This study was conducted in the district of Malang, Jawa Timur, Indonesia, from November 01, 2022, to January 25, 2023, when the government of Indonesia declared COVID-19 an epidemic disease. The COVID-19 booster vaccination program had already been launched in the district. Malang is the second-largest district in East Java Province, with a 2022 population of 2,751,761 people distributed across 390 villages (273 or 70% of which are rural and 117 or 30% of which are urban). Malang has 39 primary healthcare centers (one for every 65,000 people) and 390 village health posts (one for every 7,000 people). Malang classifies 10.15% of its population as “poor or near poor,” compared to 11.46% in all of East Java province (25). The Malang authority carried out its first COVID-19 vaccination program in January 2021 with 2 million doses of vaccine. It is reported that 2,589,407 people have been vaccinated (94.1%) (24). The second and third phases of the vaccination campaign in Malang were carried out in May 2021 and January 2022, with 5.6 million doses of vaccines reaching 2,534,372 individuals (91.2%) and 2,416,046 individuals (87.8%), respectively (18). The percentage of vaccination uptake for the first and second doses in Malang is higher than the national vaccination uptake, which is reported at 86% for the first dose and 57% for the second dose (24).

This cross-sectional study was conducted among individuals aged 15–99 years. The sampling population was determined using a stratified sampling design, with the population stratified into urban and rural areas. Based on a confidence level of 99.0% and a margin of error of 2%, we found the minimum samples for rural and urban areas to be 4,151 and 4,139 respondents, respectively. Initially, 10,050 potential respondents (5,049 in rural areas and 5,001 in urban areas) provided written informed consent and agreed to participate in this study. To encourage participants to participate in the survey, we provided a door prize for 100 randomly selected participants at the end of the survey. Of 10,050 potential respondents, 10,007 were able to complete the survey.

Before data collection, a pretest of the questionnaire was conducted in urban and rural villages at Mojokerto East Java from 12–20 August 2022. The pretest focused on questionnaire content, field editing protocols, the use of data collection apps, and general field procedures. The results of the pilot survey indicated that all respondents could easily understand the questions. Overall, the survey took 30 to 40 min to complete. We used KoboToolbox (a simple, robust, and powerful data collection app) to generate a questionnaire (26). The survey apps were used by 160 trained field researchers in charge of data collection. All recruited field researchers underwent a thorough 2-day training to learn and practice using the survey app.

Figure 1 describes the data collection process in this study. The target population of this study was Malang district people age 15–99 years (N = 2,201,408). The sampling frame of this study was a list of all registered Malang citizens aged 15–99 years who live in 390 villages retrieved from the district citizen registration official report 2021. We applied a stratified sampling design, with the population stratified into urban and rural areas (N target population for urban area = 1,540,986 individuals, N target population for urban area = 660,422 individuals). Based on a confidence level of 99.0% and a margin of error of 2%, we found the minimum samples for rural and urban areas to be 4,151 and 4,139 respondents, respectively.

Trained field researchers in each village contacted potential participants to confirm their willingness to participate in the study. The field researchers were able to identify 10,050 potential respondents (5,049 in rural areas and 5,001 in urban areas) who agreed to participate in this study. Sixth, collecting data: trained field researchers collected data from all potential participants using the survey app. Written informed consent was obtained from all respondents before data collection. Prior to interviewing, respondents were informed about the importance of participating in the survey. Confidentiality and anonymity were also ensured during data collection. Due COVID-19 pandemic situation, the field researchers were equipped with several items of personal protective equipment for COVID-19 protection, including medical masks (N95 3 M Type 9,010), face shields (headgear with clear visor), surgical gloves (Golden Glove latex), and hand sanitizer. Of 10,050 potential respondents, 10,007 were able to complete the survey.

Quality control was done in the field as well as in the University of Brawijaya office. In the field, it was the responsibility of the supervisor and data editor to listen to the recording interview for selected random interviews. In the first two enumeration areas, they had to listen to up to two interviews of each field researcher and thereafter randomly selected interviews. Supervisors also had the responsibility to do observation and verification of 10% of interviews. Verification was done by listening to parts of interview recordings. We also had a team of people in the University of Brawijaya office who listened to random parts of these recordings for random interviews and then compared answers to the electronic data. When discrepancies were found they got back to the teams, generally within the first week of the original interview for field researchers to re-check questionable answers.

