Household factors and gestational age predict diet quality of pregnant women

Abstract Adequate diet during pregnancy has positive effects on the mother and pregnancy outcome. Assessment of diet quality during pregnancy is particularly important in areas where household food security is suboptimal, to enable appropriate targeting and intervention. This study assessed diet quality and identified predicting factors among pregnant women in northern Ghana. A cross‐sectional study involving 403 pregnant women was conducted in May 2018. Pregnant women attending antenatal care clinics (ANC) were selected using simple random sampling technique. We assessed socio‐demographic characteristics, 24‐h recall and household food security. The minimum dietary diversity for women (MDD‐W) was used as a proxy measure for diet quality based on Food and Agricultural Organization (FAO) guidelines. Logistic regression models were fitted to determine the predictors of diet quality. The mean dietary diversity score (DDS) of 10 food groups was 4.4 ± 1.1 (95% CI: 4.3–4.5). Logistic regression showed that women of high educational level (adjusted odds ratio [AOR] = 2.42; 95% confidence interval [CI] [1.21–4.84]; P = 0.01), women of high household wealth index (AOR = 1.78; 95% CI [1.14–2.77]; P = 0.01], none/mild household hunger (AOR = 2.71; 95% CI [1.26–5.82]; P = 0.01), medium household size (6–15 members) (AOR = 1.66; 95% CI [1.04–2.66]; P = 0.03) and women of gestational age 20–35 weeks (AOR = 1.89; 95% CI [1.05–3.40]; P = 0.03) were more likely to have quality diets after adjusting for potential confounding variables. Diet quality among pregnant women was low and was predicted by educational level, household wealth, gestational age and food security. Women education and improvements in household food security could impact diets of pregnant women in northern Ghana.


| INTRODUCTION
According to World Health Organization (WHO) (2012), pregnant women's diet should be balanced. Diet of pregnant women should be nutrient dense, as adequate nutrition during foetal development is important for long-term physical health of the child. It is at this stage that adequate nutrient supply is critical for both the mother and her unborn child. Pregnancy is specifically a critical period for the programming of future populations (Danielewicz et al., 2017).
Maternal diet quality as measured by dietary diversity is crucial for both the mother and child. In particular, diet in the first trimester is important for the development and differentiation of various organs, whereas diet later in pregnancy may be important for overall foetal growth as well as brain development (Rifas-Shiman et al., 2009). Inadequate nutrition, especially during the first trimester of pregnancy, may restrict foetal growth and has long-term consequences on both the mother and child.
Maternal dietary diversity in Northern region of Ghana has been linked to low birth weight among children (Saaka, 2013). It has also been linked to nutrition sufficiency, micronutrient adequacy (Ali et al., 2014) and pregnancy outcome (Cheng et al., 2009;Gao et al., 2013;Shamim et al., 2016;Willy et al., 2016). As food insecurity is high among households in Northern region and has adverse implications on nutritional status (Saaka, 2016), diet quality is likely to be poor among them. Pregnant women may be worst affected as their nutrient needs are higher.
Iron deficiency anaemia is an important public health problem among pregnant women. During pregnancy, anaemia is harmful to both the woman and child, and it is associated with a higher risk of maternal and foetal morbidity and mortality (Kalaivani, 2009). Women need to have adequate storage of nutrients to meet the nutritional demands of pregnancy (Marangoni et al., 2016;World Health Organization & UNICEF, 2001). Assessments of diet quality among pregnant women in Ghana in most studies have focused on the effect of specific nutrients rather than overall diet quality. A single nutrient approach is a major limitation because multiple nutrient deficiencies are more likely to occur than single nutrient deficiencies in a lowincome setting like Ghana (Darnton-Hill & Mkparu, 2015). Therefore, it is essential to shift from examining the effect of single nutrients to assessing the overall diet quality.
Assessing the determinants of diet quality during pregnancy is vital in ensuring optimal nutrition for both mother and the child (Madden, 2015). Thus, it is important to understand the factors associated with the intake of different food groups, particularly, intake of micronutrient-rich foods by pregnant women who are most at risk of having diets with poor diversity. Such information may help policymakers and programme managers in designing interventions to address the problem of poor diet quality. Data on factors associated with inadequate dietary diversity among pregnant women in the Savelugu Nantong district in the northern region of Ghana are scarce. Therefore, this study assessed diet quality and its association with maternal factors including household wealth, food security, nutrition knowledge, education level, household size and gestational age in the district.

