**Background:** Acute respiratory infections (ARIs) are a leading cause of morbidity and mortality among children under five globally, with an estimated 652,572 deaths due to lower respiratory tract infections in 2016. Ethiopia ranks 5th among the 15 countries with the highest burden of lower respiratory tract infections in children under five. The pooled prevalence of ARI from Ethiopian studies is 18%. While previous work assessed spatial distribution of ARI using 2016 EDHS data, it did not evaluate spatially-varying predictors or incorporate survey weighting. This study aimed to investigate spatial patterns and spatially-varying factors associated with childhood ARI across three waves of Ethiopian Demographic and Health Surveys (2005, 2011, and 2016).
**Methods:** Secondary data from the EDHS 2005, 2011, and 2016 were analyzed. All surveys used two-stage stratified cluster sampling, with enumeration areas selected proportional to size and households systematically sampled. A weighted sample of 31,568 children under five was included (10,109 in 2005, 11,042 in 2011, and 10,417 in 2016). ARI was defined as cough accompanied by short/rapid or difficult breathing based on maternal report. Spatial autocorrelation was assessed using Global Moran's I. Hot spot analysis employed Getis-Ord Gi* statistics. Spatial clusters were identified using Kulldorff's spatial scan statistic with a Bernoulli model (maximum cluster size <25% of population, 9999 Monte Carlo simulations). Spatial regression used a random effects eigenvector spatial filtering (RE-ESF) model and Moran spatially varying coefficient (M-SVC) model, implemented in R version 4.1.3 using the spmoran package.
**Key Results:** The proportion of children with ARI decreased from 12.6% (95% CI: 0.113–0.138) in 2005 to 7.1% (95% CI: 0.061–0.079) in 2011 and 6.6% (95% CI: 0.055–0.077) in 2016, representing a 47.6% decline over the study period. Spatial clustering was not significant in 2005 (Moran's I = −0.011621, p = 0.798) but was significant in 2011 (Moran's I = 0.334486, p < 0.001) and 2016 (Moran's I = 0.091175, p = 0.002). Hot spot analysis revealed consistent high-risk areas in northern Ethiopia across all three surveys. In 2016, hot spots were located in Western, North Western, and Central zones of Tigray; North Gondar, South Gondar, West and East Gojam zones of Amhara; Horo-Guduru, Illiubabur, and Jimma zones of Oromia; and Keffa zone of SNNPR. The primary spatial cluster in 2016 (LLR = 45.3, p < 0.001) was centered at 8.130989 N, 35.637974 E with a 51.36 km radius and relative risk of 6.72, encompassing Ilubabor and Jimma zones of Oromia and Sheka and Keffa zones of SNNPR. The RE-ESF regression model explained 20.5% of spatial variability in ARI (adjusted R² = 0.205). Significant spatial predictors were household use of biomass fuel for cooking (estimate = 0.050, SE = 0.024, t = 2.068, p = 0.039) and no early initiation of breastfeeding within one hour of birth (estimate = 0.047, SE = 0.021, t = 2.239, p = 0.026). The M-SVC model showed these associations were strongest in northern and western regions. Approximately 98% of households (30,980) used biomass fuel for cooking across all surveys.
**Clinical Implications:** The substantial decline in ARI prevalence (47.6%) between 2005 and 2016 suggests that Ethiopia's implementation of the Global Action Plan for the Prevention of Pneumonia and Diarrhoea (GAPPD) and Integrated Management of Neonatal and Childhood Illness (IMNCI) strategies may be contributing to improved child survival. However, persistent geographic disparities—with ARI prevalence ranging from 1% to 20% across districts in 2016—indicate that current interventions are not reaching all populations equally. The strong association between biomass fuel use and ARI in northern and western regions underscores the urgent need for clean cooking energy interventions, particularly given that biomass is burned indoors in unventilated spaces, producing hazardous pollutants. The protective effect of early breastfeeding initiation (within one hour of birth) reinforces the importance of promoting immediate postnatal breastfeeding, especially in high-risk northern regions. The lack of significant associations with vaccination, stunting, wasting, and underweight at the cluster level suggests that ARI risk is multifactorial and that individual-level analyses may be needed to fully understand these relationships. Region- and district-specific policies targeting biomass fuel use and delayed breastfeeding initiation are needed to achieve equitable reductions in childhood ARI across Ethiopia.