Procedures and outcomes

Data were collected on maternal socioeconomic status, maternal and obstetric history, infant clinical conditions, imaging assessment, and microbiology of blood and cerebrospinal fluid specimens using predefined case report forms. Post-mortem examinations (both complete diagnostic autopsy and MITS) were done with additional consent from the parent or legal guardian. Following additional training, the MITS process was introduced in the second half of the study period. The clinical management of recruited patients followed the national guidelines developed by the Federal Ministry of Health of Ethiopia.¹⁵

To establish cause of death for each case, a panel consisting of international and national experts in neonatology, obstetrics, and pathology (including RP, SD, RLG, and HB) met face to face every 6 months and reviewed the deaths. For each death, the available clinical, laboratory, MITS, and autopsy data, including the sequence of clinical events during the infant's hospital stay, were available to panellists to establish the primary and contributory fetal or neonatal and maternal causes of mortality using a protocol.¹⁴

The panel was divided into pairs comprising an Ethiopian and international clinician; each pair established cause of death for a set of cases. Criteria were prospectively defined for the primary cause of death and prioritisation when more than one possible cause was present.¹⁶ The panel assigned the primary cause of death using ICD-10.¹² The primary or underlying cause of death is the “disease or injury that initiated the train of events leading directly to death, or circumstances of accident or violence which produced the fatal injury”, unlike the immediate cause of death, which was defined as “the disease or complication which directly preceded or directly led to death”. If the two panel members were unable to reach consensus, the case was presented to the full expert panel for adjudication. If insufficient or conflicting information was present, the panel could decide that the cause could not be established with the available data. In addition to the primary fetal or neonatal cause of death, the panel also identified one or more neonatal and maternal conditions that contributed to each death.

Gestational age was established using a hierarchy of three methods: ultrasound before 28 weeks of gestation when available, mother's report of last menstrual period when judged reliable, and the New Ballard Score completed at younger than 7 days of age.¹³ The so-called best gestational age, which was used for the study purposes to document whether the infant met the criteria of preterm (<37 weeks gestation at delivery) was calculated using the hierarchy: (1) ultrasound at younger than 28 weeks of gestation; (2) ultrasound at more than 28 weeks of gestation with agreement of a reliable last menstrual period or New Ballard Score; (3) reliable last menstrual period and New Ballard Score; (4) if discrepancy between last menstrual period and New Ballard Score was greater than 2 weeks, the last menstrual period date was used; and (5) if no ultrasound and last menstrual period estimate was not reliable, New Ballard Score was used.¹³ If the infant was more than 37 weeks or if the gestational age could not be established with these methods, the infant was excluded from the study.

For follow-up of recruited infants, all hospitalised infants were assessed daily using a standardised tool and those at home were monitored once a week, either through a face-to-face meeting or by telephone. The research nurses and doctors underwent initial training on the case report forms, and their skills were reinforced every 3 months and with on-the-spot assessments. Each hospital team held once a week meetings to discuss and resolve challenges.

In the study hospital NICU, infants were evaluated once to twice per day by a research nurse who documented the findings on case report forms. A supervisor monitored the collection of data and biospecimen for laboratory evaluation. Relevant laboratory, radiology, and pathology personnel were informed about the samples sent. Visits to each hospital were done regularly by the investigators to check data quality and provide support as necessary.