| Study area
The study was conducted in the Savelugu-Nantong district in the northern region of Ghana; it has a single farming season, a potential for livestock and crop production (maize, millet, groundnut, cowpea, soybeans, yam, guinea corn and cassava) with a total population of 139,283 (Gss & Macro, 2009 were selected using simple random sampling technique from a list of all health facilities in the district using Excel-generated random numbers. We used the probability proportional to size (PPS) methodology to select 403 study participants from the four health facilities. Lists of pregnant women attending antenatal care clinics (ANC) in the four facilities were compiled to form the sampling frame.
Using the formula for estimating sample size of single proportions (Cochran, 2007) with a 95% confidence interval (CI), an assumed

Key messages
• Previous studies have focused on the effect of specific nutrients without taking account overall diet quality; this study assessed overall diet quality and its predictors among pregnant women.
• We found that diet quality among pregnant women was low and was predicted by gestational age, household food security, educational level and household wealth status.
• In this low-income setting in West Africa, our findings could have important policy implications for improved female education, food security and antenatal care for improved diets during pregnancy and improve birth outcome.
prevalence of 50% of pregnant women meeting minimum dietary diversity, a margin of error of 0.05 and a 5% contingency, the sample size was 403.

| Measurement of study variables
The dependent variable was diet quality measured using a 24-h dietary diversity score (DDS). The independent variables included the socio-demographic characteristics of pregnant women such as age, highest education completed, occupation, religion, marital status, household size (grouped into 1-5, 6-15 and above 15 household members who live together and eat from a common cooking pot), household wealth and gestational age (weeks of pregnancy recorded from pregnant women's antenatal cards), household food security and nutrition knowledge. A brief description of the key study variables is as follows.

| Determination of household wealth status
Household wealth index was used as a proxy indicator for socioeconomic status (SES) of households. This was based on an earlier concept by Garenne and Hohmann-Garenne (2003) and Saaka et al. (2017), using absolute measurement of household items such as housing quality, water availability and type of toilet facilities, and ownership of durable goods and livestock. Households were given scores depending on their possession of household assets. Household wealth index was the summation of individual scores of each household items; the scores ranged from 0 to 18. Households with wealth scores less than or equal to 13 were considered as having low wealth index score and households with wealth scores greater than 14 as high wealth index. A dichotomous variable had favourable model fit characteristics compared with the usual quantile system, which would have increased the number of parameters estimated for household wealth.

| Measurement of diet quality
Diet quality was measured using DDS because it reflects intake of nutrient adequacy (Mirmiran et al., 2006). The Food and Agricultural Organization defines dietary diversity as the number of food groups an individual eats over the past 24 h (Kennedy et al., 2011). Dietary diversity of the pregnant women was assessed using the Food and Agricultural Organization (FAO) women dietary diversity questionnaire to determine women dietary diversity score (WDDS) (FAO, 2016). Emphasis was on dietary diversity; hence, food frequency or amount of food taken was not considered. The food groups that were included in the questionnaire were as follows: (1) grains, white roots and tubers and plantain (maize, rice, yam and wheat or foods made from these, e.g., bread, TZ, porridge and banku); (2) Vitamin A-rich dark green leafy vegetables (Amaranth, okra leaves, baobab leaves, ayoyo, alefu and bra); (3) other Vitamin-A rich vegetables (tomatoes, carrots, orange flesh potatoes, ripe mangoes, dawadawa pulp and watermelon); (4) other vegetables (garden eggs, okra fruits, garlic, green and red pepper and onion); (5) other fruits (orange, lemon, guava, ebony fruits, blackberry and banana); (6) flesh meat (beef, lamb, chicken and any organ meat, e.g., liver and offals); (7) eggs (chicken and guinea fowl); (8) nuts and seeds (groundnuts, cashew nuts, bungu [sesame] and melon seeds); (9) pulses (beans, peas and lentils); (10) milk and milk products (milk, cheese and yogurt). Respondents were given a score of one for consuming food from a specific food group and 0 for otherwise, this was used to calculate the total women dietary diversity score (WDDS) by summing scores from all foods eaten. The maximum women dietary diversity score (WDDS) was 10 for consuming food from all food groups and 0 for not consuming food from any food group at all. Respondents who had dietary diversity score of 5 and above were considered to have met the women dietary diversity (MDD-W), because dietary diversity was used as a proxy for measuring diet quality, respondents who met their diet quality were considered to have quality diets.

