Background: Maternal and child under-nutrition is the underlying cause of death for millions across the globe. Anemia during pregnancy is among the leading nutritional disorders with serious short and long term consequences to both the mother and fetus.

Objective: Examine the effect of dietary diversity on maternal anemia, nutritional status and key pregnancy outcomes of pregnancy.

Methods: A prospective cohort study design, involving a total of 432 eligible pregnant women, in their second antenatal care visit was conducted between August 2014 to March, 2015. The individual dietary diversity Score (IDDS) was used as the exposure variable to select, enroll and follow the mothers. Epi-data, SPSS and STATA software are used to enter and analyze the data. Chi-square test, independent 't'-test, and GLM are used to calculate risk, association and differences between key variables at P < 0.05

Introduction:

Nutrient intake, food choices, and dietary diversity are the key determinants of nutritional status and associated health outcomes in human beings. When people eat a variety of foods, they are more likely to meet their needs for a wide range of essential nutrients. Yet for many poor people across the globe, dietary diversity is very low, and the daily diet is dominated by one main staple food. Poorly diversified monotonous foods, originating from plant sources are typical characteristics of the diets most peoples in low income countries, particularly the in sub-Saharan Africa.

Dietary diversity is a measure of the number of individual foods or food groups consumed in a given time period. Particularly, the Individual Dietary Diversity Score (IDDS) is a useful proxy measure of the nutritional quality of an individual's diet and a reflection of nutrient adequacy. It can be triangulated with other food-related information to contribute towards providing a holistic picture of the food and nutrition status of an individual and can be used as a simple and quick indicator of the micronutrient adequacy of a diet.

Maternal and child under-nutrition is considered as the underlying cause of 3•5 million annual deaths, and responsible for the overall increase in disease burden of the low and middle-income countries. As child health and pregnancy outcomes are largely influenced by maternal dietary practices, in nutritionally deprived populations, maternal nutrition is identified as a strong predictor of poor pregnancy outcomes including low birth weight (LBW), still birth and pre-term deliveries. Furthermore, not only the nutritional status, but also anthropometric, biochemical, behavioral and other characteristics of the mother has also been strongly associated with pregnancy and pregnancy outcomes.

Anemia during pregnancy is among the leading maternal nutritional problem associated with a number of poor pregnancy outcomes. Globally, over 42% of pregnant women were identified as anemic and nearly 60% of this anemia is assumed to be due to iron deficiency in non-malarias areas, and 50% in malaria areas.(16). The prevalence, complications, severity and associated risks of anemia among pregnant and lactating mothers in the low and middle income countries are alarmingly high.

Ethiopia is a sub-Saharan African country with unacceptably high level of maternal and neonatal mortality rates. The prevalence of adverse pregnancy outcomes; low birth weight, preterm and still birth in the country are also very high. Contrarily, there is paucity of evidence regarding the problem and solutions to undertake. Only very few cross-sectional studies have investigated the level and determinants of maternal anemia and prenatal outcomes. Given the dynamic physiology of pregnancy, these cross-sectional studies are ill-equipped to address the problem occurring naturally. Hardly that the available studies also analyzed the association between of maternal dietary diversity practices and anemia with subsequent pregnancy outcomes. There is also inconsistency of the findings of these studies in the magnitude and predictors identified.

Therefore; this study examined the association between of maternal dietary diversity practices during pregnancy with anemia and subsequent adverse pregnancy outcomes: low birth weight, preterm and still birth.

Methods:

Study design and study area. A prospective cohort design was employed to select, enroll, categorize and follow pregnant mothers from second antenatal care visit to the end of delivery, in eight randomly selected health centers of four rural districts, in Ethiopia. The Zone is divided into four agro-climatic areas mainly due to variation in altitude. It is dominantly characterized by moderately cool (40 %) and cool (34 %) agro-climatic zones, while the remaining 1/4th accounts moderately warm (20%) and cold (6%) temperature zones. Arsi Zone is one of the surplus producing areas in Ethiopia, with very few food insecure households. Recent studies has showed that majority of the rural population has inadequate or moderate DD practices and only less than 10% of the same practiced high (3,27,28).

Sample size Sample size was calculated using a Open Epi Kelsey statistical software . The following assumptions were taken into consideration: a 95 % two-sided significance level, 80% power, a one to one ratio of exposed to unexposed and 23% anemia among exposed (mothers with inadequate dietary diversity) and a predicted 10% lower rate of anemia among unexposed mothers. This yielded a total of 196 per each arm and a total of 392 pregnant women. Adding to allow a 10% drop-out rate during the study, some 216 pregnant women per category or a sum of 432 women as the final sample for the study was calculated.

