (1,2). It is estimated to account for 40–60% of the variance in vaccination intentions (10), and a study demonstrated that intention to vaccinate was strongly tied to perceived severity of the disease (attitude), trust in public health guidelines (SN), and logistical barriers (perceived control) (11).

The DOI theory

Proposed by Rogers (11), the DOI Theory is one of the most popular theories for understanding how innovations spread and are adopted within a population. It divides individuals into categories such as innovators, early adopters, early majority, late majority, and laggards, based on their willingness to accept new ideas (11).

DOI theory has been applied to understand health behaviors and technology adoption (12). Emphasizing the role of information dissemination and social networks in early adoption, DOI has provided a framework for understanding the spread of COVID-19 vaccination within populations (12). Early adopters, characterized by strong positive attitudes and high PBC, are influenced by opinion leaders and social networks (13,14). They prioritize benefits and observability, effectively reducing perceived risks (15). Late adopters, however, face distinct barriers, such as concerns about vaccine safety and weak social pressure, often exacerbated by misinformation (16). Structural issues, like limited access to vaccines, further complicate their decision-making (17). They require stronger social proof and simplified processes to overcome hesitancy (11).

COVID-19 vaccination is distinct from regular behavior, especially as COVID-19 vaccines have been newly invented and granted emergency use authorization (18). Therefore, not only vaccine uptake (just like adopting a new technology) but also vaccination timing (i.e., the timing of adoption in individuals’ decision making) is critical for establishing herd immunity. People tend to observe their peers’ choices as a reference in the early stage, particularly for new vaccines perceived as having uncertain safety and efficacy (19). Hence, the more others have been vaccinated (notably, when vaccination reaches above 20%), the more likely individuals are to choose vaccination (20).

Integrating DOI and TPB to enhance explanatory power for vaccine adoption timing

TPB and DOI share several common behavioral variables and conceptual underpinnings. However, while TPB traditionally examines adoption as a binary outcome (1), it rarely addresses the critical dimension of timing. In contrast, DOI explicitly categorizes adopters based on the timing of adoption (11).

Vaccine adoption therefore involves both compliance and timing. In the DOI model, the timing of adoption can be viewed as analogous to compliance in TPB—both constructs capture consistent behavioral patterns though they are interpreted differently. For example, early adopters demonstrate strong compliance by vaccinating promptly; the “movable middle”, who initially hesitate but eventually comply, display delayed adoption timing; and vaccine refusers remain at zero on the timing scale, refusing vaccination altogether. This alignment underscores the interplay between timing and compliance in understanding vaccine uptake behaviors (21-23).

Integrating DOI into TPB offers targeted intervention opportunities. For instance, strategic messaging—such as emphasizing community vaccination rates to induce bandwagon effects—can mobilize the “movable middle” (15). For early adopters, campaigns may amplify their positive experiences and leverage trusted influencers to normalize vaccination, while for late adopters, interventions should address both structural and psychological barriers by reframing vaccines as compatible with prevailing cultural values (24).

Research objectives

This study conceptualizes COVID-19 vaccination as an innovation adoption behavior within an integrated DOI and TPB framework. It aims to identify the key TPB factors associated with the timing of vaccination uptake among tourism industry workers, who have borne significant economic risks during the COVID-19 pandemic. With a specific focus on characterizing early adopters and exploring the determinants of early vaccine uptake, this research seeks to inform strategies for expanding the early majority segment, thereby enhancing overall vaccine acceptance and reducing hesitancy. The findings are expected to contribute to the growing literature on vaccination hesitancy and the adoption of new health technologies, addressing persistent challenges in public health policy and practice. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1313/rc).


Methods

Study location

Macao Special Administrative Region of China (Macao), with a population of 680,000 and 39.4 million visitors in 2019 (25), stands as one of the most densely populated cities globally. The tourism satellite industries—including gaming, retail trade, food and beverage, hotels, passenger transportation, and travel agency services—employed approximately 203,000 individuals, constituting nearly half of Macao’s working population in 2019 (26).

Despite the active engagement of workers in these industries in COVID-19 preventive measures (27) and their expressed desire for an end to the pandemic, hesitancy towards adopting this novel vaccination practice has been observed. To achieve the goal of herd immunization, the Macao health authority has put significant efforts into promoting vaccination (28). By the end of 2022, approximately 94.3% of Macao residents had completed the primary vaccination series, and 57.2% had received at least one booster dose (28).

After the discontinuation of the “Zero-COVID” policy in mainland China in December 2022, Macao experienced an unprecedented surge in Omicron infection cases. It is estimated that around 70% of Macao’s population was affected by the virus, with a mortality rate of approximately 0.012% and a total of 57 reported deaths during the outbreak. Consequently, investigating the tourism satellite industries in Macao offers a valuable opportunity to examine COVID-19 vaccine adoption.

Participants and data collection

A structured questionnaire was developed based on the TPB framework (29) and adapted from established guidelines for assessing attitudes toward COVID-19 vaccination. The instrument incorporated a conformity trait scale developed by Mehrabian and Stefl (30). In addition, it collected demographic information, including gender, age, education level, marital status, employment status, residence, cohabitation with children or elderly individuals, work sector, years of experience in travel-related industries, and monthly income level. To ensure content validity, the questionnaire was reviewed by subject matter experts, and a pilot study involving 47 participants from the travel service industry was conducted to verify its effectiveness. A quality control question (“What year is this year?”) was embedded to screen out inattentive respondents and ensure data integrity.

