Explanatory variables

Both individual (neonatal and maternal factor) and facility (community)-level variables were included as a predictor in the model. Sex, gestational age, place of birth, mode of delivery, and assigned cause of death were included as neonatal factors in the model. The medical cause of death was incorporated as an individual-level factor after the underlying cause of death was assigned using the International Classification of Diseases -Perinatal Mortality (ICD-PM) [45]. From the maternal factors, variables such as maternal age, maternal parity, educational status, number of ANC (antenatal care) visits, a score of delay one, and maternal health conditions were included in the model. Maternal health conditions were assigned per the guidance of ICD-PM. The score of delay one, which is a delay in deciding to seek care [46], was computed using the row sum of seven variables included under the domain; namely 1) family poverty,2) bad experience with previous health care,3) failed to recognize the danger of pregnancy,4) unaware where to go,5) had no one take care of other children,6) reliant on traditional practice and 7) lack of decision to a health facility. All of them were binary variables with ‘Yes’ and ‘No’ responses and after summation of the score and to keep the normality of the data a square root transformation was done [47]. Finally, the transformed variable was treated as continuous variables to make a simple and parsimonious model [48].

At a facility (community) level; variables such as residence, type of region, type of health facility, a score of delay two, and a score of delay three were taken into consideration. The type of region was classified into three categories (city, agrarian, and pastoralist) based on the cultural and socio-economic backgrounds of the population [49]. Furthermore, the type of facility was codified into three classes (primary, secondary, and tertiary facilities) according to their manpower, medical equipment, and service provision [50]. Moreover, the score of delay two (delayed arrival to health facility) and delay three (delayed provision of adequate care in a health facility) [46], were computed similarly to the score of delay one. The score of delay two was computed using four items: namely 1) absence of transportation, 2) expensive cost of transportation, 3) no facility within a reasonable distance and 4) poor road condition. Similarly, the score of delay three was also computed using four items; namely,1) long travel time from health facility to health facility, 2) long waiting time before treatment was received, 3) mistake during an assessment, diagnosis, and treatment and 4) shortage of equipment and supplies. Both delays (two and three) were measured using binary variables and the responses were set as ‘Yes’ and ‘No’ options. The delays were included in the final model in a similar fashion as delay one, i.e., they were treated as a continuous variable, to smoothen the model building process.