The data used in this analysis were derived from the 2023 Indonesian Health Survey (9), conducted by the MOH, Republic of Indonesia, using a multistage systematic random sampling method. Since 2007, the MOH has conducted a cross-sectional household health survey every five years to establish baseline data and monitor health-related indicators at the district, provincial, and national levels. The 2023 survey sample size was designed to allow for district-level estimations (9). A detailed explanation of the survey methodology has been provided elsewhere (9).

The 2023 Indonesian Health Survey included a representative sample of households from 38 provinces and 514 districts or cities. It covered 345,000 general households for the basic health survey and an additional 104,000 households with children under five years of age for the nutritional survey. These households were selected from 34,500 census blocks.

At the individual level, 1,191,692 participants were interviewed. For this analysis, we used data from 3,261 mothers with infants aged 0–11 months residing in eastern Indonesia. The region classified as eastern Indonesia comprises Nusa Tenggara Barat, Nusa Tenggara Timur, Sulawesi, Maluku, and Papua (Figure 1). The data were accessed on 1 December 2024, for the purposes of the current research.

The primary outcome was maternal non-useuse at ANC during the pregnancy in infants younger than 12 months at the time of the survey. The non-use was determined based on the question: “Did you seek ANC from a healthcare provider?” Mothers who did not seek ANC services were assigned a score of 1, while those who attended ANC were assigned a score of 0.

We included 18 potential predictors of ANC non-use, categorized into five main groups: community factors, household characteristics, parental demographics, maternal exposure to health information and healthcare, and maternal obstetric characteristics. We included the child's age at the time of the survey to account for potential recall bias since mothers of older infants may recall ANC visits less accurately, and to account for contextual differences that may influence ANC use.

Two variables were included in the community characteristics: region (West Nusa Tenggara, East Nusa Tenggara, Sulawesi, Maluku, and Papua) and type of residence (urban or rural). In terms of household characteristics, the number of children under five years old (one, two, or three or more) and the household wealth index (richest, rich, middle, poor, poorest) were included. The household wealth index was constructed using Principal Component Analysis16 based on 16 housing characteristics and assets owned by respondents: (1) predominant wall material, (2) predominant floor material, (3) source of drinking water, (4) type of latrine, (5) method of waste disposal, (6) source of electricity, (7) type of cooking fuel, (8) ownership of a washing machine, (9) ownership of a refrigerator, (10) ownership of a mobile phone, (11) ownership of an air conditioner, (12) ownership of a computer, (13) ownership of a flat-screen television (minimum 30 inches), (14) ownership of a motorcycle, (15) ownership of a car, and (16) ownership of at least 10 grams of gold.

For the parental demographic characteristics, we considered four variables: maternal education (academy or university, completed senior high school, completed junior high school, incomplete elementary school), maternal occupation (housewife or not working outside the home, formal worker, informal worker, or student), paternal education (academy or university, completed senior high school, completed junior high school, incomplete elementary school), and paternal occupation (formal worker, informal worker, or unemployed).

Three variables were included in the group of maternal exposure to health information and care: health-seeking behavior (once or more every six months, once a year or less, never), knowledge and awareness about pregnancy danger signs (high, low), and knowledge and awareness about stunting (high, low). To calculate the respondents' level of knowledge about the danger signs of pregnancy, we used the information from seven questions. These included signs such as vaginal bleeding, swelling of the legs, hands, or face accompanied by headache, high fever, convulsions, early rupture of membranes, reduced or no fetal movement, and persistent vomiting or loss of appetite. The correct response for each question was assigned a score of one. The total knowledge score was calculated, and those scoring above the median were classified as having a high level of knowledge, whereas those scoring at or below the median were categorized as having a low level of knowledge.

The knowledge and awareness of stunting was assessed based on four knowledge groups: definition (nine correct answers), causes (six correct answers), impact (six correct answers), and prevention (eight correct answers) of stunting. The total knowledge score ranged from 0 (all responses were incorrect) to 29 (all correct). Respondents with scores above the median were classified as having a high level of knowledge, whereas those with scores at or below the median were categorized as having a low level of knowledge.

Finally, we included six variables for maternal exposure to health information and care: age at childbirth (<20, 20–29, 30–39, ≥40 years), pregnancy order (first pregnancy, second or third pregnancy, and fourth or more), number of childbirths, history of abortion (yes/no), presence of pregnancy complications (yes or no), and pregnancy intention (yes, no, or only then).

First, descriptive statistical methods were used to examine the distribution of all variables included in the analysis. This was followed by bivariate analyses to explore the distribution of these variables according to ANC attendance status. We then used multilevel analysis with two sequential models that incorporated random intercepts to examine the factors associated with the non-use of ANC.

In the multilevel analysis, a three-level logistic regression model was applied to account for the hierarchical structure of the data, with individuals (Level 1) nested within Primary Sampling Units (PSUs) (Level 2), which were further nested within districts (Level 3) and provinces (Level 4). The null model (also called an empty model), which included no explanatory variables, was constructed to examine the extent to which the primary sampling unit, district, and province levels were associated with ANC attendance without adjusting for individual, household, or community-level predictors. The median odds ratio (MOR) was calculated for each level to quantify its contribution to ANC nonattendance.