Participants were recruited using a convenience sampling approach, targeting full-time adult employees in Macao’s tourism satellite industries (e.g., gaming, retail trade, food and beverage, hotel, passenger transportation, and travel agency services). Initial outreach was facilitated through industry associations and the human resources departments of major gaming companies, which were asked to distribute random invitations to complete the online survey.

The questionnaires were hosted on Google Forms and supported by a poster featuring a quick response (QR) code and website link for ease of access. Data collection spanned from July 28th to August 20th, 2021, with a targeted minimum sample size of 400 based on the sampling formula. After the collection period, the dataset was subjected to rigorous error-checking procedures to ensure accuracy and reliability. This methodology is consistent with the approaches described by Wu et al. (31).

Measures

A self-administered questionnaire was used to assess (I) COVID-19 vaccination timing; (II) COVID-19 vaccine attitude (VA), SN, and PBC; (III) conformity trait, collective responsibility (CR), and influenza vaccination history; and (IV) socio-demographic characteristics. The items for (II) were designed based on the TPB and previous studies, whereas the items for (III) represent the extended variables of TPB. A panel of seven public health experts evaluated the content validity of the questionnaire, and its construct validity was examined using exploratory factor analysis (EFA) via principal components analysis (PCA).

Internal consistency reliability was evaluated using Cronbach’s alpha. The comprehensive procedures employed in this study have been documented in our preceding article (Wu et al., 2022) (31).

Outcome variable

COVID-19 vaccination timing

COVID-19 vaccination timing was adopted as the outcome variable because it integrates both the timing and compliance of vaccine uptake actions within the framework of the TPB (20-23). In this study, Vaccination Timing was assessed by capturing both the act of receiving the vaccine and the timing of the administration of the first dose. The timing score ranged from 0 to 7: respondents were sorted based on the month in which they received the vaccine (with data spanning 7 months), where the earliest respondents received a score of 7, and those vaccinated in the latest month received a score of 1. A score of 0 indicated that the individual had not received the COVID-19 vaccine.

Based on the adopter timing classification framework from the DOI theory, participants were grouped into three categories: the earlier vaccination group (EVG), later vaccination group (LVG), and unvaccinated group (UG). Participants who received the first dose of the COVID-19 vaccine before June 2021 were classified as EVG, given that only 16.2% of the population had been vaccinated by that date (32). Those who received their first dose in June 2021 or later were assigned to the LVG, while individuals who had not received any COVID-19 vaccine were placed in the UG.

Explanatory variables

COVID-19 VA

COVID-19 VA was assessed using a belief-based measure, and the reliability of the summed scale was evaluated with Cronbach’s alpha as described in a previous study (29). Participants were presented with two positive behavioral beliefs (related to natural and additional benefits) and two negative behavioral beliefs (related to side effects and safety hazards). For each belief, participants provided an evaluation of the corresponding outcome. All items were rated on a 5-point unipolar scale (1 to 5). The belief-based measure of COVID-19 VA was computed by summing the four products of paired behavioral beliefs and outcome evaluations, resulting in a total score ranging from 4 to 100, with higher scores indicating a more positive VA. The validity and reliability of the scale were established—with two factors accounting for 63.26% of the total variance—and the overall Cronbach’s alpha was 0.68, indicating acceptable reliability (31).

SN

SN was measured using a belief-based approach that encompassed four normative beliefs related to the influence of one’s family, employer, trusted medical professionals, and trusted associations (31). In addition, participants’ motivation to comply with these referents was assessed. Each normative belief and its corresponding motivation to comply were rated on a 5-point unipolar scale (1= strongly disagree to 5= strongly agree). The SN score was derived by summing the products of each normative belief with its corresponding motivation to comply, yielding a total score between 4 and 100. A higher score indicates a stronger SN. The scale demonstrated robust validity—with one factor explaining 64.74% of the total variance—and an excellent internal consistency (Cronbach’s alpha =0.92) (31).

PBC

PBC regarding COVID-19 vaccination was assessed using a single-item measure. Participants were asked to rate the ease or difficulty of obtaining the vaccine on a 5-point scale, where 1 indicated “very difficult” and 5 indicated “very easy” (29,33).

Conformity trait

Conformity trait was measured using the Conformity Scale developed by Mehrabian and Stefl (30). The Chinese version of the scale comprises 10 items and has demonstrated sufficient validity and reliability. EFA yielded a two-factor solution that accounted for 47% of the variance, and the scale exhibited a Cronbach’s alpha of 0.74 (34). Each item was scored on a 7-point bipolar scale ranging from −3 (strongly disagree) to +3 (strongly agree). Six items were positively worded (e.g., “I often rely on and act upon the advice of others”), reflecting a stronger tendency toward conformity, while the remaining four items were negatively worded and subsequently reverse-scored. The total score ranged from −30 to +30, with higher scores indicating a higher level of conformity.