| Measurement of maternal nutrition knowledge
The knowledge questions were adapted from the guidelines for assessing nutrition-related knowledge, attitudes and practices (KAP) manual published by FAO (Macías & Glasauer, 2014). The questionnaires were field tested and validated in several countries. The focus of our assessment was on basic maternal nutrition that is usually taught in nutrition education sessions to pregnant women. The content included food groups and their functions, balanced diet, nutrition during pregnancy and lactation, causes and prevention of anaemia, iron absorption enhancers and inhibitors and foods sources of key nutrients (iron and Vitamin A).
Maternal nutrition knowledge was measured on a continuous scoring scale. The assessment was based on nine questions on knowledge of nutrient rich foods required during pregnancy and on anaemia.
A score of 1 was assigned when a correct answer was given otherwise 0 to give a maximum score of 9. Overall nutritional knowledge was rated by calculating the total of all the valid responses pregnant women made. The overall composite nutrition knowledge index ranged from a minimum of 2 to a maximum of 9. The median score was 6.0. The scores were divided into two categories: pregnant women whose overall nutrition knowledge score was less than the median score were considered to have low knowledge and those with scores of 6 and above as high knowledge.

| Assessment of anthropometry and nutritional status
Measurement of mid-upper arm circumference (MUAC) was used to assess nutritional status, as this measurement is independent of pregnancy and therefore it is an effective indicator for a woman's nutritional status throughout the reproductive life. MUAC was measured following WHO standard procedure (WHO, 1995). MUAC measurement was used to classify pregnant women as normal or malnourished; pregnant women with MUAC less 23 cm were considered malnourished, and those with MUAC greater than or equal to 23 cm were considered to be normal (Tang et al., 2016).

| Measurement of household food security
Household food security was measured using a modified FANTA household food insecurity and access scale (HFIAS) (Coates et al., 2007). The HFIAS has previously been modified to simply reflect household hunger, the household hunger scale (HHS), using three questions and three sets of responses (Deitchler et al., 2011;Saaka et al., 2017;Silventoinen, 2003). The three responses (never, rarely or sometimes and often) were scored and used to classify food insecurity/hunger. Never was assigned a score of 0; rarely or sometimes had a score of 1; and often was assigned a score of 2. Therefore, 0 was the minimum score, and 6 was the maximum score. Scores of 0-1 indicate little to no household hunger, 2-3 indicate moderate household hunger, and 4-6 indicate severe household hunger (Deitchler et al., 2011). In regression analysis, we combined none and mild household hunger and those with medium or and severe as the second to reduce the possibility of data sparsity arising from empty cells in the fully adjusted model.

| Quality assurance
Questionnaire was translated to Dagbanli and pretested on pregnant women attending antenatal care at health facilities. The questionnaires were administered by four trained undergraduate nutrition officers who attended a 3-day training on administering the questionnaire and anthropometric assessment (MUAC). Data were collected at designated places at various health facilities to ensure privacy and were then checked on site for completeness on daily basis; incomplete questionnaires were identified, and respondents were followed up to complete missing data.

| Data analysis
Data analysis was performed using SPSS version 21 (IBM Inc.). Means and standard deviations were used for continuous data. Frequencies and percentages were used for categorical data. Bivariate analyses were performed using chi-square/Fisher exact test as preliminary analysis to identify predictors of women's dietary diversity. Variables showing evidence of association with the outcome variable from the preliminary analysis were included for multivariate logistic regression analysis. We used the variance inflation factor (VIF) to assess collinearity among variables in the regression model, variables did not show evidence of significant collinearity (VIF < 5) (Ringle et al., 2015). Independent factors such as gestational age, educational level, household wealth index and household size were accounted for in the multivariate logistic regression modelling.

| Ethical statement
Ethical clearance for the study was obtained from the School of Allied Health Sciences, University for Development Studies, Tamale, Ghana.
Informed consent was obtained from each participant before they enrolled on the study.

| Socio-demographic characteristics of study participants
The study involved 403 pregnant women with more than half (58.3%) of them between 21 and 30 years. The predominant religion in the setting was Islam (97.8%). Six out of every 10 respondents did not have formal education (61.5%), 96.8% were married and were mostly from the Dagomba ethnic group. Trading was the main occupation of the respondents (28.0%); however, 32.8% of were housewives. More than half (61.3%) of the pregnant women had high wealth index, and most (89.1%) of them did not experience food insecurity (Table 1).

