Cause of neonatal deaths in Northern Ethiopia: a prospective cohort study

A prospective cohort study including live births between 1 April 2014 and 31 July 2014 was conducted in seven randomly selected hospitals in Tigray Region of Ethiopia.

The Ethiopian health care has a decentralized four-tier system consisting of primary health care, district, zonal, and specialized hospitals. The primary health care includes rural health posts, nested into health centers serving 25,000 populations. District hospitals are expected to serve catchment populations of 250,000. Zonal and specialized hospitals serve 500,000 and five million people, respectively. There are a total of 15 public hospitals: six zonal, one specialized and eight district hospitals in Tigray region. Comprehensive maternal and newborn services, including cesarean section and delivery services, are provided free of charge.

In the present study, seven hospitals were randomly selected from the 15 hospitals using a lottery method [18, 19].

All mothers who had a live birth in the study hospitals during the study period were eligible for inclusion. Exclusion criteria included those arriving at the hospitals after six hour of delivery and/or had a psychiatric illness.

This is a sub-study undertaken to estimate NMR in the region. The study included 1162 mother-neonate dyads. Details of sample size calculation and sampling procedure have been published previously [20].

Data related to neonates, mothers, and pregnancies were collected using a pretested structured questionnaire. The International Classification Disease Version 10 (ICD-10) verbal autopsy questionnaire was adapted to collect verbal autopsy data [21].

Data were collected by midwifes working in the respective hospitals. The survival status of neonates after discharge was ascertained by making follow-up phone calls every week to the end of the neonatal period. Those who could not be reached by phone were visited by health extension workers (i.e., primary health workers who live in the community) every week to ascertain the status of the neonate who then notify the outcome to the study midwife. Clinical characteristics related to the mother and her baby were either extracted from medical record or collected at baseline. Details of follow up have been published previously [20].

For neonates who died at home after discharge, probable causes of death were assigned by physicians using the ICD-10 coding [21]. For neonates who died in the health facility, cause of death was assigned by physicians, who work in the study hospitals, based on national guidelines [22] and the Wigglesworth classification with the revised decision tree modified from the neonate and stillbirth single cause of death classification [23, 24] and a criteria derived and slightly modified from elsewhere [25] to assign single cause of death for all neonates. When death occurred at home, data related to the illness were recorded using the standard WHO verbal autopsy questionnaire and supplemented with clinical hospital records from delivery and discharge to provide additional information on the cause of death. Two physicians independently reviewed the completed verbal autopsy questionnaires to assign underlying cause of death using the ICD-10coding [21]. Physicians assigned up to three underlying causes using the three digit ICD-10 coding. Consensus was sought to establish a cause of death. When disagreement arose, the verbal autopsy form was sent to a third independent physician and cause of death was reported if this physician agreed with either of the two physicians. In instances when no agreement was reached, the principal cause of death was reported as ‘unknown’.

Data collectors received one day training on data collection, on how to deal with the hospital-based clinical diagnostic data, the verbal autopsy form, and the recruitment and follow- up of mothers with their neonates. Physicians were also trained for one day on how to review the questionnaire in order to assign ICD-10 code for cause of neonatal death.

Birth weight was measured using standard weight measuring balances used in the study hospitals. Low birth weight was defined as a birth weight of <2500 g, and normal 2500 g to 4000 g, and macrosomia ≥ 4000 g. Gestational age was determined based on the mother’s last menstrual period. Probable causes of death were defined as: 1) prematurity if the neonate was born before 37 completed weeks, had surfactant deficiency, intraventricular hemorrhage, respiratory distress syndrome, and/or necrotizing enterocolitis; 2) asphyxia included a failure to initiate spontaneous respirations and/or Apgar score of less than 3 in the first 5 min from delivery, and neonatal hypoxic/ischemic encephalopathy yielding an Apgar score at 5 min of less than 7;3) infection included any possible source, such as sepsis, pneumonia, meningitis, and/or respiratory tract infections; and 4) congenital abnormality included a potentially lethal congenital malformation (e.g., congenital cardiac defect, spina bifida, congenital syndromes, gastro-intestinal malformation). Further, “unknown” cases included deaths which did not have a specific cause, have little information, or no determined cause based on the ICD-10 and hospital-based clinical diagnosis.