CR

CR was measured using the three-item subscale of the 5C scale. Each item was rated on a 5-point scale (1= strongly disagree to 5= strongly agree), with one item (“When everyone is vaccinated, I don’t have to get vaccinated too”) reverse-scored to maintain consistency. The three items were averaged to create a composite score, with a higher average indicating a higher sense of CR. The subscale demonstrated acceptable reliability with a Cronbach’s alpha of 0.71 (35).

Influenza vaccination history

Influenza vaccination history was measured by a single item asking, “Did you receive last season’s influenza vaccine?” Responses were recorded as “yes” or “no”.

Socio-demographic characteristics

Socio-demographic variables included gender, age, educational level, marital status, monthly income, cohabitation with older adults or children, and industry sector.

Statistical analysis

Cluster analysis

Cluster analysis is an unsupervised machine learning technique used to identify natural groupings within data. In public health research, it is valuable for exploring heterogeneity in behaviors and attitudes, particularly regarding health behaviors like vaccination uptake. By identifying clusters of individuals with similar characteristics, this method may help uncover new subgroups within vaccine-hesitant populations (Kaufman & Rousseeuw, 2009) (36).

In this study, cluster analysis was employed to investigate variations in attitudes and behaviors toward COVID-19 vaccination among distinct population groups. Specifically, K-means clustering was applied to four variables: conformity, COVID-19 VA, SN, and PBC. The optimal number of clusters (k) was determined using an elbow plot and considerations of practical interpretability. Once the number of clusters was finalized, each participant was assigned to one of the resulting groups. Cluster membership frequencies and mean scores for the variables were then analyzed to explore participant heterogeneity.

Hierarchical generalized linear model (GLM)

A hierarchical GLM was conducted to examine predictors of COVID-19 vaccination timing (i.e., the timing of the first-dose vaccination). In the first block (Model 1), socio-demographic characteristics (e.g., gender, age, education, income) were entered. The second block (Model 2) added variables derived from the TPB (COVID-19 VA, SN, and PBC), along with conformity trait and CR. In Model 3, influenza vaccination history was included, and in Model 4, interaction terms (e.g., COVID-19 VA × conformity trait) were introduced. Model fit was assessed using pseudo R2, deviance, and log-likelihood.

Ethical issues

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the School of Business, Macau University of Science and Technology (No. MUST/MSB/2021/03) and informed consent was obtained from all individual participants when they voluntarily participated and completed the questionnaire. At the beginning of the survey, participants were informed that their participation was anonymous and voluntary, that their privacy would be strictly protected, and that no personal information would be tracked or stored in any form.


Results

Participant characteristics among the three COVID-19 vaccination groups

A total of 620 questionnaires were collected, of which 608 (98.06%) were deemed valid for analysis. Among these respondents, 135 participants received the first dose of the COVID-19 vaccine before June 2021 (EVG), 245 participants received the first dose in June 2021 or later (LVG), and 228 participants remained unvaccinated (UG).

As shown in Table 1, a greater proportion of participants in the EVG had received the influenza vaccine (58.5%) compared to those in the LVG, while the lowest influenza vaccination rate was observed in the UG (4.4%). Moreover, participants in the EVG scored highest on measures of COVID-19 VA, SN, PBC, and CR, whereas those in the UG had the lowest scores. Interestingly, Table 1 also reveals that participants in the LVG and UG groups exhibited significantly higher scores on the conformity trait than those in the EVG, with no significant difference between the LVG and UG groups.

Table 1

Demographic characteristics and comparison among three COVID-19 vaccination timing groups

Characteristics Summary statistics (n=608) COVID-19 vaccination timing groups χ2 Post-hoc tests
EVG (group I, n=135) LVG (group II, n=245) UG (group III, n=228)
Age (years) 38.26±10.36 [18–72] 38.82±10.15 [18–62] 38.37±10.39 [20–72] 37.80±10.47 [22–69] 0.435
Gender 4.196
   Male 222 (36.5) 56 (41.5) 94 (38.4) 72 (31.6)
   Female 386 (63.5) 79 (58.5) 151 (61.6) 156 (68.4)
Education 0.080
   High school or below 284 (46.7) 63 (46.7) 116 (47.3) 105 (46.1)
   Diploma or above 324 (53.3) 72 (53.3) 129 (52.7) 123 (53.9)
Marriage 1.249
   Married 372 (61.2) 85 (63.0) 154 (62.9) 133 (58.3)
   Single 236 (38.8) 50 (37.0) 91 (37.1) 95 (41.7)
Living with older adults or children 1.695
   Yes 391 (64.3) 85 (63.0) 152 (62.0) 154 (67.5)
   No 217 (35.7) 50 (37.0) 93 (38.0) 74 (32.5)
Influenza vaccination history 136.384***
   Yes 193 (31.7) 79 (58.5) 104 (42.4) 10 (4.4) I > II > III
   No 415 (68.3) 56 (41.5) 141 (57.6) 218 (95.6)
Monthly income (in local currency MOP) 34.689***
   ≤10,000 98 (16.1) 35 (25.9) 32 (13.1) 31 (13.6) I > II, I > III
   10,001–20,000 237 (39.0) 55 (40.7) 108 (44.1) 74 (32.5) II > III
   20,001–30,000 198 (32.6) 27 (20.0) 71 (29.0) 100 (43.9) I < III, II < III
   30,001–40,000 49 (8.1) 10 (7.4) 25 (10.2) 14 (6.1)
   ≥40001 26 (4.3) 8 (5.9) 9 (3.7) 9 (3.9)
Working industries 32.724***
   Travel agency 125 (20.6) 38 (28.1) 46 (18.8) 41 (18.0)
   Gaming 167 (27.5) 22 (16.3) 73 (29.8) 72 (31.6) I < II, I < III
   Food & beverage 178 (29.3) 49 (36.3) 83 (33.9) 46 (20.2) I > III, II > III
   Others 138 (22.7) 26 (19.3) 43 (17.6) 69 (30.3) II < III
Conformity trait −2.09±6.59 −4.10±7.80 −1.33±5.72 −1.70±6.47 7.762*** I > II, I > III
CR2 3.72±0.64 3.98±0.65 3.86±0.56 3.44±0.60 44.768*** I > III, II > III
COVID-19 VA 42.12±12.82 51.06±12.70 44.32±11.76 34.47±9.15 430.385*** I > II > III
SN 59.28±25.03 72.84±23.42 63.36±23.26 46.88±22.08 61.183*** I > II > III
PBC 4.11±1.33 4.69±0.92 4.49±1.11 3.36±1.41 79.470*** I > III, II > III

