Study Design
This study was an unmatched case-control design conducted between October 2016 and May 2017 at Tikur Anbessa Hospital (TAH) and Gandhi Memorial Hospital (GMH). For this study, women with live births during the study period who were willing to participate were randomly selected as controls. Women with antepartum, intrapartum, and early neonatal mortality willing to participate were selected as cases. Neonates who were born at the hospital but died at home after discharge were excluded from this study. This was due to difficulties with follow-up as many patients are transported from regional and district hospitals and do not seek continuity of care at these tertiary hospitals. Neonates delivered outside the hospitals and referred to the neonatology ward at the hospital were excluded from this study. A total of 1077 participants were recruited during the study period: 366 cases and 711 controls. Unfortunately, due to financial and time constraints of the midwife recruiters and data collectors, the study was unable to continue for an entire year and an exact 2:1 ratio for controls to cases was unable to be met.
Study Setting
The study was conducted in the labor wards of two government-funded tertiary hospitals: Tikur Anbessa Hospital (TAH) and Gandhi Memorial Hospital (GMH). Both hospitals serve as teaching hospitals that provide a range of obstetric and gynecological services and accept a large number of referrals from Addis Ababa and the surrounding districts. Thus, both hospitals experience a higher rate of delivery than those in their surrounding areas. Women with high-risk pregnancies may be admitted to the maternity ward before labor for planned delivery. After delivery, neonates who need special care are transferred to the neonatology wards at the Neonatal Intensive Care Unit.
Participants
All women with stillbirth or early neonatal mortality occurring at Tikur Anbessa Hospital and Gandhi Memorial Hospital during the study period willing to participate were included as cases. For every woman with stillbirth or early neonatal mortality, two randomly selected women with live births willing to participate were included as controls. Controls were selected from the general patient population at each hospital immediately after delivery. If for some reason a patient initially selected as control then experienced early neonatal mortality, it was re-categorized as a control.
Data Collection
For data collection, a recruiting team composed of midwives was created under the guidance of the principal investigator. Nine midwives from the labor ward in TAH and GMH were trained as recruiters and data collectors for the study. At least one of these recruiter midwives was assigned to each shift at their respective hospitals. The entire obstetrics and gynecology staff at both hospitals were given presentations on the study at morning meetings. Recruiter midwives were introduced to the staff as leaders of recruitment. The majority of deliveries at the tertiary hospitals are conducted by resident OBGYNs with the assistance of nurses and midwives. If a stillbirth or early neonatal mortality occurred, the recruiter midwife on that shift was consulted. Mothers with stillbirth and early neonatal mortality were given a mourning period followed by a counseling session. After a careful mental, emotional, and physical stability assessment, a midwife-recruiter would approach the mother for the interest and consent of the study. Additionally, the midwife recruiters randomly selected women with live births for the control population and approached them for consent after delivery before discharge. Study controls were selected by a random systematic sampling technique from a list of women who delivered at the hospitals each day. Once randomly selected, the control patients were approached by the midwife recruiter. Data for the study were collected using a pre-designed data collection form by trained nurse assistants.
For both case and control participants, data on potential risk factors were retrieved from medical records including health passports, delivery records, and treatment charts shortly before discharge. The midwives asked for clarification of any incomplete data from the health care worker in charge of the case or the mother herself. Questions were asked of mothers only after they had proper time to mourn, received counseling, and were in a stable emotional, mental, and physical state.
Study Instruments
Data were collected utilizing a data capture sheet composed of three sections. Section 1 captured data on demographic information, prior obstetric history, antenatal care, and complications of current pregnancy and labor. Section 2 captured data on stillbirth and early neonatal cases. All available information from laboratory tests, other investigations, and autopsies are also recorded in this section. If the cause of mortality was unknown, an autopsy was requested only if consent was received from the mother. The purpose of section 2 was to utilize all available information to assign a standard primary cause of mortality to each case based on the Wigglesworth Classification. Section 3 captured data on participants with multiple pregnancies and/or deliveries.
