Demographic and Socioeconomic Variations in Diet Quality Among Pregnant Women in Mbeya, Tanzania


 BackgroundSub-optimal diets are the primary risk factor for mortality among pregnant women. In Tanzania, many pregnant women have micronutrient deficiencies in iron, iodine, and vitamin A. Promoting healthy diets, with a focus on food quality and the consumption of fortified foods, may help to reduce mortality amongst pregnant women. However, more evidence is needed to inform the development of effective food based approaches. The aim of the study was to evaluate associations between demographic and socioeconomic factors with diet quality among pregnant women in Mbeya, Tanzania. MethodsData from a cross-sectional study conducted, in 2020, in Mbeya which included a population-based sample of 420 pregnant women was analyzed. Dietary intake was assessed using the Prime Diet Quality Score (PDQS) with data from 24-HR diet recalls. Socioeconomic variables were also collected. Chi square-tests and one way ANOVAs were used to assess differences across demographic and socio-economic predictors of PDQS. Multivariate analyses were conducted to adjust for confounders and effect modifiers. All tests were two-tailed and significance level was set at 5%. ResultsStudy participants had a mean age of 25.49 ± 6.37 years. The mean PDQS was 16.2±2.7. For the PDQS, the consumption of at least 4 servings per week of food groups was as follows: dark green leafy green vegetables (29.2%), other vegetables (14.7%) and vegetable liquid oil (57.2%). Consumption of refined grains was high (48.0%). Pregnant women who visited antenatal clinics (ANC) had a reduced diet quality. Occupational status and household wealth index were significantly associated with PDQS in high quintile groups. Marital status was negatively associated with the PDQS and, pregnant women who reported to visit ANC were positively associated with PDQS. ConclusionsPregnant women aged 15-49 years in Mbeya region have low PDQS scores due to high intakes of refined grains, limited consumption of red meats, and low intakes of healthy cruciferous vegetables, whole citrus fruits and poultry. These findings suggest that public health action is needed to improve diet quality amongst pregnant women to improve health outcomes of this population.


Background
Inadequate nutrition during pregnancy is a serious public health concern, with long-term effects both on mother and children's health [1,2]. Pregnant women are vulnerable to inadequate nutritional status because of the high nutrient demands during pregnancy. In low income countries, evidence suggests that malnutrition during pregnancy is associated with socio-economic constraints, as well as with environmental, cultural and political factors [3] . There is a considerable body of evidence supporting the association between inadequate nutrition during pregnancy and adverse pregnancy outcomes, poor infant survival, and risk of chronic diseases in later life [4].
Urbanization has a negative impact on the nutritional health of people living in areas of high deprivation. In 2019, the United Nations reported that more than half of the world's population (4.2 billion people) lives in urban areas and by 2041, this gure might increase to 6 billion people [5]. In many low and middleincome countries (LMICs) including Tanzania, urbanization is associated with a nutrition transition from traditional diets to more Western-style diets [6] [7] [8] [9]. These eating patterns are categorized by increased consumption of red meat, saturated fatty acids (SFA), sugar and re ned grains, coupled with reduced intakes of ber, fruits and vegetables [7] [8]. Evidence indicates that in LMICs, urban populations living in areas of low deprivation are more likely to embrace unhealthy Western style diets [10] [11], compared to the rural population, which has preserved a traditional diet categorized by consuming more vegetables, oils and whole grains, and less meat and SFA [11].
The assessment of dietary intake in LMICs is challenging. However, it is important to do so since the accurate assessment of dietary intake and nutritional status may aid in the development of population speci c nutritional interventions [12]. Moreover, such assessments will enable dietary de ciencies and excesses within the population to be identi ed, highlighting potential areas to target for nutritional supplementation and forti cation or dietary intervention and identi cation of at-risk groups.
One of the critical barriers to the development of population speci c nutritional interventions in LMICs, especially for vulnerable populations, is the lack of up-to-date, comprehensive and reliable data on individual dietary intake [12]. Unfortunately, most LMICS view dietary data as too complex or expensive to collect [12]. Furthermore, mechanisms that would enhance the collection and use of dietary data, including the use of new technology and a national infrastructure for collecting and using dietary data, in these countries rarely exists [12].
In Tanzania national dietary surveys are limited, the most widely cited is the National Nutrition Survey (TNNS) which was conducted in 2014 and 2018 [13]. The TNNS includes anthropometric measurements and data on supplementation and infant feeding practices, however, it does not include detailed information on dietary intake at either the household or individual level [13]. The key challenge in dietary assessment , especially in Tanzania, is di culties in assessing portion sizes of foods. The available methods such as food record, 24-hour dietary recalls and food frequency questionnaires are timeconsuming, costly, and skill/labour intensive [12]. Weighed Food Records (WFR), also called weighed food diaries or simply weighed records, are strong methods for estimating dietary intake [14], but they are also laborious for participants. WFR also require participants to have good literacy skills so that food intake over a period of three to seven full days can be accurately recorded [15]. An alternative method, the food frequency questionnaire, also requires participants to be literate and capable of estimating quantities of food and frequency of intake [15,16]. As such, these methods are challenging to use in LMIC countries, where much of the adult population is illiterate. An acceptable alternative is method is a 24 hour food recall questionnaire which is administered by trained eld workers [17]. Although these questionnaires are relatively easy to administer, this method may be subject to misreporting of energy and micronutrient intake [18]. A lack of speci c cultural and/or detailed knowledge of local foods on the part of the researchers, can also limit the accuracy of data collected by recall.
Previous studies have demonstrated that there is an urgent need for the improvement of diet quality among pregnant women. However, there is currently limited investment in understanding the role and interactions of socio-economic and demographic predictors to dietary quality among pregnant women in LMICs. Thus, the present study aims to evaluate the demographic and socioeconomic predictors of nutrient and food intakes, as well as of the overall diet quality among pregnant women aged 15-49 years in Mbeya region. The synthesized information provides novel insights into the in uence of socioeconomic and demographic predictors to dietary quality among pregnant women.