Data are presented as mean ± SD [range] or n (%). Statistics were based on Brown-Forsythe test, ***, P<0.001. COVID-19, coronavirus disease 2019; CR, collective responsibility; EVG, earlier vaccination group; LVG, later vaccination group; MOP, Macanese Pataca; PBC, perceived behavioral control; SD, standard deviation; SN, subjective norm; UG, unvaccinated group; VA, vaccine attitude.

The pairwise relationships among behavioral variables are illustrated in Figure 1. K-means clustering identified three distinct groups based on conformity, VA, SN, and PBC. The clusters consisted of 100 (Cluster 0), 226 (Cluster 1), and 282 (Cluster 2) respondents, respectively.

Figure 1 Pairwise distributions of conformity, VA, SN, and PBC across the three clusters. COVID-19, coronavirus disease 2019; PBC, perceived behavioral control; SN, subjective norm; VA, vaccine attitude.

Cluster 0 exhibited the lowest conformity (−12.10) and highest PBC (4.30), indicating greater autonomy and ease of vaccination. Cluster 1 showed moderate conformity (−0.34), the lowest VA (−7.34), and moderate SN (41.12), reflecting a group with low vaccination motivation. Cluster 2 was characterized by the highest SN (71.64), VA (10.15), and PBC (4.57), suggesting strong social influence and positive VA.

Analysis of variance (ANOVA) results confirmed statistically significant differences across clusters for all variables (conformity: F=325.13, P<0.001; VA: F=168.06, P<0.001; SN: F=142.12, P<0.001; PBC: F=271.87, P<0.001).

Predictors of COVID-19 vaccination timing

A hierarchical GLM was conducted to identify key predictors of COVID-19 vaccination timing (i.e., timing of vaccination). Model 1 included only the socio-demographic characteristics (e.g., gender, marriage status, identity, living place, education, monthly income, age). Model 2 added the core TPB constructs—PBC, VA, and SN—as well as conformity trait. Model 3 further included influenza vaccination history (yes =1, no =0). Finally, Model 4 introduced a single interaction term (PBC × VA) to test moderating effects between perceived ease and attitude.

Table 2 presents the resulting coefficients, standard errors, and fit statistics for each model. In Model 1 (pseudo R2 =0.1282), certain socio-demographic variables (identity category, age, some education levels, and Monthly income_3) were initially associated with earlier vaccination. In Model 2 (pseudo R2 =0.4491), adding PBC, VA, SN, and conformity trait substantially improved the explanatory power. Model 3 (pseudo R2 =0.5496) included influenza vaccination history, which remained a highly significant predictor of earlier COVID-19 vaccination. In Model 4, the introduction of the interaction term PBC × VA did not further increase the overall pseudo R2 beyond 0.5582.

Table 2

Hierarchical GLM predicting COVID-19 vaccination timing (reversed)