The dependent variables in this study were COVID-19 vaccine uptake and COVID-19 booster vaccine uptake. COVID-19 vaccine uptake was measured using the question: “Have you received at least one dose of a COVID-19 vaccine as of today?.” Likewise, COVID-19 booster vaccine uptake was measured using the question: ‘Have you received at least three doses of a COVID-19 vaccine as of today?” These questions were to be answered as either “Yes—have had at least one dose of vaccine/Yes—have had at least three doses of vaccine” or “No—have not had a first dose of vaccine/No—have not had at least three doses of vaccine.” These closed-ended questions indicated vaccine uptake and booster vaccine uptake (27).

The mediating variable was willingness to receive a COVID-19 vaccine. Respondents who responded “No—have not had a first dose of vaccine” were asked the question: “If a COVID-19 vaccine is available, are you willing to receive it?” Respondents were to answer “Willing to receive vaccine,” “Not willing to receive vaccine,” or “Undecided.” Responses of “Not willing to receive vaccine” and ‘Undecided” were assigned as “unwilling” (28, 29).

The main independent variable was knowledge about COVID-19; this was assessed using a six-item questionnaire developed by Zhong et al. (30) and adopted in other similar studies (31, 32). The questionnaire included three questions about the clinical characteristics of the disease (i.e., primary symptoms, availability and effectiveness of treatment, and severity). Two survey questions addressed transmission (i.e., infection through contact with animals and transmission through respiratory droplets), and one item covered prevention and control (i.e., wearing medical masks for prevention). The possible responses were: “Yes,” “No,” or “Do not know.” The knowledge scores were calculated by assigning one point to each correctly answered question and an aggregate score was calculated (range 0–6), with higher scores indicating more knowledge about COVID-19 (31).

We also included comorbidities and sociodemographic factors in the models. Comorbidities were measured through a respondent’s history of hypertension, cardiovascular diseases, diabetes, stroke, autoimmune disease, kidney failure, cancer, gastritis, obesity, chronic obstructive pulmonary disease, and respiratory failure as diagnosed by a medical doctor (33). Sociodemographic factors included gender, age, education, and monthly household income. Each respondent’s educational level was classified based on the Indonesian education system: no schooling, elementary school, junior secondary school, high school, college, and university (28). Household monthly income was classified into four categories based on standard monthly wages in Malang: <1.8 million IDR, 1.8–3 million IDR, 3–4.8 million IDR, and > 4.8 million IDR (34).

To ensure that the sample was representative of people living in Malang at large, we generated descriptive statistics (percentages and 95% confidence intervals [CIs]) for the outcomes using cross-sectional weights. Linear regression was used to examine factors associated with knowledge about COVID-19. Logistic regression was performed to examine the association of knowledge about COVID-19 on vaccine uptake. Generalized structural equation modeling (GSEM) was used to examine whether willingness to receive vaccine mediates the association between knowledge about COVID-19 and vaccination uptake. The maximum likelihood (ML) estimator was used to estimate all models; for the probability model, we reported the odds ratio (OR), 95% confidence intervals (95% CIs), and a two-sided value of p of <0.05 (35). We used Delta, Sobel, and Monte Carlo tests to determine whether the reduction in the effect of the independent variable after including the mediator variable in the model was significant and, therefore, whether the mediation effect was statistically significant. STATA 17.1 was used to clean and analyze the data. Listwise deletion was used to remove missing data from the analyses, allowing each model to include a different number of participants.

The survey was prefaced with a participant information statement and consent form in simple Bahasa (the local language). A trained interviewer read the statement and consent for every participant via the KoboToolbox survey app and confirmed that participants had understood the participant information statement to proceed to the survey; completion of the survey constituted consent. Ethics approval was granted by the Brawijaya University Ethical Board (Reference: 11/EC/KEPK/04/2021).