Sampling Procedure Pregnant mothers were recruited and enrolled to the study during the second trimester of pregnancy with gestational age ranging between 24 - 28 weeks. They were also followed from the time enrollment to final end of pregnancy or delivery for documenting remaining maternal and fetal outcomes. The timing for enrollment at second trimester of pregnancy was based on previous studies in Ethiopia showed the fact that rural pregnant mothers visit health facilities for antenatal care lately during the second or third trimester of pregnancy. Though formal enrollment to the study is made in the second visit, relevant obstetric, anthropometric and biochemical data from medical records for those mothers who initiated antenatal visits earlier during the first trimester of pregnancy were documented.

It was the maternal DD score used as a main criteria to recruit, enroll and follow these mothers by their respective study categories. Pregnant mothers were categorized either to the adequate dietary diversity ''adequate'' or inadequate dietary diversity ''inadequate'' groups in a one to one ratio based on the individual dietary diversity scores (IDDS). Mothers having IDDS of < 4 were categorized to the inadequate while those having 4 - 9 IDDS were assigned to the adequate group (Figure. 1). During analysis; however, the adequate group was re-categorized into moderate and high dietary diversity sub-groups as it covers a wide range of IDDS from four to nine. All classification and categorization was strictly based on the recommendations of recent FAO's guidelines.

Inclusion and Exclusion Criteria To select appropriate study subjects, criteria like selecting a pregnant women who are willing to stay in the study throughout the whole course of pregnancy and visit same health center for the antenatal and delivery care services were used. Additional criteria used are weather the pregnant mother has lived for at least six months in the study area or not, if she has previous known medical, surgical or obstetric problems were taken into account for inclusion or exclusion.

Food Consumption (Dietary intake) According to FAO's latest and revised guidelines techniques and methods for measuring household and individual dietary diversity, a qualitative recall of all foods consumed by the women during the previous 24-h period was performed during two distinct seasons (pre-harvest of the main harvest season, which occurs commonly between August to October and the peak harvest season, November - January). A total of four 24-hours visit were completed for each mother on monthly basis starting from enrollment to delivery.

Each woman involved in the study was asked to recall all the dishes, snacks, or other foods she had eaten during this period, regardless of whether the food was eaten inside or outside the compound. The woman was also asked to spontaneously describe her food consumption without any probes. It was after collecting all spontaneous responses that she is prompted to be sure that no meal or snacks had been forgotten. Next, a detailed list of all the ingredients of the dishes, snacks, or other foods mentioned, was collected from either the person in charge of their preparation or directly from the woman being interviewed.

Dietary diversity scores Dietary diversity score (DDS) was defined as the number of food groups consumed over a period of 24h. The diet was classified according to nine food groups as recommended by FAO, which included: (1) cereals, roots and tubers; (2)vitamin-A-rich fruits and vegetables; (3)other fruit; (4) other vegetables; (5) legumes and nuts; (6) meat, poultry and fish; (7) fats and oils; (8) dairy; and (9) eggs. Other remaining items such as tea, sugar and sweets were not used in DDS calculations.

Dietary diversity scores were calculated by a tally of food groups consumed by the responding pregnant mothers. After the respondent recalls all the foods and beverages consumed as described above, data collectors (midwifes at the health centers) underlined the corresponding foods in the list under the appropriate food group writing "1" in the column next to the food group if at least one food in this group has been consumed. If the food is not listed in any group, it was written in the margin and discussed later with the investigator or immediate supervisors assigned to oversee the data collection activity. Though happened rarely, mothers were also asked about the ingredients and food substances of which the diet consumed when a mixed diet consumption was found. This was settled based on agreed upon classification that was selected ahead of data collection, and sometimes consulting for investigators or supervisors.

Anthropometry The anthropometric measurements were performed using the standardized procedures recommended by WHO. Pregnant women were weighed to the nearest 100 g on electronic scales with a weighing capacity of 10 to 140 kg. Their height was measured to the nearest mm with portable devices equipped with height gauges (SECA 206 Body meter) locally calibrated and standardized. The mid-upper arm circumference (MUAC) of the left arm was measured to the nearest mm with a non-stretch measuring tape. Women with unreliable measurements due to a physical handicap were excluded from all analyses using anthropometric measures.

