Predictors of low birth weight and preterm birth in rural Uganda: Findings from a birth cohort study

Background Approximately 20.5 million infants were born weighing <2500 g (defined as low birthweight or LBW) in 2015, primarily in low- and middle-income countries (LMICs). Infants born LBW, including those born preterm (<37 weeks gestation), are at increased risk for numerous consequences, including neonatal mortality and morbidity as well as suboptimal health and nutritional status later in life. The objective of this study was to identify predictors of LBW and preterm birth among infants in rural Uganda. Methods Data were derived from a prospective birth cohort study conducted from 2014-2016 in 12 districts across northern and southwestern Uganda. Birth weights were measured in triplicate to the nearest 0.1 kg by trained enumerators within 72 hours of delivery. Gestational age was calculated from the first day of last menstrual period (LMP). Associations between household, maternal, and infant characteristics and birth outcomes (LBW and preterm birth) were assessed using bivariate and multivariable logistic regression with stepwise, backward selection analyses. Results Among infants in the study, 4.3% were born LBW (143/3,337), and 19.4% were born preterm (744/3,841). In multivariable analysis, mothers who were taller (>150 cm) (adjusted Odds Ratio (aOR) = 0.42 (95% CI = 0.24, 0.72)), multigravida (aOR = 0.62 (95% CI = 0.39, 0.97)), or with adequate birth spacing (>24 months) (aOR = 0.60 (95% CI = 0.39, 0.92)) had lower odds of delivering a LBW infant Mothers with severe household food insecurity (aOR = 1.84 (95% CI = 1.22, 2.79)) or who tested positive for malaria during pregnancy (aOR = 2.06 (95% CI = 1.10, 3.85)) had higher odds of delivering a LBW infant. In addition, in multivariable analysis, mothers who resided in the Southwest (aOR = 0.64 (95% CI = 0.54, 0.76)), were ≥20 years old (aOR = 0.76 (95% CI = 0.61, 0.94)), with adequate birth spacing (aOR = 0.76 (95% CI = 0.63, 0.93)), or attended ≥4 antenatal care (ANC) visits (aOR = 0.56 (95% CI = 0.47, 0.67)) had lower odds of delivering a preterm infant; mothers who were neither married nor cohabitating (aOR = 1.42 (95% CI = 1.00, 2.00)) or delivered at home (aOR = 1.25 (95% CI = 1.04, 1.51)) had higher odds. Conclusions In rural Uganda, severe household food insecurity, adolescent pregnancy, inadequate birth spacing, malaria infection, suboptimal ANC attendance, and home delivery represent modifiable risk factors associated with higher rates of LBW and/or preterm birth. Future studies on interventions to address these risk factors may be warranted. Study approval was obtained from the Makerere University Research Ethics Committee at the School of Public Health in Kampala, Uganda; the Uganda National Council for Science and Technology in Kampala, Uganda; the Tufts Health Sciences Institutional Review Board in Boston, MA; and the Institutional Review Board at Harvard T.H. Chan School of Public Health, Boston, MA. Before enrollment into the study, written informed consent was obtained from all participants. The Uganda Birth Cohort Study (UBCS, {“type”:”clinical-trial”,”attrs”:{“text”:”NCT04233944″,”term_id”:”NCT04233944″}}NCT04233944) was a prospective birth cohort study conducted from 2014–2016 in 12 districts/16 sub-counties in rural northern and southwestern Uganda. The study, which enrolled 5,044 pregnant women, was designed to assess the impact of the Uganda Community Connector Program (UCCP), a five-year agriculture, livelihoods, and nutrition program funded by the United States Agency for International Development (USAID) which aimed to improve the nutritional status of women and children and the livelihoods of vulnerable populations in rural Uganda. The enrollment period for the UBCS lasted approximately 12 months. Eligible women, who were identified as pregnant from a urine pregnancy test administered by village health team (VHT) workers, were referred to study staff for enrollment into the main study. Following enrollment, which occurred primarily during the second or third trimester of pregnancy, mother-infant pairs were prospectively followed every three months until infants reached six or nine months of age. Data collected in the UBCS included information on demographics and household characteristics [e.g., water, sanitation and hygiene (WASH) practices, food security, agricultural production; and gender dynamics], maternal dietary intake and diversity, pregnancy history and outcomes, breastfeeding and complementary feeding, and infant morbidity and mortality. Maternal and infant anthropometry, including infant birth weight, were also collected. Pregnant women 15–49 years of age were eligible to participate in the UBCS if they planned to reside in the study area through the completion of follow-up and provided written informed consent. The inclusion, exclusion, referral, and termination criteria for the UBCS are presented in S1 Table. The target enrollment for the UBCS was 5,152 pregnant women (i.e., 322 pregnant women in each of the 16 participating sub-counties). This sample size allowed for a detection of a 0.14-unit difference in child length-for-age Z-score (LAZ) (the primary outcome variable of the parent study) with 80% power and a 0.05 level of significance, assuming 30% attrition between enrollment of pregnant women and delivery for reasons such as maternal death, fetal loss, household migration, temporary relocation of the mother for delivery, withdrawal, and loss to follow-up. Furthermore, it assumed 10% attrition among live births between