Table 1 describes the characteristics of the respondents. The percentage of respondents who reported having received at least one dose of a COVID-19 vaccine was 94.8%, while the percentage of those who reported having received at least three doses was 88.5%. These numbers were higher than the national average for COVID-19 vaccine and booster vaccine uptake. The average age of respondents was 43.6 years old [standard deviation (SD) = 15.0], which is slightly older than the average age of the same age range in the district in 2022. In 2022, the proportion of females in Malang’s population was 49.6%, which is slightly higher than the proportion of female respondents in our study (47.9%). The educational status of respondents was similar to the educational status of Malang’s population in 2022: the largest percentage of the population graduated from high school (35.5%). The greatest number of respondents reported a monthly household income of under 1.8 million rupiahs (49.9%). This percentage was also similar to the 49.6% found in the general population of the district in 2022. The proportion of respondents who reported being unemployed was 8.2%, while in 2022 the open unemployment rate in the district was 7.7% (25). Most respondents reported having no comorbidities (91.8%). Among 10,007 respondents, 4.8% reported having hypertension, 0.9% reported having cardiovascular diseases, 1.1% reported having diabetes, 0.5% reported having chronic obstructive pulmonary disease and respiratory failure, and less than 0.5% reported having had a stroke or autoimmune disease, kidney failure, gastritis, and obesity.

Most respondents answered about four of six knowledge items correctly (M = 4.60, SD = 1.1). Respondents appeared to be knowledgeable about transmission through respiratory droplets from infected people (92.9% answered correctly, 1.3% incorrectly, and 5.8% reported that they did not know). The highest prevalence of misunderstanding was discovered in the knowledge item regarding infection through eating or having contact with wild animals (Table 2). Only 24.5% correctly answered that transmission does not occur in this way and that the statement was therefore false, while 59.0% believed that it was true and 16.4% responded that they did not know. Most of the respondents (98.2%) correctly replied that wearing a general medical mask aids in prevention, 0.8% answered incorrectly, and 1.0% did not know.

Knowledge scores varied according to age, educational level, income, and employment status (Table 3). Older respondents were less likely to have accurate information about COVID-19 [β = −0.003, 95% CI = −0.005–(−0.002)]. Respondents who were educated at the elementary, junior secondary, high school, college, and university levels were more likely to have accurate information about COVID-19. Respondents with higher economic status were more likely to have accurate information about COVID-19. Unemployed respondents were less likely to have accurate information about COVID-19 [β = −0.236, 95% CI = −0.313–(−0.160)]. Gender was not related to knowledge about COVID-19.

Table 4 describes the determinants of COVID-19 vaccine uptake. Respondents who had more accurate knowledge about COVID-19 were more likely to be vaccinated (OR = 1.528, 95% CI = 1.428–1.634). Being older (OR = 0.984, 95% CI = 0.977–0.991) and being female (OR = 0.819, 95% CI = 0.676–0.994) were associated with lower COVID-19 vaccination uptake. Respondents with a university education were more likely to be vaccinated than those with no schooling (OR = 2.408, 95% CI = 0.877–6.609). Null associations were found for respondents with elementary school, junior secondary, high school, and college education. Respondents with a monthly household income of more than 4.8 million rupiahs, 3–4.8 million rupiahs, and 1.8–3 million rupiahs were more likely to be vaccinated than respondents with incomes of less than 1.8 million rupiahs. The null association was found for respondents with a monthly household income of 3–4.8 million rupiah. Unemployed respondents were less likely to be vaccinated than employed respondents (OR = 0.348, 95% CI = 0.270–0.448). As expected, most respondents who have comorbidities were less likely to get vaccinated compared to those who did not have comorbidities.

Table 5 shows the determinants of COVID-19 booster vaccine uptake. Respondents who had more accurate knowledge about COVID-19 were more likely to have received booster vaccination (OR = 1.260, 95% CI = 1.196–1.328). Being older (OR = 0.987, 95% CI = 0.982–0.992) and being female (OR = 0.814, 95% CI = 0.715–0.928) were associated with lower COVID-19 booster vaccination uptake. Respondents with high school and university education were more likely to have received booster vaccination than those with no schooling (OR = 1.535, 95% CI = 0.984–2.397; OR = 2.408, 95% CI = 0.877–6.609, respectively). Null associations were found for respondents with elementary school, junior secondary school, and college education. Respondents with a monthly household income of 3–4.8 million rupiahs and 1.8–3 million rupiahs were more likely to have received booster vaccination than respondents with an income of less than 1.8 million rupiahs. The null association was found for respondents with a monthly household income of more than 4.8 million rupiahs. Unemployed respondents were less likely to have received booster vaccination than employed respondents (OR = 0.522, 95% CI = 0.427–0.638). As expected, respondents who had comorbidities were less likely to have received booster vaccination than those who had no comorbidities.