Variables (predictors) Hierarchical GLM
Model 1 Model 2 Model 3 Model 4
Intercept 2.329** (0.813) 1.803** (0.688) 1.742** (0.645) 1.741** (0.645)
Gender (male =1, female =0) 0.207 (0.173) 0.101 (0.147) −0.029 (0.139) −0.040 (0.138)
Marriage (single =1, married =0) –0.265 (0.193) −0.256 (0.164) −0.294 (0.154) −0.281 (0.154)
Residential status (local =1, non-local =0) −1.102*** (0.297) −0.662** (0.257) −0.596** (0.242) −0.626** (0.242)
Living place (Macao =1, outside =0) −0.502 (0.288) 0.020 (0.246) 0.022 (0.231) 0.030 (0.230)
Living with older adults or children (yes =1, no =0) −0.095 (0.184) −0.091 (0.155) −0.013 (0.146) −0.018 (0.145)
Education
   Education_2 1.493* (0.748) 1.109 (0.632) 0.671 (0.598) 0.643 (0.595)
   Education_3 1.532* (0.718) 1.071 (0.607) 0.648 (0.573) 0.617 (0.571)
   Education_4 1.810* (0.734) 1.278* (0.620) 0.827 (0.586) 0.772 (0.584)
   Education_5 1.612* (0.737) 1.250* (0.622) 0.817 (0.588) 0.719 (0.586)
   Education_6 1.135 (0.820) 0.767 (0.693) 0.258 (0.655) 0.207 (0.652)
Monthly income
   Monthly income_2 −0.245 (0.245) −0.115 (0.208) −0.132 (0.196) −0.141 (0.195)
   Monthly income_3 −0.562* (0.262) −0.197 (0.224) −0.057 (0.212) −0.067 (0.211)
   Monthly income_4 0.136 (0.358) 0.494 (0.303) 0.605* (0.286) 0.623* (0.285)
   Monthly income_5 −0.077 (0.536) −0.160 (0.454) −0.103 (0.428) −0.070 (0.426)
   Monthly income_6 −0.351 (0.649) −0.822 (0.551) −0.488 (0.520) −0.514 (0.518)
Age 0.306** (0.104) 0.101 (0.090) 0.056 (0.085) 0.046 (0.085)
PBC (perceived ease of vaccination) 0.520*** (0.077) 0.467*** (0.073) 0.519*** (0.075)
VA (COVID-19 attitude) 0.654*** (0.086) 0.572*** (0.082) 0.559*** (0.082)
Conformity trait −0.088 (0.068) −0.100 (0.064) −0.092 (0.064)
SN 0.206* (0.085) 0.134 (0.081) 0.123 (0.080)
Influenza vaccination history 1.277*** (0.146) 1.268*** (0.146)
PBC × VA 0.161** (0.061)
Log-likelihood −1,260.0 −1,153.9 −1,116.8 −1,113.3
Deviance 2,246.0 1,584.6 1,402.5 1,386.2
Df residuals 591 587 586 585
Pseudo R2 (Cragg & Uhler) 0.1282 0.4491 0.5496 0.5582
Obs. Num 608 608 608 608

Data are presented as coefficient (standard error). Significance levels are denoted as ***, P<0.001; **, P<0.01; *, P<0.05. Dependent variable: COVID-19 vaccination timing. Model 1: sociodemographic characteristics (e.g., gender, age, education, income) were entered; Model 2: added variables derived from the TPB (COVID-19 VA, SN, and PBC), along with conformity trait; Model 3: influenza vaccination history was included; Model 4: interaction terms (e.g., PBC × VA) were introduced. COVID-19, coronavirus disease 2019; CR, collective responsibility; df, degrees of freedom; GLM, generalized linear model; Obs. Num, number of observations; PBC, perceived behavioral control; SN, subjective norm; VA, vaccine attitude.

The results reported by Model 4 in Table 2 revealed several significant predictors of vaccination timing. The coefficient for PBC was 0.519 (P<0.001). Similarly, a positive COVID-19 VA showed a strong coefficient of 0.559 (P<0.001). Additionally, the history of influenza vaccination has a coefficient of 1.268 (P<0.001), which is the most powerful predictor. The interaction effect between PBC and VA (0.161, P<0.01) further emphasizes the positive effects. Meanwhile, demographic factors such as gender, living situation, and education levels did not show significant associations, although local residents in Macao (−0.626, P<0.01) were significantly less likely to be in the early vaccination group, compared to non-local residents.

The interaction between VA and conformity trait on vaccination compliance is illustrated in Figure 2.

Figure 2 COVID-19 VA moderated by conformity trait. COVID-19, coronavirus disease 2019; PBC, perceived behavioral control; VA, vaccine attitude.

Discussion

In our study, we found that PBC, COVID-19 VA, and influenza vaccination history are the key predictors of relative early COVID-19 vaccination timing, with SN and conformity trait playing marginal or insignificant roles. Individuals with higher levels of COVID-19 VA, SN, and PBC, lower levels of conformity trait, and those who had received the previous seasonal influenza vaccine tended to take vaccination earlier.

Diverse attitudes toward COVID-19 vaccination timing

As revealed in the cluster analysis of this study, there were diverse attitudes toward COVID-19 vaccination timing, which suggests the need for targeted public health interventions. For instance, Cluster 1 reflects a selective attitude toward vaccination, Cluster 2 represents high vaccination rates, and Cluster 3 demonstrates a cautious approach to COVID-19 vaccines. A study has shown that factors such as age, prior health behaviors (e.g., flu vaccination), and knowledge about the vaccine can mediate the relationship between norms and attitudes, further associated with vaccination uptake (37). These findings underscore the importance of tailored communication strategies and policy frameworks that take socio-demographic factors into account.

Theoretical integration of DOI and TPB and implications for promoting COVID-19 vaccination timing

By incorporating DOI’s focus on innovation adoption, this study enhances our understanding of how attitude and behavioral factors influence vaccination timing, particularly during the early stages of vaccine rollout (12). Previous research on VAs has primarily focused on vaccination’s natural benefits (e.g., vaccine effectiveness and safety) (8,38,39). The present study demonstrated that COVID-19 VA (β=0.559), which encompasses both the natural benefits of vaccination and additional incentives (e.g., time off work, financial rewards, and travel requirements), was among the strongest predictors of vaccination timing (40-42). Within the DOI framework, individuals classified as innovators and early adopters typically exhibit favorable attitudes and strong SNs toward vaccination. Consequently, these individuals are more inclined to pursue vaccination promptly due to their inherent propensity to embrace novel health innovations (12). Meanwhile, the majority of the population tends to rely on observing their peers’ choices as a reference (19,43,44), especially when the community vaccination rate is low. Thus, these behavioral decisions align with the TPB.

