Perinatal death in Northern Uganda: incidence and risk factors in a community-based prospective cohort study

Background: Perinatal mortality in Uganda remains high at 38 deaths/1,000 births, an estimate greater than the every newborn action plan (ENAP) target of ≤24/1,000 births by 2030. To improve perinatal survival, there is a need to understand the persisting risk factors for death. Objective: We determined the incidence, risk factors, and causes of perinatal death in Lira district, Northern Uganda. Methods: This was a community-based prospective cohort study among pregnant women in Lira district, Northern Uganda. Female community volunteers identified pregnant women in each household who were recruited at ≥28 weeks of gestation and followed until 50 days postpartum. Information on perinatal survival was gathered from participants within 24 hours after childbirth and at 7 days postpartum. The cause of death was ascertained using verbal autopsies. We used generalized estimating equations of the Poisson family to determine the risk factors for perinatal death. Results: Of the 1,877 women enrolled, the majority were ≤30 years old (79.8%), married or cohabiting (91.3%), and had attained only a primary education (77.7%). There were 81 perinatal deaths among them, giving a perinatal mortality rate of 43/1,000 births [95% confidence interval (95% CI: 35, 53)], of these 37 were stillbirths (20 deaths/1,000 total births) and 44 were early neonatal deaths (23 deaths/1,000 live births). Birth asphyxia, respiratory failure, infections and intra-partum events were the major probable contributors to perinatal death. The risk factors for perinatal death were nulliparity at enrolment (adjusted IRR 2.7, [95% CI: 1.3, 5.6]) and maternal age >30 years (adjusted IRR 2.5, [95% CI: 1.1, 5.8]). Conclusion: The incidence of perinatal death in this region was higher than had previously been reported in Uganda. Risk factors for perinatal mortality were nulliparity and maternal age >30 years. Pregnant women in this region need improved access to care during pregnancy and childbirth. This was a prospective cohort study among pregnant women recruited at ≥28 weeks and followed up for the first 50 days of life. It was nested in the Survival Pluss trial (NCT0260505369), a cluster-randomized community-based trial. The more detailed description of this study can be found in a previous publication [27]. All pregnant women in the study were identified by community volunteers and followed up until 1 week after birth. The study was conducted in Lira, a district in Northern Uganda, between January 2018 and March 2019. Lira district has 3 administrative counties, 13 sub-counties, 89 parishes and 751 villages. The district had a population of 410,000 in 2014 [28], served by 31 healthcare facilities, including one referral hospital, 3 healthcare centres with operation rooms, 17 healthcare centres with maternity ward but no surgical facilities, and 10 healthcare centres with only out-patient services (dispensary). The main economic activity in the region is subsistence farming. The study was carried out in 3 sub-counties, Aromo, Agweng and Ogur. These sub-counties were chosen based on the poor maternal and perinatal indicators and the location in a rural and hard-to-reach area of the district. Participants were identified by 250 community female volunteers who had received training on ethical conduct and record keeping. They contacted the research team via mobile communication whenever they identified a pregnant woman in their communities. A research assistant, accompanied by the community volunteer, then visited the pregnant woman at home; she was included if she was found to be at least 28 weeks pregnant, resident in the study area and willing to participate in the study. The gestational age was determined based on the last normal menstrual period. Participants were recruited irrespective of antenatal care utilization. Follow-up phone calls and visits were made to ensure that the mothers were visited within 24 hours after childbirth and 7 days postpartum. During recruitment, pregnant women who were likely to leave the study area before the end of 6 months and those with psychiatric illness were excluded. For this study, the required sample size of 1,812 was estimated using Fleiss statistical methods for rates and proportions [29]. It was assumed that 35% of the pregnant women who experienced a perinatal death had no formal education and 28% with perinatal death had a secondary education [30]. This estimation factored in 0.05 alpha, 80% power and 10% non-response. A team of 42 trained research assistants, fluent in the local language Lango collected the data using a standardized questionnaire in face-to-face interviews conducted at the participant’s home. The standardized questionnaire, with structured questions on demographic, socio-economic and current pregnancy was administered at recruitment. The sections on birth and perinatal survival status were administered within 24 hours