The indirect effects of knowledge (scorekn) on vaccine uptake (vaccineupatke) mediated by willingness to receive vaccination (willvaccine) were significant (Figure 2; Table 6). The Sobel, Delta, and Monte Carlo tests measuring the statistical significance of indirect effects also showed significance (indirect effect = 0.018, SE = 0.002, value of p = 0.000, z-value = 7.123, 95% CI = 0.013–0.023).

Several limitations of this study should be acknowledged. First, the analysis used average knowledge scores, so the specific effects of accurate responses to each individual item were not examined. Second, this study did not extensively explore other attitudinal factors associated with COVID-19 behaviors, such as perceived barriers or other communication factors including information seeking, media usage, and information processing, that may have influenced public knowledge. Third, some of the variables in this study were based on retrospective data, especially regarding respondents’ vaccination uptake and respondents’ histories of comorbidities. These data were thus subject to recall bias. Researchers conducting further studies may wish to use medical record data collected from primary healthcare centers or hospitals to address the issue of recall bias.

This study has important implications for policymakers and health practitioners with valuable insights into how to create an effective strategy to increase COVID-19 vaccination uptake in the district and LMICs with similar health contexts. First, improving knowledge of the vaccine itself, including its efficacy and safety, is not enough to improve COVID-19 vaccine and booster vaccine uptake. Policymakers and health practitioners need to improve public knowledge of COVID-19 in general by acknowledging and discussing their concerns about the disease. Although most of the extant literature has focused on knowledge specifically related to COVID-19 vaccines to improve coverage (55), some studies have highlighted the importance of knowledge about and positive attitudes toward the disease itself in increasing vaccine acceptance (56–58). To tackle vaccine hesitancy and increase uptake, policymakers in Indonesia and other LMICs thus need to design strategies to deliver accurate information not only regarding COVID-19 vaccines but also regarding the disease in general. Furthermore, rumors and misconceptions about COVID-19 and COVID-19 vaccines, especially on social media, should be dismissed and people should be exposed to scientific facts to improve COVID-19 vaccine uptake.

Second, our study shows heterogeneity in COVID-19 vaccine uptake across demographics and socioeconomic characteristics; older people, those with lower levels of educational attainment, those with lower incomes, and those who were unemployed had lower vaccine uptake than others. However, knowledge of COVID-19 and COVID-19 vaccines is lower among individuals with lower socioeconomic status, and providing information to these individuals is more challenging due to limited public health and healthcare services and other infrastructural issues such as the digital divide (58–61). Despite improving accessibility of vaccination programs to those specific socioeconomic groups, introducing more public health strategies to deliver accurate information is thus important to address communication inequalities and design public health communications that will more effectively reduce the existing disparities across segments of the population.

Finally, our findings have important implications for the rollout of booster vaccines. Despite robust immune responses after two doses of COVID-19 vaccine, comorbidities are strongly associated with hyporesponsiveness to COVID-19 vaccination. Booster vaccination is thus required to maintain high levels of protective antibodies in individuals with comorbidities. Our findings showing lower booster uptake among people with comorbidities suggest that interventions to improve access and health literacy need to be provided for these individuals in particular.

In conclusion, this study found significant positive associations between COVID-19 knowledge and vaccine uptake. Our findings suggest that interventions and public health programs aiming to improve knowledge, attitudes, and perceptions regarding COVID-19 vaccination can be implemented to improve vaccine uptake. Furthermore, our findings may contribute to developing a strategy for controlling the COVID-19 pandemic by addressing other determinants of vaccination uptake, including age, gender, and socioeconomic status.