Conformity trait can hinder the effect of COVID-19 VA on early adoption of vaccination

We also found that conformity trait was negatively associated with COVID-19 vaccination timing. Notably, participants in the LVG and the UG exhibited significantly higher conformity trait scores than those in the EVG. Previous studies have shown that individuals tend to observe and imitate their peers’ choices, especially when community vaccination rates are low (19,43,44). Our findings further support the view that a higher level of conformity trait may delay the early adoption of COVID-19 vaccines.

Moreover, a significant moderating effect of conformity trait on the relationship between COVID-19 VA and COVID-19 vaccination timing was observed, with the effect being stronger among individuals with lower conformity trait scores. According to trait activation theory (45), specific situational cues can trigger the expression of particular personality traits. Our results suggest that by strategically framing situations or optimizing the context, policymakers might encourage individuals to express their conformity trait in a manner that promotes early vaccination uptake.

Influenza vaccination history can indicate prospective COVID-19 vaccine adopters

The findings of this study indicate that participants who received the previous season’s influenza vaccine exhibited higher COVID-19 vaccination timing scores, classifying them as earlier adopters according to the DOI framework. In line with DOI, successful past adoption experiences can bolster individuals’ confidence in adopting new technologies during subsequent health events (12,21). Meanwhile, the finding supports previous literature extending the TPB, which has identified past vaccination behavior as a reliable predictor of future vaccination uptake (9,46-49).

These findings suggest that an individual’s history of influenza vaccination can serve as an effective indicator of prospective early adopters of the COVID-19 vaccine. A targeted strategy of promoting early COVID-19 vaccine coverage may analyze existing medical records to identify individuals who received influenza vaccinations in previous years.

Additionally, the association between prior influenza vaccination acceptance and COVID-19 vaccination timing indicates that promoting seasonal influenza vaccination could also serve as a preparedness strategy to enhance the uptake of COVID-19 and other pandemic vaccines in the long term (48).

Limitations

This study has several limitations. First, the cross-sectional design limits causal inference between predictors and vaccination timing. Second, PBC measures may conflate intention and capability, as non-intenders might underreport access barriers. Third, self-reported vaccination dates and histories risk recall bias, while unmeasured confounders (e.g., policy shifts) could influence results. Fourth, multiple regression cannot fully address confounding; methods like propensity score matching may better isolate causal effects. Finally, non-probability sampling restricts generalizability beyond tourism workers in East Asian culture.


Conclusions

The findings of this study underscore the critical role of psychological factors and prior vaccination behavior in influencing COVID-19 vaccination timing. To our knowledge, this is the first study to integrate the DOI and TPB frameworks in predicting COVID-19 vaccination timing. Key findings reveal that earlier adopters exhibited stronger VAs, SNs, and PBC, alongside lower conformity traits and prior seasonal influenza vaccination. This underscores the utility of combining DOI and TPB to design targeted vaccination campaigns, as their integrated constructs effectively capture cognitive, social, and innovation-related drivers of behavior.

Notably, conformity traits initially hindered early vaccine uptake but may become advantageous as community vaccination rates rise, suggesting phased messaging strategies (e.g., emphasizing autonomy early, social norms later). Furthermore, influenza vaccination history emerged as a practical indicator for identifying early COVID-19 adopters, advocating for dual-promotion strategies to bolster preparedness for future pandemics.


Acknowledgments

None.


Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1313/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1313/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1313/prf

Funding: This study was financially supported by the Science and Technology Development Fund of Macau SAR (No. 005/2022/ALC), the Self-Supporting Program of Guangzhou Laboratory (No. SRPG22-007), and the Macau University of Science and Technology Foundation 33 (No. FRG-24-040-MSB).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1313/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the School of Business, Macau University of Science and Technology (No. MUST/MSB/2021/03) and informed consent was obtained from all individual participants when they voluntarily participated and completed the questionnaire.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.


