Background: Maternal undernutrition is a pervasive health problem among Ethiopian mothers. This study aims at identifying the level of maternal undernutrition and its associated factors in Kilte Awaleo-Health and Demographic Surveillance Site (KA-HDSS), Tigray region, Ethiopia. Methods: Nutritional status of 2260 lactating mothers was evaluated using the mid-upper-arm circumference (MUAC). Data from the vital events and verbal autopsy databases were linked to the survey and baseline recensus data to investigate the association of adult mortality from chronic causes of death (CoD) on maternal undernutrition. We employed a generalized log-binomial model to estimate the independent effects of the fitted covariates. Results: The overall prevalence of maternal undernutrition based on MUAC < 23 cm was 38% (95% CI: 36.1, 40.1%). Recent occurrence of household morbidity (adjusted prevalence ratio (adjPR) = 1.49; 95%CI: 1.22, 1.81) was associated with increased risk of maternal undernutrition. In addition, there was a 28% higher risk (adjPR = 1.28; 95%CI: 0.98, 1.67) of maternal undernutrition for those mothers who lived in households with history of adult mortality from chronic diseases. Especially, its association with severe maternal undernutrition was strong (adjusted OR = 3.27; 95%CI: 1.48, 7.22). In contrast, good maternal health-seeking practice (adjPR = 0.86; 95%CI: 0.77, 0.96) and production of diverse food crops (adjPR = 0.72; 95%CI: 0.64, 0.81) were associated with a lower risk of maternal undernutrition. Relative to mothers with low scores of housing and environmental factors index (HAEFI), those with medium and higher scores of HAEFI had 0.81 (adjPR = 0.81; 95%CI: 0.69, 0.95) and 0.82 (adjPR = 0.82; 95%CI: 0.72, 0.95) times lower risk of maternal undernutrition, respectively. Conclusions: Efforts to ameliorate maternal undernutrition need to consider the influence of the rising epidemiology of adult mortality from chronic diseases. Our data clearly indicate the need for channeling the integrated intervention power of nutrition-sensitive development programs with that of nutrition-specific sectoral services.
Due to poor quality or absence of complete civil registration systems (CVRS), evidence-based health policy formulation and decision-making is challenging in most sub-Saharan Africa [30]. To date, complete and quality CVRS system has not been available in Ethiopia. However, there are other health data sources despite their limitations. In addition to the periodic and routine data sources (censuses, demographic and health survey, routine health information reports), selecting geographically defined populations, and generating longitudinal vital events (birth, migration, and death), causes of death (CoD), and marital status information could also be an alternative systematic approach. Such population-based and ongoing longitudinal health and demographic projects are known as health and demographic surveillance sites (HDSS), and are linked through The International Network for the Demographic Evaluation of Populations and their Health (INDEPTH) network [30, 31]. Pooling data from multiple HDSS, in a given country, may help to narrow the chronic data gaps in low and middle income countries [31], although this approach cannot be a replacement for a fully functional and quality CVRS. In Ethiopia, there are six HDSS projects home-based at their respective local university and networked by the support of Ethiopian Public Health Association (EPHA) and Center for Diseases Control (CDC-ETHIOPIA). Health and demographic data are pooled from these six sites and disseminated for public use on annual basis. KA-HDSS, which is located 835 km north of Addis Ababa in the Eastern Zone of Tigray, was established in 2009 with a total baseline population of 66,438 residents. The current study is based on data extracted from the KA-HDSS longitudinal data, and the nutrition and baseline recensus survey data, which was collected from July to December in 2015. The nutrition survey was conducted as a baseline for the installment of longitudinal nutrition project to the already existing main surveillance project. Our study population included all the 2260 breast-feeding mothers of KA-HDSS, who were at least 18 years old. The data collectors were high school completed permanent employees of the surveillance project (KA-HDSS). They were chosen to ensure the overall sustainability of the project and its collateral contribution to the community. They were rigorously trained, and experienced on the standard data collection tools and procedures, including on how to consent the study participants properly. Before the start of the data collection, the data collectors introduced themselves to the study participants, and gave information on the purpose of the study. The data collection started after consent from the participants was given. Furthermore, the interviewers were residents of the surveillance population, with better knowledge about the culture of their community, thus facilitating the proper implementation of the consent procedures. Face-to-face interviews were conducted together with direct observations and measurements. KA-HDSS data collection supervisors led by the field manager (from the field KA-HDSS office) and academic research members (from the KA-HDSS central home office at Mekelle University, College of Health Sciences) strictly monitored the data collection process. Maternal nutritional status was assessed using the mid-upper-arm circumference (MUAC) measurement. The midpoint between the tip of the shoulder and the tip of the elbow of the left arm was measured using a flexible, inelastic MUAC tape. Circumferential measurements were taken after making sure that the tape had a proper tension, not too loose or too tight, around the midpoint and all values were recorded to the nearest 0.1 cm. Lactating mothers with a MUAC below 23 cm were classified as undernourished, and those with less than 21 cm as severely undernourished [32, 33]. Studies have reported that MUAC is a reliably efficient alternative indicator for body mass index for evaluating adult nutritional status, particularly in developing countries where large logistical mobilization is needed to accurately measure the height and weight in population-based surveys [34, 35]. Its positive association with infant breast milk intake, breast milk volume and quality, its measurement simplicity and ability to predict mortality, particularly among the older adults, are additional advantages [35–38]. Detailed lists of agricultural asset ownership, livestock and food crops produced, were counted and converted to monetary terms and quintile classified. The lowest two quintiles were treated as “poor” and the remaining three upper quintiles as “not poor”. Maternal health practice index was constructed from two maternal health service utilization indicators: ever use of modern family planning methods and delivering at health facilities. If the score was 2 (mother used both health services), then it was operationalized as “good”. If the score was 10%), which is the case in this study [47–50]. In addition, its easiness of interpretation makes prevalence ratio preferable to the odds ratio [50]. Univariable log-binomial analyses were performed to observe the association of each independent variable with maternal undernutrition. In the next step, all the variables with a P-value of less than 0.25 in the univariable analysis were taken to the multivariable log-binomial model. Adjusted prevalence ratios and their 95% confidence intervals (95% CI) were reported at P-value of less than 0.05. In addition, we also assessed the factors associated with severe maternal undernutrition. In this case, since the prevalence of severe maternal undernutrition was less than 10%, we used multivariable binary logistic regression model to identify the potential associated factors. Similar to the above approach, variable selection was made at P-value of less than 0.25 in the univariable binary logistic regression and adjusted odds ratio (AOR) with 95% CI were produced from the multivariable model at P-value of less than 0.05. However, to encourage the interpretation of the associations from the perspective of practical (public health) significance, rather than focusing only on statistical significance, we decided to skip the P-values and instead report only the measures of association and their corresponding 95% CI. Categorical characteristics of the study participants are presented using frequencies and proportions, while the continuous characteristics are summarized using means with standard deviations (SD) and median with interquartile ranges (IQR), depending on their distribution. All analyses were performed in Stata 15.1 (StataCorp, LLC, College Station, TX).