A verbal autopsy instrument modified from the World Health Organization (WHO) instrument for the evaluation of stillbirth and early neonatal mortality was used [11, 12]. Modifications included adjustments for cultural sensitivity and exclusion of irrelevant questions. The final questionnaire was composed of different sections for basic information about the deceased neonate and stillbirth and included both narrative and close-ended questions. The instrument was translated into Amharic language and then re-translated in English to ensure content recording and validity. Pretesting of the instrument was performed to identify potential issues during instrument administration and to drive possible solutions. A senior midwife conducted a verbal autopsy at the hospital. The health care provider who attended the birth did not participate in the interview for the verbal autopsy.
Study Variable Definitions
The following variable definitions were utilized in this study:
- Live birth: the complete expulsion or extraction from the mother of a product of human conception, irrespective of the duration of pregnancy, which shows any evidence of life (ie. Heartbeats, umbilical cord pulsations, breathing, or voluntary muscle movement), regardless of whether the umbilical cord has been cut or the placenta is attached
- Stillbirth: a baby born with no signs of life at or after 28 weeks of gestation
- Antepartum Stillbirth: the death of the fetus before initiation of labor after 28 weeks of gestation as diagnosed by the attending physician
- Intrapartum Stillbirth: the death of the fetus after initiation of labor after 28 weeks of gestation as diagnosed by the attending physician
- Early neonatal death: A neonate born in the hospital but dies before the seventh postnatal day
- Perinatal Mortality: the sum of infant deaths that occur at less than 7 days of age and fetal deaths with a gestational age of 28 weeks or more
Ethical Clearance
Ethical clearance was obtained from the Research and Publication Committee (RPC) of the Department of Gynecology and Obstetrics and Institutional Review Board of the College of Health Sciences, Addis Ababa University (Record# 008/2016). Permission was also obtained from the study facilities to collect data. Participation in the study was completely voluntary and informed consent was acquired from every participant before participation. While this study involved women and newborns which are considered vulnerable populations, several efforts were made in the study design to protect their emotional and mental well-being regarding this sensitive topic. Anonymity and confidentiality of patient personal information were protected through several mechanisms.
Classification of Stillbirth and Early Neonatal Mortality
Stillbirth and early neonatal mortality were classified according to the method described by Wigglesworth et al., which was utilized to identify a single underlying cause of mortality [13, 14]. While in most cases, there are several contributing factors of mortality, the Wigglesworth Extended Classification places events that occurred first in the hierarchy for the cause of mortality than those occurring later. Available health records and verbal autopsies of cases were reviewed separately by two midwife data collectors. Each was assigned a primary cause of mortality based on the Wigglesworth classification system. If there was a consensus between the two midwife-data collectors, the cause of mortality was assigned as such. However, if there was a discrepancy, the principal investigator was consulted and reviewed the case to assign the final cause of mortality. Initially, the study had intended to categorize stillbirths as fresh or macerated; however, much of the data on this was unavailable or undocumented by the midwife recruiters so it was therefore left out of the analysis.
Statistical Analysis
SPSS Statistics Version 26 was utilized for statistical analysis. A comparison of demographic and obstetric characteristics for the case and control population was done using Pearson's chi-squared and fisher's exact test for categorical variables. Comparisons of the mean for quantitative variables between the two populations were done using a t-test with equal variance not assumed. Univariate logistic regression analysis was performed to test the association between variables and stillbirth. For those variables that showed a significant association in the univariate analysis, multivariate logistic regression was created to adjust for confounders. The variables investigated in the multivariate model were education, marital status, blood type, parity, previous stillbirth, previous preterm birth, previous child with congenital abnormalities, antenatal care for current pregnancy, hypertensive disorder during pregnancy, antepartum hemorrhage during pregnancy, PROM during pregnancy, method of delivery, and multiple births. Blood type was included as a variable because some studies have demonstrated an association between blood type and bleeding status, specifically postpartum hemorrhage [15, 16]. A data replacement method for missing variables was not utilized because of missing data composed less than 5% per variable.