References

  1. Ajzen I. The theory of planned behavior. Organ Behav Hum Decis Process 1991;50:179-211.
  2. DiClemente RJ, Salazar LF, Crosby RA. Health behavior theory for public health: Principles, foundations, and applications. Burlington: Jones & Bartlett Publishers; 2013.
  3. Godin G, Kok G. The theory of planned behavior: a review of its applications to health-related behaviors. Am J Health Promot 1996;11:87-98. [Crossref] [PubMed]
  4. Corace KM, Srigley JA, Hargadon DP, et al. Using behavior change frameworks to improve healthcare worker influenza vaccination rates: A systematic review. Vaccine 2016;34:3235-42. [Crossref] [PubMed]
  5. Liu CC, Ling J, Zahry NR, et al. Using the Theory of Planned Behavior to determine COVID-19 vaccination intentions and behavior among international and domestic college students in the United States. PLoS One 2024;19:e0293130. [Crossref] [PubMed]
  6. Wiemken TL, Carrico RM, Kelley RR, et al. Understanding why low-risk patients accept vaccines: a socio-behavioral approach. BMC Res Notes 2015;8:813. [Crossref] [PubMed]
  7. Graupensperger S, Abdallah DA, Lee CM. Social norms and vaccine uptake: College students' COVID vaccination intentions, attitudes, and estimated peer norms and comparisons with influenza vaccine. Vaccine 2021;39:2060-7. [Crossref] [PubMed]
  8. Limbu YB, Gautam RK, Zhou W. Predicting Vaccination Intention against COVID-19 Using Theory of Planned Behavior: A Systematic Review and Meta-Analysis. Vaccines (Basel) 2022;10:2026. [Crossref] [PubMed]
  9. Schmid P, Rauber D, Betsch C, et al. Barriers of Influenza Vaccination Intention and Behavior - A Systematic Review of Influenza Vaccine Hesitancy, 2005 - 2016. PLoS One 2017;12:e0170550. [Crossref] [PubMed]
  10. Zampetakis LA, Melas C. The health belief model predicts vaccination intentions against COVID-19: A survey experiment approach. Appl Psychol Health Well Being 2021;13:469-84. [Crossref] [PubMed]
  11. Rogers EM. Diffusion of innovations. 3rd ed. New York: The Free Press; 1983.
  12. Mo PK, Luo S, Wang S, et al. Intention to Receive the COVID-19 Vaccination in China: Application of the Diffusion of Innovations Theory and the Moderating Role of Openness to Experience. Vaccines (Basel) 2021;9:129. [Crossref] [PubMed]
  13. Zhu W, Zou H, Song Y, et al. Understanding the impact process of vaccine adoption for COVID-19. Hum Vaccin Immunother 2022;18:2099166. [Crossref] [PubMed]
  14. D'Souza GC, Pinto CN, Exten CL, et al. Understanding factors associated with COVID-19 vaccination among health care workers using the Diffusion of Innovation Theory. Am J Infect Control 2024;52:509-16. [Crossref] [PubMed]
  15. Vaccine equity: A fundamental imperative in the fight against COVID-19. PLoS Med 2022;19:e1003948. [Crossref] [PubMed]
  16. Walkowiak MP, Walkowiak JB, Walkowiak D. More Time, Carrot-and-Stick, or Piling Coffins? Estimating the Role of Factors Overcoming COVID-19 Vaccine Hesitancy in Poland and Lithuania in the Years 2021-2022. Vaccines (Basel) 2022;10:1523. [Crossref] [PubMed]
  17. Rydland HT, Friedman J, Stringhini S, et al. The radically unequal distribution of COVID-19 vaccinations: a predictable yet avoidable symptom of the fundamental causes of inequality. Humanit Soc Sci Commun 2022;9:61.
  18. Eroglu B, Nuwarda RF, Ramzan I, et al. A Narrative Review of COVID-19 Vaccines. Vaccines (Basel) 2021;10:62. [Crossref] [PubMed]
  19. Leong C, Jin L, Kim D, et al. Assessing the impact of novelty and conformity on hesitancy towards COVID-19 vaccines using mRNA technology. Commun Med (Lond) 2022;2:61. [Crossref] [PubMed]
  20. Phelps A, Sparks PL, de la Haye K, et al. Perceptions of social circle vaccine acceptance and COVID-19 vaccine initiation among postpriority vaccine eligible adults. Health Psychol 2023;42:276-84. [Crossref] [PubMed]
  21. Scott SD, Plotnikoff RC, Karunamuni N, et al. Factors influencing the adoption of an innovation: an examination of the uptake of the Canadian Heart Health Kit (HHK). Implement Sci 2008;3:41. [Crossref] [PubMed]
  22. Al Breiki M, Al Abri A, Al Moosawi AM, et al. Investigating science teachers' intention to adopt virtual reality through the integration of diffusion of innovation theory and theory of planned behaviour: the moderating role of perceived skills readiness. Educ Inf Technol (Dordr) 2023;28:6165-87. [Crossref] [PubMed]
  23. Chukwuma P. Integrating Diffusion of Innovations and Theory of Planned Behavior to Predict Intention to Adopt Electric Vehicles in Rwanda. European Journal of Business and Strategic Management 2023;8:1-15.
  24. Li KK, Chan MWH, Lee SS, et al. The mediating roles of social benefits and social influence on the relationships between collectivism, power distance, and influenza vaccination among Hong Kong nurses: A cross-sectional study. Int J Nurs Stud 2019;99:103359. [Crossref] [PubMed]
