Abstract
Background
Rotavirus is a major cause of childhood diarrheal disease, but evidence on vaccine uptake in Indonesia is limited. During the study period the vaccine was available only privately, on a self-paid basis. This study examined factors associated with uptake before national introduction in 2023, characterizing who is reached when a vaccine is available but not financed.
Methods
We analyzed the fifth wave of the Indonesia Family Life Survey (IFLS-5, 2014–2015), including children aged 1–5 years with caregiver-reported rotavirus vaccination status. Hierarchical binary logistic regression was used. Child age, equivalent to birth cohort in this single-wave survey and thus indexing each child’s eligibility window, was included in every model. Complete-case analysis was primary; its identifying assumption was tested directly, and four sensitivity analyses were added: inverse-probability weighting, a tipping-point analysis, multiple imputation (m = 20), and Firth penalized regression restoring two variables affected by quasi-complete separation.
Results
Among 3,320 children, 287 (8.64%) had received at least one dose. In the fully adjusted complete-case model, higher odds were observed for complete basic immunization (AOR 2.71, 95% CI 1.73–4.24) and postnatal care (AOR 1.57, 95% CI 1.17–2.11), and lower odds for children in Islam-headed households (AOR 0.31, 95% CI 0.21–0.47) and in Bali and Nusa Tenggara (AOR 0.51, 95% CI 0.31–0.84), with higher uptake in Sulawesi than Java (AOR 2.81, 95% CI 1.64–4.81). These four findings were unchanged under multiple imputation and Firth penalization. Associations with male sex, secondary maternal education, maternal age above 30 years and paternal age 41–50 years did not persist under imputation and are not robust. Child age, equivalently birth cohort, was not associated with uptake (AOR 1.02, 95% CI 0.91–1.13), and the religion association differed significantly across regions (
p
= 0.002).
Conclusion
Uptake was low and robustly associated with prior use of preventive child health services and with sociocultural and geographic context. Because the religion association varied markedly across regions, it is interpreted as a marker of local context rather than a causal effect of religious belief. Several weaker associations were not robust to multiple imputation. Findings are associations, not causal relationships.