Modeling women’s reported under-five child mortality in Ethiopia: a comparative evaluation of zero-inflated and hurdle regression models
Under-five child mortality remains an important public health concern in Ethiopia. Understanding factors associated with the number of under-five child deaths reported by women is important for identifying vulnerable populations and informing child survival strategies. However, count outcomes characterized by over dispersion and a large proportion of zero observations require appropriate statistical methods. This study aimed to identify factors associated with the number of reported under-five child deaths among women in Ethiopia and to compare alternative count regression models to determine the most appropriate model for the data. We analyzed data from 16,650 women aged 15–49 years who participated in the nationally representative 2016 Ethiopia Demographic and Health Survey, which used a two-stage stratified cluster sampling design. We modeled the number of child deaths reported by each woman using eight competing count regression models: Poisson, negative binomial, zero-inflated Poisson, zero-inflated negative binomial, zero-inflated generalized Poisson, hurdle Poisson, and hurdle negative binomial. We selected the best-fitting model using the Akaike Information Criterion, Bayesian Information Criterion, and likelihood-based comparisons, and identified determinants using adjusted effect estimates with 95% confidence intervals and p < 0.05. Among the 16,650 women included in the analysis, 8,584 (51.56%) reported no under-five child deaths, whereas 8,066 (48.44%) reported at least one under-five child death. The Hurdle Negative Binomial (HNB) model provided the best fit among the eight competing models (AIC = 37,322.6; BIC = 37,779.49) and was significantly preferred over the zero-inflated negative binomial model (Vuong statistic = 8.63, p < 0.001). In the positive-count component of the HNB model, women aged 35–39 years (IRR = 1.075, 95% CI: 1.023–1.130) and those aged ≥ 40 years (IRR = 2.466, 95% CI: 2.353–2.584) had higher counts of reported under-five child deaths compared with women aged 15–19 years. Secondary education (IRR = 0.817, 95% CI: 0.765–0.872) and higher education (IRR = 0.845, 95% CI: 0.767–0.932) were associated with lower counts, whereas having seven or more household members (IRR = 1.658, 95% CI: 1.592–1.729), not breastfeeding (IRR = 1.198, 95% CI: 1.158–1.240), an unprotected water source (IRR = 1.184, 95% CI: 1.141–1.229), and no toilet facility (IRR = 1.064, 95% CI: 1.023–1.107) were associated with higher counts. In the zero component, education, family size, breastfeeding status, household wealth, marital status, residence, drinking-water source, and toilet facility were significantly associated with the odds of reporting no under-five child deaths. Under-five mortality in Ethiopia is influenced by interconnected maternal, demographic, socioeconomic, healthcare, and environmental factors, with substantial regional variation. The superior performance of the Hurdle Negative Binomial model underscores the importance of accounting for both excess zeros and overdispersion when modeling child mortality. Maternal education, household wealth, breastfeeding, birth spacing, vaccination, place of delivery and access to safe water and sanitation emerged as important determinants of child survival. Integrated and equitable interventions addressing these factors, particularly among socioeconomically disadvantaged and underserved populations, are essential to further reduce under-five mortality in Ethiopia.
Authors
- Desalegn Tamiru
- Abraham Lomboro Dimore (ORCID: https://orcid.org/0000-0002-8429-2746)
- Yibeltal Siraneh (ORCID: https://orcid.org/0000-0003-4416-9572)
- Teklemariam Ergat Yarinbab (ORCID: https://orcid.org/0000-0001-7652-3375)
- Amanuel Mengistu Merera (ORCID: https://orcid.org/0000-0003-2466-463X)
- Addis Birhanu (ORCID: https://orcid.org/0000-0002-7566-1773)
- Daniel Geleta (ORCID: https://orcid.org/0000-0001-6516-0185)
- Buzuneh Tasfa Marine (ORCID: https://orcid.org/0009-0004-0180-9669)
- Gurmessa Tura
- Warkitu Sileshi Ensermu
- Dawit Regassa
- Temesgen Soboka
- Fira Abamech
Institutions
- Jimma University (ET)
- Federal Ministry of Health (ET)
Publication Details
- Journal
- BMC Public Health
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1186/s12889-026-30368-y
- Primary Topic
- Global Maternal and Child Health
- Type
- article
- Field-Weighted Citation Impact
- 0.00