Zero-inflated negative binomial modelling of antenatal care utilization and it’s predictors in Somaliland: evidence from the 2020 demographic and health survey

Background Antenatal care (ANC) is a critical component of maternal health, directly linked to achieving global goals like Sustainable Development Goal 3. However, Somaliland faces significant challenges, with a historically high maternal mortality ratio and a high prevalence of ANC non-utilization. This study aims to determine the level and associated factors of ANC utilization, specifically accounting for the high rate of non-use, among pregnant women in Somaliland. Methods We analyzed data from the nationally representative 2020 Somaliland Demographic and Health Survey (SLDHS). A sample of 3,152 women aged 15–49 who had a recent live birth was included the study. Given the high proportion of zero ANC visits and data overdispersion, a Zero-Inflated Negative Binomial (ZINB) regression model was employed. This method simultaneously models the factors associated with the likelihood of having any ANC visit (zero-inflated part) and the factors influencing the number of visits among users (count part). Results are presented as Incidence Rate Ratios (IRR) for the count model and Adjusted Odds Ratios (AOR) for the zero-inflated part. Results The analysis revealed a low overall utilization rate, with 51.73% (weighted) of women reporting zero ANC visits during their last pregnancy, and a low mean of 1.7 visits. The ZINB model demonstrated a superior fit to the data. Factors significantly associated with an increased number of ANC visits (Count Part, IRR) were belonging to the Middle (IRR = 1.259), Fourth (IRR = 1.259), and Highest (IRR = 1.373) wealth quintiles. Conversely, residence in Sool (IRR = 0.786) and Sanaag (IRR = 0.732) regions, and having 3–4 children (IRR = 0.882) were associated with a reduced number of visits. Factors that significantly reduced the odds of being a non-user (Zero-Inflated Part, AOR) included primary, secondary, or higher educational attainment and mobile phone ownership (AOR = 0.425). Conversely, nomadic residence (AOR = 1.746) was associated with significantly higher odds of having zero ANC visits. Conclusion This study demonstrates that antenatal care utilization in Somaliland remains suboptimal and reveals substantial disparities driven by regional location, nomadic lifestyle, socioeconomic status, education, parity, and access to health information. By applying a zero-inflated negative binomial model, the analysis distinguishes true structural non-users from under-users, offering deeper insight than previous approaches.