| Dietary diversity and frequency of consumption from specific food groups among pregnant women
The mean dietary diversity score (DDS) for 10 food groups at 95% CI

| Nutritional knowledge of pregnant women
The proportion of pregnant women with high nutritional knowledge was 56.8%. Four out of every five pregnant women knew that breakfast was the most important meal and could mention at least one cause of anaemia. The proportion of pregnant women who knew that frequent suckling produces more breast milk was low (11.4%). Less than a half of the participants knew that coffee or tea could decrease iron (48%), although all of them knew that Vitamin C-rich foods enhanced iron absorption. Most (82.4%) of pregnant women could mention at least three food sources of micronutrient in diets, and 72.0% of them knew the importance of feeding their children with fruits and vegetables. (Table 3).

| Factors associated with MDD-W
Logistic regression analysis showed that maternal educational level, household wealth index, household food security, household size (6-15 members) and gestational age were factors associated with having five or more food groups in the day (Table 4)

| DISCUSSION
In the present study, we sought to assess diet quality of pregnant women and characterize the most important predictors of their diet. We found that high education level, high household wealth, food security, household size and women of young gestational (20-35 weeks) age were more likely to have quality diets.
In the current study, gestational age was associated to MDD-W.
As suggested in an earlier study, balanced nutrition during gestation has been identified as a prerequisite for a healthy pregnancy and birth outcomes (Bawadi et al., 2010).  as pregnancy progresses, it is likely difficult to give the appropriate diet counselling to motivate pregnant women to keep to a quality diet made of a variety of food groups. In addition, it is also possible that family food supplies and resources may be mobilized at this stage in preparation for delivery ceremonies such as naming ceremonies that are highly valued and tend to be food and resource intensive in this setting.
Similar to our study, previous studies (Amugsi et al., 2016;Shamim et al., 2016;Willy et al., 2017) reported that dietary diversity is influenced by education and is significantly high among pregnant women who have higher educational achievement. Also, Bodnar and Siega-Riz (2002) confirmed that women who had better education consumed higher percentages of vegetables improving their diet quality. Another study by Ali et al. (2014)  SES also plays a critical role in pregnancy outcome (Saaka, 2013).
A significant association was found between SES and diet quality (Mbwana et al., 2016;Saaka, 2013 This study had some limitations. As diet quality was assessed using a recall method, we cannot rule out the potential for recall bias. If more educated women knew about the benefits of eating a diverse diet and therefore over recalled their consumption of specific foods to increase their diet diversity, our estimates on the relationship between high education and diet quality would be biased away from the null, implying that the real effect of higher education on diet quality is lower than reported. In addition, 24-h recall might not usually be feasible as a result of participants not being entirely truthful or complying; nonetheless, they provide representative data of the population as a result of lesser burden on study participants relative to diet record method (Thompson & Subar, 2017). Even though women of all pregnancy stages were included in this study, the use of health facilities as target outlets for sampling pregnant women likely introduced some selection bias by gestational age where women of early pregnancy in particular were less likely to attend health facilities for antenatal care. This possibly underrepresented women of early gestational age in our sample.
Our findings could have important policy and further research implications. Policies that address household food security problems are likely to impact diet quality of pregnant women and possibly birth outcome. Encouraging girl child education can also have impacts on diets. Interestingly, the government of Ghana recently implemented a policy of free education for all school going age from basic level to the second cycle level with appropriate feeding interventions (Abizari et al., 2020). This study can provide supportive evidence to encourage girl child education in this part of Ghana where female education is low. Household wealth increase can also be improved to ensure improved diet quality pregnant women. Policies targeting poverty reduction could include provision of basic and durable household assets. Larger prospective studies should further investigate factors identified in this study and the effect of wider environmental exposures such as climate change on diet quality of pregnant women.

| CONCLUSION
Diet quality among pregnant women was low and was predicted by educational level, household wealth, gestational age and food security.
Women education and improvements in household food security could impact diets of pregnant women in northern Ghana.

ACKNOWLEDGMENTS
We are thankful to the respondents for their participation. We are grateful to the Savelugu Nantong Health Directorate for giving us permission to conduct the study among pregnant women of the district.
We thank the staff of all the health facilities in the district for their assistance in organizing pregnant women and supporting in our data collection. We did not receive funding for this study.