Following the null model, we constructed Model 1 by introducing all the explanatory variables. This allowed us to assess the extent to which the explanatory characteristics explained the variation in ANC non-use while accounting for the nested structure of the data. We applied backward elimination to remove factors that were not significantly associated with the outcome, using a significance threshold of 0.05. Four variables— the child's age, region, type of residence (urban or rural), and household wealth index—were selected a priori and retained in the final model regardless of their statistical significance. Adjusted odds ratios (aORs) were used to estimate the association between each potential predictor and the study outcome.

The final model presented adjusted odds ratios (aORs) and 95% confidence intervals (95% CIs) for all predictors. This analysis accounted for the complexity of the sample design. Multilevel models were developed using the Stata/MP software (version 14.2; StataCorp, College Station, TX, USA) with a melogit routine.

This study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. Ethical approval for the 2023 Indonesia Health Survey was obtained from the Health Research Ethics Committee of the Health Polytechnic of the Ministry of Health Jakarta II (Komisi Etik Penelitian Kesehatan Poltekkes Kemenkes Jakarta II), under reference number LB.02.01/I/KE/L/287/2023.

Our analysis highlighted the role of environmental, household, and maternal factors in ANC non-use among mothers with infants aged 0–11 months residing in eastern Indonesia. ANC attendance was lower among mothers of younger infants (<6 months) than among those of older infants. At the environmental level, the geographic region emerged as a significant predictor of ANC non-use. At the household level, the household wealth index was significantly associated with ANC use, with lower wealth linked to lower service utilization. Among maternal-level factors, limited health-seeking behavior, poor knowledge about stunting, and absence of pregnancy complications were all significantly associated with non-use. These findings underscore the need for targeted, context-specific interventions to improve ANC coverage, especially among the disadvantaged and underserved populations in eastern Indonesia.

Our findings showed that ANC attendance was lower among mothers with infants aged <6 months than mothers with older infants. This finding may possibly reflect the evolving impact of the Coronavirus Disease 2019 (COVID-19) pandemic. Although we did not have data on the exact timing of pregnancy in relation to the pandemic phases, it is plausible that heightened concern over COVID-19 in early 2022 encouraged greater ANC engagement as a protective health-seeking behavior (9). However, as fear declined by late 2022, care-seeking behavior also appeared to decrease. The transition from emergency to routine health system operations led to the reallocation of resources, staffing shortages, and lingering healthcare worker burnout, which may have disrupted ANC availability (17, 18). The long-term economic effects of the pandemic likely forced some women to deprioritize healthcare visits due to financial constraints (19).

Our study confirmed regional disparities in ANC utilization. Mothers in the Sulawesi, Maluku, and Papua regions were less likely to attend services than those living in Nusa Tenggara Barat. Several potential barriers hinder ANC attendance in these regions, such as geographic isolation, limited health infrastructure, workforce shortages, cultural beliefs, low health literacy, and economic hardship (14, 20). In contrast, it was reported that the Nusa Tenggara Barat benefitted from stronger infrastructure, a well-connected network of community health centers or Puskesmas and hospitals, and targeted maternal programs, supported by better transportation and higher population density (21, 22). We found that ANC attendance in Papua remained among the lowest, which may be attributed to factors such as rugged terrain, dispersed settlements, and a relatively weak health system (22). In particular, Central Papua, a newly established province, and Highland Papua face additional challenges, including administrative and logistical delays in health system development (23). Furthermore, ongoing geopolitical instability and security concerns related to armed conflict in these regions may further limit access to healthcare. In some areas, the presence of extremist groups and intermittent violence has made it unsafe for residents to leave their homes, leading to reluctance to visit primary health centers or other health facilities. These complex barriers likely contribute to the persistently low ANC coverage observed in these provinces. Similarly, women living in Maluku Province might also face challenges related to its archipelagic geography and high inter-island travel costs (24). Our findings showed that even within eastern Indonesia, ANC use varied notably across provinces, indicating that challenges in access are not uniformly experienced. These intra-regional disparities further emphasize the broader gap between eastern and western Indonesia, where the latter benefits from stronger infrastructure and better maternal health outcomes. Addressing these layered inequalities requires locally tailored strategies that respond to both regional and province-specific needs.

The increased likelihood of mothers from the poorest households not attending ANC services indicates the importance of socioeconomic disparities in maternal healthcare access. This is consistent with findings from previous studies conducted in Indonesia and other low- and middle-income countries (10, 14, 25–27). In Indonesia, antenatal care is officially provided free of charge at both integrated health posts, locally known as Pos Pelayanan terpadu or Posyandu, and primary-level facilities, mainly on Puskesmas, particularly the beneficiaries of the Indonesian National Health Insurance (NHI). Despite the high NHI coverage in Indonesia, the uptake among its beneficiaries when accessing health services was reported to be suboptimal (27, 28). This is consistent with our previous study from Maluku Province, which revealed that less than 15% of NHI beneficiaries used NHI when accessing health services.¹ Moreover, although ANC services are provided free of charge, other indirect costs, such as transportation, childcare, and lost income due to time off work, could continue to burden low-income households.