delivery and completion of follow-up. Fig 1 shows the study profile for this analysis. In total, 5,044 women met the eligibility criteria and were enrolled into the UBCS. Of these, women were excluded from this analysis if they had a missing enrollment visit (n = 95) or a missing birth visit (n = 851). Furthermore, they were excluded if their infant was not born alive (n = 120) or if they had a multiple birth (n = 8). Women were excluded from the LBW analysis if birth weight of the newborn was not recorded within 72 hours (n = 633) after birth and from the preterm birth analysis if gestational age data were missing (n = 129). After exclusion criteria were applied, a total of 3,337 women were included in the LBW analysis and 3,841 in the preterm birth analysis. S2 Table presents the breakdown of enrollment by region, district, and sub-county for both the UBCS and this analysis. The UBCS questionnaires consisted of 13 modules which were programmed onto handheld Android devices using Open Data Kit (ODK) software. Trained enumerators conducted household visits every three months from the time of enrollment until the child reached six or nine months of age. With the exception of pregnancy and birth outcome characteristics, data for this analysis came from the UBCS questionnaire administered at the enrollment visit, which occurred during the second or third trimester of pregnancy. Household food security status was assessed using the Household Food Insecurity Access Scale (HFIAS) [15], a validated tool for use in populations across different cultural contexts, including in rural East Africa [16]. The HFIAS covers a recall period of 30 days and consists of two types of questions: nine “occurrence” and nine “frequency-of-occurrence” questions. The respondent is first asked if a given condition was experienced (yes/no) and, if it was, then with what frequency (rarely, sometimes, or often). The resulting responses can be transformed into either a continuous or categorical indicator of food security. Categorically, households are characterized into four distinct categories: food secure, mildly food insecure, moderately food insecure, or severely food insecure. Dietary diversity during pregnancy was assessed from dietary recall data collected using the Food and Agriculture Organization (FAO) Minimum Dietary Diversity for Women (MDD-W) index [17]. Scores were computed as the sum of 10 food groups (grains, white roots and tubers, and plantains; legumes; nuts and seeds; dairy; meats, poultry and fish; eggs; vitamin A rich dark green vegetables; other vitamin A rich fruits and vegetables; other vegetables; and, other fruits) based on whether or not they were consumed in the previous 24-hours. At the enrollment visit, tests for maternal malaria infection and hemoglobin status in pregnancy were conducted by trained nursing staff at participants’ households. Malaria infection was diagnosed using a rapid diagnostic test (RDT, SD Bioline Malaria Ag P.f/Pan test, Standard Diagnostics, Inc., Republic of Korea), and hemoglobin levels were measured using a portable hemoglobinometer (HemoCue 301; HemoCue America, Brea, CA, USA). Depending on the results, appropriate counseling, treatment, and/or referral to local health facilities were provided in accordance with UBCS standard operating procedure (SOP). Gestational age was calculated from the first day of mothers’ last menstrual period (LMP). Maternal height was measured to the nearest 0.1 cm using a portable height board (ShorrBoard® infant/child/adult portable height-length measuring board; Weigh and Measure, LLC, Olney, MD). Infant birth weight was measured within 72 hours to the nearest and 0.1 kg using an electronic scale (Seca model 874, Seca Corporation, Hanover, MD). In all anthropometric measures, triplicate measurements were averaged to provide a single measurement. For the purpose of this analysis, infants born <2,500 grams were considered LBW, and infants born <37 weeks gestation were considered preterm. Adolescent pregnancy was defined as <20 years old. A binary indicator for adequate dietary diversity (i.e., ≥5 food groups in the previous 24 hours) was calculated per FAO’s recommendation [17]. Per recommendations from the World Health Organization (WHO), adequate birth spacing was defined as >24 months [18] and adequate ANC care was defined as ≥4 visits per the previous four-visit ANC (FANC) model [19]. Prior to regression analyses, categorical summary statistics for household (location, household head sex, household head marital status, household head education, household food security, water source, UCCP participation), maternal (age, height, education, dietary diversity, gravida, birth spacing, ANC visits, deworming medication, iron tablets, hemoglobin, malaria status, HIV status), and infant (sex, location of delivery) characteristics were cross tabulated among LBW and non-LBW infants and among preterm and non-preterm infants. Bivariate logistic regression analyses were conducted to identify the relationship between independent household, maternal, and infant characteristics and the birth outcomes of interest (i.e., LBW and preterm birth). All variables in the bivariate analysis were considered for multivariable logistic regression analysis. Backward stepwise logistic regression models, which produced adjusted odds ratios (aORs), with a 0.05 cut-off for inclusion, were used to test for the predictors of LBW and preterm birth. All analyses were conducted using STATA 15 software (Stata Corps, College Station, TX, USA). In all cases, p<0.05 was considered statistically significant.