Modelling factors associated with infant mortality in Northern Region of Ghana: survival analysis approach

Infant mortality remains a significant public health concern in Ghana, particularly in the Northern Region. This study investigated factors associated with infant survival using facility-based data from Kpandai Polyclinic and Tamale West Hospital. The Kaplan-Meier estimator showed no statistically significant survival differences across most categorical covariates, including child sex, maternal education, place of residence, marital status, and occupation; however, a significant difference was observed for mode of delivery. The mean gestational age at delivery was 35.6 weeks, which is below the standard full-term threshold, suggesting a relatively high prevalence of preterm births within the cohort. The standard Cox proportional hazards model identified gestational age and mode of delivery as significantly associated with infant mortality. However, the directions of these associations were counterintuitive and should be interpreted with caution. In particular, the association with mode of delivery is likely influenced by confounding by indication, as caesarean sections are typically performed in response to complications or high-risk pregnancies. Similarly, the effect of gestational age may be influenced by measurement limitations and the absence of detailed clinical variables, such as neonatal complications and birth conditions, which may obscure the true relationship between gestational age and survival. The piecewise time-dependent Cox model revealed that only a few variables particularly mode of delivery, gestational age, and place of residence exhibited notable time-varying associations, whereas the majority of demographic and socioeconomic factors remained largely non-predictive. The extended model provided a superior fit to the data compared to the standard Cox model.