Hemoglobin : Hemoglobin level was taken twice, during the initial antenatal care visit by the midwives in the health center to document baseline data which is part of the standard practice. Hemoglobin levels were also collected at term (fourth antenatal care visit). It was measured using HemoCue (AB Leo Diagnostics, Sweden) at the laboratories of the health centers. Before potential use of the of hemoglobin levels obtained for analysis, we made adjustments based on WHO recommendations (34) for altitudes of each district.

Gestational age: Gestational age was estimated by midwives in the health center, from the last menstrual period (LMP) and fundal palpation during antenatal care which is also verified at delivery. Records on gestational age were taken five times; at baseline (ANC 1), enrolment (ANC 2), third and four ANC visits and at delivery.

Allocation to cases and controls: A mother is assigned to either the adequate or the inadequate group based on consistency of dietary diversity practices across the four visits considered. She will be to a group if remained to fall in that category in three of the four visits or showed at most one shift. Otherwise, considered incomplete and eliminated from analysis.

Potential sources and controlling for Bias : Anticipating the fact that the major source of bias could be due to key variables like major socio-demographic and economic factors and intake /compliance to iron folic acid supplements. To control this, random allocation and post enrollment analysis was made and to check differences across study these groups. Controlling of these variables was also made during data analysis.

Outcome Ascertainment: The outcomes measured include: maternal anemia level, low birth weight, pre-term birth and still birth. Maternal anemia during pregnancy was ascertained by the hemoglobin levels at baseline and term. Birth weight recorded by the midwives immediately after birth in the study health centers to the nearest 10g,. Still birth and pre-term birth were also ascertained by the same professionals at birth.

Other information, data on maternal characteristics such as age, education, reproductive history and morbidity were collected. Information on the household's composition, expenditures and other socioeconomic, environmental and health indicators were also documented

Data collection Study tools (questionnaires and others) were adopted from other studies and pre-tested on 5% of similar population but where the actual study was not conducted. Accordingly, relevant obstetric and nutritional data were collected using these tools by twenty four well trained midwives working permanently in the antenatal care service provision units of respective health centers. Over two third (seventeen) of these nurses were diploma holders and the remaining had first degree in health sciences. All of them had at least four years of work experience in the same unit.

A five days training on participant selection process, enrollment, follow-up and data collection tool was given to all of these midwives involved in the data collection and supervision process. In each of the health centers selected, one supervisor (usually head of the health center) was assigned to oversee the data collection. The investigator frequently traveled to the health centers from enrollment to final data collection, at least once every other week in each facility from the start of data collection to final outcome ascertainment and completion of data collection activities.

Prior to beginning data collection, the questionnaire was adapted to the local survey context by appropriate translation into local languages (Afan Oromo and Amharic) and adaptation of the food lists of the standard questionnaire to reflect locally available foods. It is also agreed on a common meaning and translation of terms used to describe key concepts. The most appropriate food group classification were based on foods which can be classified into more than one food groups. Mixed dishes were dis-aggregated in order to record all of the individual components in their respective food groups.

Eligible participating women were approached by the midwives verbal consent was obtained after proper explanation on study procedures for enrollment and follow - up. During subsequent visit, all relevant data were collected as the data collection guide prepared by the investigator to be used by the data collecting midwives at the health centers.

Ethical Considerations. The study was approved by Ethics Committees of both the Addis Ababa University, College of Natural Sciences as well as the Regional Health Bureau. A formal and official cooperative letter was written hierarchal from the region to the zonal health office and then to , district to health centers and finally to Kebeles (villages). Prior to undertaking interviews, written consent was obtained from all mothers.

Statistical Analysis. Data entry was performed with Epi-data statistical software. Data quality was maintained by quality checks during both data collection and entry (double entry) and further cleaning. All statistical analyses were carried out using SPSS Statistics (version 20.0). Exploratory data analysis techniques were used to uncover the distribution structure of the study variables and identify outliers or unusually entered values.

Distribution of continuous variables was tested for normality using Smirnov - Kolmogrov test. Variables not assuming normal distribution were transformed to log distribution before actual analysis is made. Independent 't'-test was used to test for mean differences between the different dietary diversity groups. General linear model and Chi-square test or Fisher exact test were used to test for independence in distribution of categorical variables (demo- graphic characteristics, categorized nutritional variables, and dietary intake) between study groups.

Inclusion Criteria:

Pregnant women who are :

• Willing to stay in the study throughout the study
• Visit same health center for the antenatal and delivery care services
• pregnant mother has lived for at least six months in the study area

Exclusion Criteria:

• Pregnant women who have known medical, surgical or obstetric problems

Pregnant women in rural Ethiopia