This retrospective case-control study was conducted in Moi Teaching Referral Hospital (MTRH) in Kenya. The study was undertaken in Uasin Gishu County which is in the Rift Valley Province of Kenya. This hospital services approximately seven million women [19]. The MTRH has a reproductive health wing called the Riley Mother Baby Unit, which contains both the labour ward and the New Born Unit (NBU). The NBU has three sections: the born before arrival (BBA) unit, the acute ward or critical ward and the general ward [20]. The NBU has several incubators, newborn trolleys, a capacity of 60 beds, and is staffed with consultant paediatricians (10), registrars (6), intern doctors (one) and nurses (30) specially trained for NBU care [20]. According to the MTRH records department the total number of monthly admissions during the study period was about 120-140 newborns [20].
Client flow in the hospital is such that pregnant/post-partum patients are seen in a designated room (triage area) on the labour ward floor by a nurse, medical officer intern and resident doctor. Patients in active labour are usually admitted to the labour ward where they are managed by a team of obstetricians, residents, interns and midwives/nurses. Patients in a latent phase of labour, or with medical conditions, are usually admitted to the antenatal ward, and other patients with medical issues after delivery are admitted to the post natal ward. Patients without complications are discharged. Neonates who have complications after birth are immediately taken to the NBU which is adjacent to the labour ward. The neonatologist and paediatricians subsequently manage the neonate [20].
A medical record review was undertaken on admissions to the newborn unit in MTRH between January 2005 and March 2011. Cases were defined as neonates who were born dead (fetal deaths) or died within seven days of birth (early neonatal deaths). Most recent cases were selected retrospectively until the desired sample size was achieved. Two controls were obtained for every case. The controls were surviving neonates born immediately preceding and following the cases within the first week of life. Exclusion criteria were late neonatal deaths (more than seven days after birth).
A structured data collection instrument was used to collect data from medical records identified from the NBU admission register. It was not possible to blind the data abstractors to case/control status because mortality information was in the medical records.
The primary cause of death was identified using the information from hospital medical records and post mortem reports. Where interpretation was required, the information on the cause of death was verified by the study physician. Data were coded and double entered into two separate password protected databases, which were later compared for consistency. Where there were inconsistencies, the patient’s file was obtained to verify the information.
The outcome of this study was death of the neonate or fetus at birth or within seven days of birth. Explanatory variables were classified as maternal and obstetric characteristics, maternal and obstetric complications, neonatal complications and neonatal characteristics.
Maternal and obstetric characteristics included information on women of reproductive age (15 to 49 years) such as: mother’s age (15-24 years, 25-34 years, or 35-45 years); gravidity categorised as primigravida (1), multigravida (2-4), or grandmultigravida (above 5); qualification of birth attendant (doctor (consultant/registrar/intern), or midwife); gestational age (less than 37 weeks, 37-42 weeks or above 42 weeks); mode of delivery (spontaneous vertex delivery, assisted vaginal delivery, or caesarean section); number of antenatal visits (ANC) (0-1, 2-3, or above 4), and booking status on admission (yes = attended ANC at MTRH or no = did not attend ANC at MTRH).
Maternal and obstetric complications included the presence of common causes of death, as documented in the patient records and post mortem reports; premature rupture of membranes (PROM); dystocia (prolonged or obstructed labour); pre-eclampsia; haemorrhage; and other complications (cardio respiratory diseases, previous scar, Human Immunodeficiency Virus or HIV, malaria, retained placenta, anaemia, abortion). A complication was assumed there was an indication to this effect in the notes, otherwise it was assumed that there were no complications.
Newborn complications included the presence of causes of newborn death as documented in the records including asphyxia, congenital malformation, sepsis, Respiratory Distress Syndrome (RDS), and other complications (hypothermia, diarrhoea, jaundice, hypoglycaemia, meconium aspiration syndrome, and HIV). One single cause of death was identified for each neonate case and where there were multiple causes the final cause of death documented in the post mortem reports was used.
Neonatal characteristics including sex, Apgar score at five minutes and weight in grams were also recorded.
Ethical approval was provided by the University of Newcastle Human Research Ethics Committee (HREC) and the Institutional Research and Ethics Committee (IREC) in Kenya. Permission was obtained from the MTRH administration to undertake the study.
Statistical methods
Statistical analysis was performed using STATA version 10 (StataCorp, College Station, TX, USA). Exploratory data analysis involved checking the data for implausible relationships, outliers and errors using frequency distributions, tables and graphs. Checks comprised visualizing the distributions by use of graphics including histograms overlaid with a normal curve, normal probability plots and box plots to identify potential errors. Bar charts were also used to compare distributions between groups.
All variables were categorical. Categories were combined where cell sizes were small. Bivariate analysis was undertaken using the Chi-squared test to compare characteristics of cases and controls. Initially, a modelling process was undertaken by including all variables with p < 0.2 in separate models for each of the four groups of potential risk factors (maternal and obstetric characteristics, neonatal complications, maternal and obstetric complications and neonatal characteristics). Each group of potential risk factors was analysed in separate multivariable logistic regression. A backward stepwise method was used whereby, at each step, variables with a p-value of >0.1 on the likelihood ratio test were removed. The remaining variables were combined in a final model. Unadjusted and adjusted odds ratios, confidence intervals and p-values are reported for all models.
For a ratio of cases to controls of 1:2, 80% power, a 5% significance level and 40% probability of exposure (i.e. risk factor prevalence) in controls, a sample of 600 neonates (200 cases and 400 controls) was required to detect an absolute difference in risk factor prevalence of at least 12%, or an odds ratio of approximately 0.6 or 1.7.