Study area, period and design
The study was conducted in Tigray regional state, Ethiopia. It is located at 780 km from the capital of Ethiopia, Addis Ababa. According to the 2018 Tigray regional health bureau report, there are 7 zones, 52 woredas/districts, 780 health posts, 227 health centers, and 18 hospitals. The major agricultural products found in Tigray regional state are cereals (Taff, barley, maize, and wheat), grains (bean, soya bean, and pea), vegetables, fruits, honey, and roots, animal products like meat, poultry, and milk, and milk products. We applied a prospective cohort study design from October 2019 to June 2020.
Population, recruitment and follow up
The source population was all pregnant women attending antenatal care (ANC) in hospitals of Tigray regional state whereas the sample population was all pregnant women attending ANC in the randomly selected hospitals of Tigray regional state. The inclusion criteria were first ANC visit with gestational age not more than 16 weeks, willingness to attend routine ANC visits and permanent residence, and more than six months residency in Tigray regional state. Pregnant women with severe illness, overweight or obesity (BMI ≥ 25 kg/m2), and multiple pregnancies were excluded from the study. Pregnant women were recruited at first ANC contact (≤16 gestational weeks). Then, they were followed at ANC2 (20–26 gestational weeks), and ANC3 (30–40 gestational weeks). Finally, data on birth outcomes were taken during the institutional delivery. We had planned to exclude still births but such cases were not found during our data collection period.
Sample size determination
In this prospective cohort study, the ratio of exposed to unexposed pregnant women group was 1:1. We calculated the sample size using Epi-Info version 7.2.4 with the assumptions of 95% significance level (2-sided), 80% power. Besides, we took the incidence of low birth weight for exposed (11.8%) and unexposed (10.4%) pregnant women, and a relative risk of 1.9 from a study conducted in Tigray regional state in 2014 . Then, we considered a 10% loss to follow up and our total sample size for this study was 540 (270 for exposed and 270 for the unexposed pregnant women).
Firstly, we used a simple random sampling technique to select a total of six hospitals (Mekelle hospital Wukro hospital, Adigrat hospital, Adwa hospital, Aksum hospital, Suhul hospital) from a total of 18 hospitals in Tigray regional state. Secondly, we took a total number of study participants from each hospital based on their proportion to population size (PPS) i.e. proportional to their average client size attended per month by referring to the registration books of each antenatal care unit. Lastly, we recruited participants at the antenatal care unit using a consecutive sampling technique until the required sample size was attained. The recruited pregnant women were followed until they gave births.
The dependent variable was the incidence of low birth weight and the independent variables include; socio-demographic and economic factors (age, educational status, marital status, parity, monthly income, residence, and religion), diseases (anemia, syphilis, hepatitis B virus, and HIV), dietary-related factors (diet diversity scores, food insecurity, drinking of tea/coffee), hygiene and sanitation-related factors (latrine, hand washing, and access to safe water), and anthropometry measurements (maternal height and weight, mid-upper-arm circumference (MUAC), gestational weight gain, and newborn weight). Gestational age at first contact was determined by the last normal menstruation period (LNMP), fundal height measurement, and or ultrasound. Low birth weight is weight < 2.5 kg at birth . MUAC was used to assess the nutritional status of pregnant women. Those with a MUAC value less than 23 cm were considered as undernourished (exposed) and those with MUAC≥23 cm were normal nourished (unexposed) [14, 15]. Diet diversity scores (DDS) of the pregnant women were calculated using the 24-h dietary recall method. DDS is a proxy indicator for the quality of consumed diet, which in turn reflects the consumption of micronutrients. DDS out of 10 points was computed by combining the values of all the groups. The DDS was categorized as low (< 5) and recommended (≥5) . Household food insecurity scores were classified into one of the four categories: food secured, mildly food insecure, moderate food insecure, and severely food insecure .
Data collection tools, procedures and quality control
Data were collected by face-to-face interview using a pre-tested and structured questionnaire developed from EDHS and other literature [5, 12, 13, 16, 18]. Initially, the questionnaire was prepared in English and contextualized/adapted in a culturally relevant and comprehensive form. Then, it was translated into Tigrigna (local language) and translated back to English by language experts to check consistency. The questions were simple, clear and unambiguous. Nutrition and health experts have participated in developing and commenting the questionnaire. Some wording and sequences of the questions were modified after pre-test the questionnaire. The questionnaire had baseline questions concerning socio-demographic and economic, diseases, diet, hygiene, and sanitation-related factors and anthropometry measurements. The questionnaire had follow-up questions on MUAC, gestational weight gain, diet diversity score (DDS), and iron folate supplementation. Finally, the questionnaire had questions about birth outcomes like the sex of the newborn, live/dead, and birth weight.
Anemia was identified using the WHO hemoglobin concentration cutoff point, less than 11 g/dl . There were counseling and testing services to diagnose HIV, syphilis and hepatitis B virus in all hospitals. MUAC was measured by non-stretchable measuring tape. A tape was fixed at the mid-point between the elbow and the shoulder (acromion and olecranon). The tape measure was placed around the non-dominant arm; usually the left arm. The weight of the mother was measured in kilograms with a weighing scale (Seca designed by Germany) and rounded off to the nearest 0.1 kg. The height of the mother was measured with a stadiometer (Seca-2000, mechanical height meter), without shoes, and rounded off to the nearest 0.1 cm. The weight of the baby at birth was measured in kilograms on digital baby scales (Seca 354 Hamburg, Germany). The infants were weighed wearing no clothing. To estimate DDS, the pregnant women were asked to recall all the food items consumed in the previous 24 h preceding the data collection date. Then, the reported food items were classified based on the ten food groups (Cereals, Pulses, Nuts and Seeds, Dairy, Meat, Eggs, Dark green leafy vegetables, other vitamin A-rich fruits and vegetables, other vegetables, and other fruits). Consuming a food item with a minimum of 15 g (one teaspoon) from any of the groups was assigned a score of “1” and a score of “0” was assigned if no or less than 15 g food was consumed. Household food insecurity was assessed by a tool adopted from food and nutrition technical assistance (FANTA). The tool contains 18 questions; the first 9 questions were answered by yes or no based on the occurrence of the condition in the past four weeks. If the answer was yes for the occurrence question, a question for frequency of occurrence was asked to determine whether the condition occurred rarely (once/twice), sometimes (three to ten times), or often (more than ten times) . Standard procedures were used in anthropometry measurements. All anthropometric measurements were taken three times and the average was calculated to ensure reliability. Data collectors and supervisors were BSc midwives. Two days’ training was given for data collectors and supervisors. All instruments were calibrated regularly using standard measurements. Data collectors were under close supervision and the collected data were reviewed and checked for completeness, clarity and accuracy on a daily basis prior to data entry.
We used a statistical package for social sciences (SPSS) version 25 to analyze the collected data. We cleaned the data by sorting and tabulating simple frequency tables. Low birth weight was dichotomized into 1 = Yes and 0 = No. Then, we computed descriptive statistics for the study variables. Categorical variables were reported using frequencies and percentages. We checked the normality for the distribution of continuous variables using the Shapiro-Wilk test. We applied crosstabs to estimate the cumulative incidence of low birth weight. Chi-square test was used to assess the significant differences in the cumulative incidence of low birth weight. The difference was considered statistically significant at P-value < 0.05. We used a log-binomial model to estimate the adjusted risk ratio and its 95% confidence interval (CI) of the risk factors for low birth weight.