Mothers who did not engage in regular health checkups were less likely to use ANC services. This finding aligns with previous studies (29, 30) indicating that the absence of routine interactions with the healthcare system could lead to low awareness, limited familiarity with medical services, and reduced trust in healthcare providers—all of which might discourage ANC attendance (31).

Our study also confirmed the role of maternal knowledge and awareness in ANC uptake, as previously reported (32). We found that mothers with limited knowledge of stunting were also less likely to attend ANC services, potentially due to a lack of understanding of the role of ANC in preventing stunting and promoting child health. A study conducted in West Java, Indonesia, found that improving maternal knowledge significantly increased interest and motivation to use ANC services (11).

Interestingly, mothers without obstetric complications during pregnancy had a reduced likelihood of attending ANC, as reported in previous studies (30, 32, 33). This suggests that women without complications might perceive less need to seek care during pregnancy. Consequently, mothers who did not have any pregnancy complications might perceive their pregnancies as low risk, leading them to underestimate the importance of ANC services. This perception could result in lower ANC attendance, as they might not recognize the role of routine ANC checkups in detecting and preventing potential complications.

Although not explored in this study, cultural values and local wisdom could influence maternal health-seeking behavior in eastern Indonesia, as reported in several qualitative studies (34–36). Traditional beliefs, such as reliance on family elders or traditional birth attendants, could delay or deter ANC visits. Gender roles, trust in modern healthcare, and local norms also shape service uptake. Therefore, further research to understand and integrate these sociocultural factors is essential to improving ANC uptake and ensuring health programs are culturally appropriate (37–40).

The MOR values from our multilevel analysis revealed important insights into the contextual variation in ANC nonattendance across geographic levels. The high MOR in the null model at the province, district, and PSU levels indicated substantial clustering, suggesting that regional and local factors were significantly associated with ANC nonattendance. After adjusting for the significant explanatory variables, the MOR at the provincial level indicated that the explanatory variables largely explained inter-provincial differences. However, MOR remained high at the district level and increased at the PSU level. This shows persistent unexplained variations at more localized levels. These findings suggest that structural, operational, or contextual barriers, such as uneven health facility distribution, staffing shortages, or access issues at the community level, might continue to limit ANC uptake.

Our findings indicate the need for sustained efforts to improve ANC uptake in eastern Indonesia. Raising community knowledge and awareness of the universal benefits of ANC, regardless of perceived pregnancy-related risk, is essential. Various health promotion strategies have been reported to be effective in improving maternal and child health knowledge, both individual and community-based interventions, such as through home visits (24), group counselling (41), and culturally tailored community outreach (42). Furthermore, health workers should use every contact opportunity to educate women and their families about the importance of early and regular ANC and the services available to support a healthy pregnancy.

Efforts to ensure that ANC is financially accessible are also critical. Policies that reduce the financial burden on pregnant women, such as subsidies, conditional cash transfers, or integration of ANC costs into national health insurance schemes, should be considered (27). Additionally, our findings underscore the importance of strengthening mobile health outreach (e.g., mobile clinics, home visits, or temporary outreach posts) to bring services directly to communities (43, 44). The use of telemedicine services (45), provision of incentives for rural health workers (46, 47), and ensuring reliable transportation support (24) could help overcome geographical and logistical barriers that will improve service accessibility. Future preparedness strategies should also ensure uninterrupted access to essential maternal health services during public health crises such as pandemics. This requires strengthening health system resilience and implementing adaptable service delivery models that could continue operating during future public health emergencies (18, 48). Our study also suggests that efforts to improve ANC utilization must extend beyond individual-level interventions and address health system gaps and service delivery challenges at community and sub-district levels.

These findings highlight the need for regionally tailored health policies to address both inter- and intra-regional disparities in ANC access. Localized approaches are crucial to accommodate the diverse geographical and administrative contexts in eastern Indonesia. Strengthening subnational health systems—particularly in provinces like Central Papua and Maluku is essential. Bridging the development gap between eastern and western Indonesia requires equitable resource allocation and stronger oversight of maternal health programs is important to ensure progress toward national health goals.

This study used nationally representative data with a large sample size, enabling a robust analysis of the factors associated with ANC non-use in Indonesia. A key strength is its focus on the eastern region, where access to healthcare is limited. The findings highlight regional disparities and context-specific barriers and offer valuable insights for targeted interventions. The use of multilevel modelling also allowed for a comprehensive assessment of provincial-, district-, and individual-level influences on ANC utilization. However, this study had several limitations. Its cross-sectional design precludes causal inference, and the lack of variables on service access, availability, and affordability limits the analysis. It also lacks qualitative insights into maternal beliefs and perceptions of care. Remote areas may be underrepresented, and reliance on self-reported data may introduce a recall bias.