after childbirth and at 7 days postpartum. Research assistants (university graduates) underwent a one-week intensive training on data collection procedures and tools before deployment in the field. This training was facilitated by doctors, nurses and a data analyst. The research assistants that were in two groups lived within the community in Aromo and Agweng sub-counties. They were given mobile cell phones and motorcycles to make timely follow-up visits. Data were checked on a daily basis by the coordinators and supervisors for completeness and consistency before submission. A verbal autopsy was carried out in cases of perinatal death, using standard verbal autopsy questionnaires developed by the WHO [31]. A verbal autopsy was taken within 2 weeks if relatives felt able and willing to provide information. Deaths were grouped according to timing; if it occurred during the antepartum period, intrapartum period or in the early neonatal period. The cause of death was assigned by two paediatricians after independently reviewing the collected data. A consensus was reached on the verbal autopsy data collected to determine the probable cause of death, which were later assigned by one of the authors to the new International Classification of Death – Perinatal Mortality (ICD-PM) grouping [32]. All data collection tools were translated in Lango, pretested, and adjusted as necessary. The primary study outcome for this study was perinatal death. According to the WHO definition, a stillbirth was a baby born with a gestation of 28 weeks or more and an early neonatal death was a live-born baby that died within the first 7 days of life. Perinatal death was used as an umbrella term for both stillbirth and early neonatal death [33]. The birth weight of stillborn infants was not used to determine gestational age, as this was culturally unacceptable in the study area. Therefore, gestational age was determined based on the last normal menstrual period. The perinatal mortality rate was defined on the WHO definition as the number of perinatal deaths per 1,000 births included in the study (≥28 weeks) [18]. The stillbirth rate was defined as the number of stillbirths per 1,000 total births [18]. The early neonatal death rate was the probability that a child born alive within the study period died during the first 7 days after birth, expressed per 1,000 live births [34]. Twin pregnancies were counted as one (last twin), irrespective of the number who died. An asset-based wealth index was used to estimate the economic status of the participant’s household, using the principal component analysis. The ‘wealth index’ was computed using the first principal component and based on the availability of nine different household assets. The wealth index was later reduced into three groups; lowest 40%, middle 40% and top 20%. Maternal age was categorized as <20 years, between 20 and 30, and >30 years. Marital status was regrouped as married or unmarried. Maternal education was categorized as no education, only primary education, and secondary education or higher education. Parity (at enrolment) was defined as the number of deliveries a woman had had and grouped as Para 0 (nullipara), Para 1–4 (multipara) and Para 5+ (Grand multipara). Antenatal care utilization was collected as ‘yes/no’ at enrolment. Obesity was defined as a body mass index (BMI) >30 kg/m2. Participants’ place of residence was included to assess for the difference in perinatal mortality between the two administrative divisions (sub-counties). A variable representing the intervention or the control arm of the parent trial was included in the analysis as a potential confounder. Data were collected using an android-based mobile application (Open Data Kit: https://opendatakit.org) and analysed using STATA version 14.0 (StataCorp; College Station, TX, USA). Categorical variables were summarized as proportions and continuous variables as means with their standard deviations as appropriate. We used generalized estimating equations of the Poisson family, with a log link, taking into account clustering, and assuming an exchangeable correlation to calculate the risk ratios estimating the magnitude of any associations between exposure variables and perinatal death. Based on scientific literature and biological plausibility, we included the following factors into our model: maternal age, marital status [10,11], maternal education, wealth index, antenatal care attendance [9], parity [11,13,14], intervention of the community trial [12] and maternal obesity [15,35]. We also included the place of residence to assess whether perinatal death varied in the two sub-counties represented in our study. All the variables in the model were assessed for collinearity, which was considered present if the variables had a variance inflation factor (VIF) of >10. In situations of collinearity, we retained the variable with the greater biological plausibility. Multivariable regression analysis was used to take into account potential confounding.