Study design
A cross sectional study conducted among pregnant women aged 15 and 49 years in the seven districts of Mbeya Region Tanzania. The study was carried out from September to October 2020.

Study area
The Mbeya region has a population of 2,204,543 (1,068,615 male and 1,135,928 female) and 557,574 women of reproductive age. The total deliveries in Mbeya region in 2020 were 72,076. There are 17 hospitals, 23 health centers, and 278 dispensaries where 251 health facilities provide reproductive and child health services. Our study was conducted at 42 Reproductive and Child Health (RCH) Clinics in seven (7) districts of Mbeya region. The selected RCH clinics in this study are estimated to provide services to approximately 1036 pregnant women.

Study Population
All pregnant women aged between 15 to 49 years old, within the rst and second trimesters (less than 28 weeks of gestation) and who attended antenatal visits were invited to participate in the study. A total of 574 pregnant women were invited and 420 (response rate 73%) agreed to participate. Pregnant women who refused to consent or who were unable to communicate due to illness and, those who were taking medication were excluded from this study.

Sample size and Sampling Procedure
A sample size (n = 420) was considered su cient based on the Lwanga and Lemeshow formula [18]. Prior to carrying out the study, the proportion of women of reproductive age with good diet quality was estimated to be 45%, with a margin error of 5%, a con dence level of 95% and a design effect of 1.5.
Another 10% was added to the sample size to account for non-responses. A sample size of 420 was considered satisfactory assuming a difference of 0.10 between groups of pregnant women with poor dietary quality and, those with good dietary quality, a signi cance level of 5%, a power of 90% and a design factor of 1.5.
The sampling procedure involved two steps. A list of 251 governmental and faith-based health facilities providing RCH services in Mbeya region was obtained and used in a random selection of the health facilities from each district based on probability proportional to size sampling. A total of 42 facilities from a pool of 251 were randomly selected for the survey. An additional two reserved clusters were included in the survey. Given the sampling frame of public health facilities in Mbeya, the probability proportional to size was performed to allocate the number of facilities per district for inclusion in the survey. Therefore, a total of 44 health facilities offering RCH services located in Mbeya region were visited and surveyed.