In this study, early neonatal mortality was defined as the probability of death before 7 completed days of life and late neonatal mortality defined as the probability of dying between 7 completed days and before 28 completed days. Similarly, the cause specific mortality rate was defined as death of a neonate during the first 28 days of life attributed to specific cause per 1000 live births.

Reported income was categorized into low (<500 Ethiopian birr (ETB)), medium (500–1500 ETB) and rich (>1500 ETB). Residence was classified as ‘urban’ and ‘rural’. Urban is defined as a locality with 2000 or more whose inhabitants are primarily engaged in non-agricultural activities. All participants not meeting the urban definition were considered ‘rural’.

Educational status was first categorized as ‘unable to read and write’, ‘able to read and write but no formal education’, ‘grade 1–4’, ‘grade 5–8’, ‘grade 9–12’, ‘college and above’. During analysis, we merged the select categories due to small sample size. Specifically, our final categories were ‘unable to write & read or no formal education’, ‘primary education (grades 1–8)’, ‘secondary education (grades 9–12)’ and ‘tertiary education (college and above)’.

Intermediate factors included maternal, neonatal, and health service related considerations. These were operationalized as gender of the neonate categorized as female/male; distance from respondents’ home to the health facility (measured or calculated in walking time converted to kilometers(km) reported as <5 km, 5–10 km and > =10 km distance).

Descriptive statistics included mean, standard deviation, frequency and associated percentages where appropriate. Associations between sociodemographic factors and cause of death were examined using chi-square. We classified each neonatal death in the hierarchical order of prematurity, asphyxia related, infection, congenital abnormalities, and other (unknown) cause of deaths. Data cleaning and analysis was done using Stata™ Version 12.

Responses were obtained from 1162 mother-neonate dyads (see Table 1 for demographic details). Ten observations were excluded from the analysis due to incompleteness of the data. The mean (±standard deviation) maternal age at delivery was 27 ± 5.4 years. About a third of the live births (32%) were to teenage mothers (under age of 18 at delivery). Of the total included 1152 live births, 68 neonates died yielding a NMR of 63 per 1000 live births. The majority of mothers [996 (86%)] were followed by phone call, 112 (10%) were followed by HEW, and the remaining 44 (4%) died while in the study hospitals before discharged. There were 60 dyads lost to follow up, withdrawal, or refusal.

Of the total neonates, 121 (10.5%) were low birth weight. Of the neonates who died, 46 (68%) died in health facilities. Of the total neonates, 61 (5.3%) were hypothermic, with 97% of these infants controlled with immediate interventions (see Table 2).

We found the leading causes of death to be prematurity [23 (34%)], and asphyxia [21 (31%)] which accounted for 2 of every 3 deaths. The remaining deaths were caused by infections 8 (12%), congenital abnormalities 5 (7%), and other causes 11 (16%). Three-fourths (51) of the total deaths took place during the early neonatal period, primarily due to prematurity and asphyxia. All neonates who died due to congenital abnormalities [5] did so during the early neonatal period. In the late neonatal period, asphyxia, infection, and prematurity were the leading causes of death at 35, 23, and 23% respectively. Though the absolute number of deaths due to infection was the same in both periods (4), the percentage of deaths from this cause was higher in the late (23%) than in the early (8%) neonatal period. Overall, prematurity was the highest reported cause of death in 20 per 1000 live birth whilst congenital abnormality was the least reported cause at 4 per 1000 live births (Table 3).

There were no significant difference in distribution of deaths between males and female neonates (P = 0.943). Infection was higher among newborns from teenage mothers and all infections were occurred among mothers who lived in the rural areas. A substantial number of deaths [13 (43%)] in the urban areas were caused by asphyxia. Nearly half [12 (47%)] of premature infants were found in mothers of the lowest reported monthly income category. On Pearson chi-square analysis of socio-demographic characteristics with cause of death, age at marriage (P = 0.004), residence (P = 0.05) and distance (P = 0.043) were found to be significantly associated (Table 4).