Association between Wood and Other Biomass Fuels and Risk of Low Birthweight in Uganda: A Cross-Sectional Analysis of 2016 Uganda Demographic and Health Survey Data

In utero exposure to household air pollution (HAP) from polluting cooking fuels has been linked to adverse pregnancy outcomes including low birthweight (LBW). No previous study in Uganda has attempted to investigate the association between the different types of biomass cooking fuels and LBW. This study was conducted to investigate the association between wood and other biomass cooking fuel use with increased risk of LBW, using the 2016 Uganda Demographic and Health Survey for 15,270 live births within five years prior to interview. LBW, defined as birthweight of <2500 g, was estimated from maternal recall and health cards. Association between household exposure to the different solid biomass cooking fuels and LBW was determined using multivariable logistic regression. Biomass cooking fuels were used in 99.6% of the households, with few (0.3%) using cleaner fuels and 0.1% with no cooking, while the prevalence of LBW was 9.6% of all live-births. Although the crude analysis suggested an association between wood fuel use and LBW compared to other biomass and kerosene fuel use (AOR: 0.82; 95% CI: 0.67–1.00), after adjusting for sociodemographic and obstetric factors, no association was observed (AOR: 0.94; 95% CI: 0.72–1.22). LBW was significantly more likely among female neonates (AOR: 1.32 (95% CI: 1.13–1.55) and neonates born to mothers living in larger households (AOR: 1.03; 95% CI: 1.00–1.07). LBW was significantly less likely among neonates delivered at term (AOR: 0.39; 95% CI: 0.31–0.49), born to women with secondary or tertiary level of education (AOR: 0.80; 95% CI: 0.64–1.00), living in households with a higher wealth index (AOR: 0.69; 95% CI: 0.50–0.96), Eastern (AOR: 0.76; 95% CI:0.59–0.98) and Northern (AOR: 0.75; 95% CI: 0.57–0.99) regions. The study findings suggest inconclusive evidence regarding the association between the use of wood compared to other biomass and kerosene cooking fuels and risk of LBW. Given the close observed association between socioeconomic status and LBW, the Ugandan government should prioritize public health actions which support female education and broader sustainable development to improve household living standards in this setting. The UDHS (2016) is a cross-sectional population-based national dataset funded by the U.S. Agency for International Development, with the birth recode (a file produced by DHS where each observation is an individual birth within the last five years) and relevant variables from the individual recode being extracted for this study [28]. The birth recodes contained birth history data, while individual recode (each observation is every woman within the survey) provided information on socio-demographic and household characteristics [28]. A two-stage stratified sampling methodology was employed to randomly select a representative sample [28]. Any woman residing in the selected household of reproductive age (15–49 years) was interviewed and asked to report their birth history (including live and still births) for the five years preceding the survey. A total of 19,588 households and 18,506 women were surveyed [28], with response rates of 97% (18,506/19,088) for the individual (women’s) dataset and 67% (10,429/15,522) of the birth records respectively. Additional information regarding the UDHS has been described elsewhere [28]. Singleton live births, occurring at term (≥37 weeks gestation) and/or pre-term (<37 weeks gestation) which occurred in the last five years (2012–2016) from the time of interview were included in the study. Multiple births were excluded from the analysis because of the high risk of LBW among multiple pregnancies [31]. The wealth index provided by DHS is calculated through principal component analysis (PCA), including assets, toilet facility, drinking water sources, cooking fuel, and house construction as predictor variables [32], with the final variable containing wealth quintiles (lowest, low, middle, high, and highest). As cooking fuel was the exposure of interest within this study, the wealth index (categorized as low, second, middle, fourth, and higher) was recalculated using the methods provided by the DHS [33] in SPSS [34] to remove cooking fuel to prevent circularity [35]. Self-reported main household cooking fuel was categorized into cleaner fuels (LPG), electricity, biogas, and no cooking), and biomass and kerosene fuels (kerosene, coal/lignite, charcoal, wood, straw/shrubs/grass, agricultural crop, animal dung). Households, where cooking was not done, were classified as using cleaner fuel because it was assumed that HAP levels in these households would be comparable to households that used cleaner fuels. The exposure variable was the biomass or kerosene cooking fuels, categorized into wood and other biomass and kerosene fuels (kerosene, charcoal, straw/shrubs/grass, agricultural crop, animal dung). The outcome variable was LBW defined as a birthweight of less than 2500 g, obtained from either the health card (34%) or the maternal recall of child’s weight at birth (66%). Covariates were identified from the literature [21,36,37,38] as those potentially associated with HAP or LBW. The covariates from household and contextual characteristics included age of the household head, access to electricity, place of residence (urban, rural), geographical region (central, east, north, west), household smoking status (yes, no), place of cooking (in the main house, separate house, outdoors) and wealth index (low, second, middle, fourth, or highest). The 15 sub-regions in Uganda were categorized into four regions which included central, east, west, and north, which are defined based on ethnicity, poverty index, and geographical location [39]. Information from the respondents included age (15–19, 20–34, 35–49 years), level of education (no or primary education, secondary or tertiary). Pregnancy-related maternal covariates considered included parity (primigravida or multigravida), birth order (continuous variable), sex of the baby (male or female), and body mass index (BMI) (<18.5 or ≥18.5 kg/m2). Circumstantial pregnancy-related covariates included mode of delivery (spontaneous vaginal delivery or caesarian section), the timing of first antenatal care visit (ANC) (≤5 or >5 months), number of ANC visits (≥4 or <4), sulphadoxine-pyrimethamine (SP) (yes or no), birth interval (<24 months or ≥24 months), iron-folate supplementation (yes or no), deworming during pregnancy (yes or no), birth interval, based on WHO categorization, of less than 24 months or birth intervals of ≥24 months was used [40]. BMI, measured in kg/m2, was categorized as low when BMI was <18.5 or normal when BMI was ≥18.5 [41], as there is a higher risk of LBW for BMI of less than 18.5 [42,43]. Categorical variables were summarized using frequencies and proportions. Skewed continuous variables were summarized using the median and inter-quartile range (IQR), while normally distributed continuous variables were summarized using means and standard deviations. Bivariate and multivariate logistic regression, using survey commands to adjust for the complex sampling structure, was deployed to determine the association between exposure to wood and other forms of polluting fuels and LBW. The odds ratios (OR), 95% confidence intervals (95% CI), and p-values were reported. Clinically relevant variables, those with a p-value was less than 0.2, and variables without high levels of missing, in the bivariate analyses were included in the multivariate logistic regression model. Missing values were handled by case-wise deletion. Sensitivity analyses were undertaken to ensure robustness of study findings and to further investigate confounding factors (e.g., BMI) that could not be accounted for in the main analysis due to a large proportion of missing data (Figure 1). Further stratified analyses were undertaken according to residence (rural or urban), cooking location (indoor, outdoor), and maternal BMI (≤18.5). Multivariable linear regression model was performed using birthweight as a continuous variable. Stata software (version 16.1) [44] was used to analyze the data. Description of sensitivity analysis. USAID obtained ethical approvals from the relevant authorities in Uganda to collect the data. Permission was obtained from the USAID to gain access to the anonymized and aggregated freely available dataset from the DHS online data archive [45].