  25. Government of Macao Special Administration Region Statistics and Census Service. Macao in Figures. 2021. Available online: https://www.dsec.gov.mo/en-US/Home/Publication/MacaoInFigures
  26. Government of Macao Special Administration Region Statistics and Census Service. Tourism Satellite Account 2019. 2021. Available online: https://www.dsec.gov.mo/getAttachment/9a0e9b4f-bbd4-4029-acd5-21d8a53e77c1/E_TSA_PUB_2019_Y.aspx
  27. Wang H, Cheong PL, Wu J, et al. Health Literacy Regarding Infectious Disease Predicts COVID-19 Preventive Behaviors: A Pathway Analysis. Asia Pac J Public Health 2021;33:523-9. [Crossref] [PubMed]
  28. Macao SAR Government Health Bureau. COVID-19 vaccine information page. Available online: https://www.gov.mo/en/promotions/covid-19-vaccine-information-page/
  29. Ajzen I. Constructing a theory of planned behavior questionnaire. 2006. Accessed on 8 July 2022. Available online: https://people.umass.edu/aizen/pdf/tpb.measurement.pdf
  30. Mehrabian A, Stefl CA. Basic temperament components of loneliness, shyness, and conformity. Social Behavior and Personality: an International Journal 1995;23:253-63.
  31. Wu J, Chen CH, Wang H, et al. Higher Collective Responsibility, Higher COVID-19 Vaccine Uptake, and Interaction with Vaccine Attitude: Results from Propensity Score Matching. Vaccines (Basel) 2022;10:1295. [Crossref] [PubMed]
  32. Our World in Data. Coronavirus (COVID-19) vaccinations. Available online: https://ourworldindata.org/covid-vaccinations?country=MAC
  33. Hayashi Y, Romanowich P, Hantula DA. Predicting Intention to Take a COVID-19 Vaccine in the United States: Application and Extension of Theory of Planned Behavior. Am J Health Promot 2022;36:710-3. [Crossref] [PubMed]
  34. Zhou R, Horrey WJ, Yu R. The effect of conformity tendency on pedestrians' road-crossing intentions in China: an application of the theory of planned behavior. Accid Anal Prev 2009;41:491-7. [Crossref] [PubMed]
  35. Betsch C, Schmid P, Heinemeier D, et al. Beyond confidence: Development of a measure assessing the 5C psychological antecedents of vaccination. PLoS One 2018;13:e0208601. [Crossref] [PubMed]
  36. Kaufman L, Rousseeuw PJ. Finding groups in data: an introduction to cluster analysis. Hoboken: John Wiley & Sons; 2009.
  37. Eiden AL, Barratt J, Nyaku MK. Drivers of and barriers to routine adult vaccination: A systematic literature review. Hum Vaccin Immunother 2022;18:2127290. [Crossref] [PubMed]
  38. Lin C, Tu P, Beitsch LM. Confidence and Receptivity for COVID-19 Vaccines: A Rapid Systematic Review. Vaccines (Basel) 2020;9:16. [Crossref] [PubMed]
  39. Hagger MS, Hamilton K. Predicting COVID-19 booster vaccine intentions. Appl Psychol Health Well Being 2022;14:819-41. [Crossref] [PubMed]
  40. Klüver H, Hartmann F, Humphreys M, et al. Incentives can spur COVID-19 vaccination uptake. Proc Natl Acad Sci U S A 2021;118:e2109543118. [Crossref] [PubMed]
  41. Giles EL, Robalino S, McColl E, et al. The effectiveness of financial incentives for health behaviour change: systematic review and meta-analysis. PLoS One 2014;9:e90347. [Crossref] [PubMed]
  42. Macau Business Media. More gaming operators advance with vaccination leave schemes. 2021. Available online: https://www.macaubusiness.com/more-gaming-operators-advance-with-vaccination-leave-schemes/
  43. Leng A, Maitland E, Wang S, et al. Individual preferences for COVID-19 vaccination in China. Vaccine 2021;39:247-54. [Crossref] [PubMed]
  44. Liao Q, Ng TWY, Cowling BJ. What influenza vaccination programmes are preferred by healthcare personnel? A discrete choice experiment. Vaccine 2020;38:4557-63. [Crossref] [PubMed]
  45. Tett RP, Guterman HA. Situation trait relevance, trait expression, and cross-situational consistency: Testing a principle of trait activation. J Res Pers 2000;34:397-423.
  46. Brewer NT, Chapman GB, Rothman AJ, et al. Increasing Vaccination: Putting Psychological Science Into Action. Psychol Sci Public Interest 2017;18:149-207. [Crossref] [PubMed]
  47. Galanis P, Vraka I, Katsiroumpa A, et al. First COVID-19 Booster Dose in the General Population: A Systematic Review and Meta-Analysis of Willingness and Its Predictors. Vaccines (Basel) 2022;10:1097. [Crossref] [PubMed]
  48. Leuchter RK, Jackson NJ, Mafi JN, et al. Association between Covid-19 Vaccination and Influenza Vaccination Rates. N Engl J Med 2022;386:2531-2. [Crossref] [PubMed]
  49. Kwok KO, Li KK, Leung CLK, et al. The way forward to achieve high COVID-19 vaccination and revaccination coverage in a city amid a period of tranquility. Front Public Health 2022;10:935243. [Crossref] [PubMed]
Cite this article as: He W, Wu J, Chen CH, Finley A, Wang H, Huang H, Ng C, Chui T, Zhang J, Hon C. Predicting COVID-19 vaccination timing by integrating the theory of planned behavior and the diffusion of innovations: a cross-sectional survey in Macao, China. J Thorac Dis 2025;17(5):2813-2826. doi: 10.21037/jtd-24-1313

Download Citation