Data Collection
Prime Diet Quality Score (PDQS) Diet quality is an optimal measure of nutrients for wellbeing. The Prime Diet Quality Score (PDQS) was recently developed using a modi ed PrimeScreen questionnaire as a way to characterize diet quality globally and was associated with a lower risk of coronary heart diseases (CHD) in a large population in USA [20]. PDQS contains 21 food groups; 13 are healthy food and, 7 are unhealthy food. PDQS was assessed using 24-hour recalls, which re ected the feeding practice from the previous morning to the morning of the interview.
A standard structured questionnaire of PDQS constructed in English was translated into Kiswahili, the main language in Tanzania, spoken pro ciently by almost 95% of the population. The questionnaire was translated; from English into Kiswahili by bi-lingual Kiswahili/English, and then back translated to English by independent translators. IMAN project staff in the eld reviewed for semantic, experiential and conceptual equivalence to the original version. Sensitivity to culture and selection of appropriate words were considered. The structured questionnaire was piloted to a separate group of women (not part of this study) to evaluate the quality of the translations in terms of comprehensibility, readability and relevance to assess face validity.
Participants were asked 'from when you woke up yesterday till you woke up this morning did you consume the following food items: dark green leafy vegetables, cruciferous vegetables, dark orange vegetables and fruits, other vegetables, citrus fruits, other fruits, legumes, nuts and seeds, poultry, sh, whole grains, vegetable liquid oils, white roots and tubers, red meat as a main dish, processed meats, re ned grains and baked products, sugar-sweetened beverages, fried foods away from home, sweets and ice cream, low-fat dairy?' Responses were given on a 5-point Likert scale; 0= not at all, 1= once, 2= twice, 3= thrice and 4= fourth or more.We considered each occasion of consumption of a food group as a serving. We then computed the mean number of servings over the available recall days for each participant. The mean number of servings for each food group was multiplied by 7 to standardize to the number of servings per week, from which points for each food group could be assigned based on whether the food was categorized as healthy or unhealthy [19]. Points were assigned for consumption of healthy food groups as follows: 0-1 serving/week, 0 points; 2-3 servings/week, 1 point; and ≥4 servings/week, 2 points. Scoring for unhealthy food groups was assigned as follows: 0-1 serving/week, 2 points; 2-3 servings/week, 1 point; and ≥4 servings/week, 0 points [19]. To preserve the original scoring of up to 42 points (representing a high-quality diet), all participants were assigned two points for the egg's component, keeping it as a neutral component.

Demographic and socio-economic factors
A structured coded questionnaire programmed into the Open Data Kit (ODK) and administered using Android tablets were used to collect demographic and socio-economic variables. The variables considered were: age, marital status, education level, occupation status, and household income of pregnant women that attended ANC clinics in Mbeya region. Possible potential confounders were considered in the analysis such as maternal anthropometric measures mean upper arm circumference (MUAC), number of visits to the ANC and, anemia status.

Data analysis
The data were analyzed by using Stata 15 statistical package. The PDQS dependent variables were assessed both as continuous and ordinal variables (i.e. 0 = Low quintile; 1 = Medium quintile and, 3= High quintile). Ordinal PDQS variables were preferred to dichotomize PDQS. The three groups of PDQS were used to ensure that the loss of information was small. Independent variables such as household wealth index were assessed as an indicator of socio-economic status according to a standard approach in equity analysis [21]. Durable household assets indicative of wealth (i.e. telephone, radio, TV, refrigerator, lantern, cupboard, houses with electricity, motorcycle, bicycle, cart etc.) were recorded as (1) "available and in working condition" or (0) "not available and/or not in working condition." These assets were analyzed using principal components analysis, PCA. The rst component resulting from this analysis was used to categorize households into two approximate group's i.e. lowest and highest group.
Numeric variables were summarized using mean and standard deviation and categorical variables were summarized using frequency and proportions. At bivariate analysis, chi square-test and one way ANOVA was used to test the signi cant difference across demographic and socio-economic predictors of PDQS. In multivariate analysis, for adjusting confounders and effect modi er, both multi-linear regression and multinomial logistic regression was used. All tests were two-tailed and signi cance level was set at 5%.

Results
Demographic and socio-economic characteristics Study participants had a mean age of 25.49 ± 6.37 years. Table 1 depicts the socio-economic and demographic characteristics of the study population. More than half of participants were 15-24 years old. 71.7% had completed at least primary education and, 84.3% were self-employed. About one-third (31.0%) of the participants lived in houses with electricity, 72% had access to improved sources of drinking water and, 65% were not sharing toilet facilities. Furthermore, 57% of the participants were married.
Prime Diet Quality Score (PDQS) The mean PDQS was 16.2±2.7, with a range of 8 to 24 out of possible 42 points. Table 2 shows the consumption of PDQS food groups by pregnant women in the study. Only 29.2% of participants consumed ≥4 servings of dark green leafy green vegetables, 14.7% other vegetables and, 57.2% vegetable liquid oil per week. slightly similar trends for other healthy foods, including cruciferous vegetables, dark orange vegetables and fruits, whole citrus fruits, other whole fruits, legumes, nuts and seeds, poultry, sh, whole grains, white roots and tubers and low fat diary were also consumed infrequently by pregnant women. Re ned grains, baked goods, and sugar sweetened beverages were the most commonly consumed unhealthy foods. Consumption of red meats was higher in pregnant women in the lowest quintile (73.9%) as compared with pregnant women in the highest quintile of the PDQS (5.9%). Table 3 depicts the bivariate analysis that was conducted using a chi-square test. PDQS index was signi cantly associated with marital status, occupation status, household wealth index and visits to ANCs (p<0.05). However, the age group of pregnant women, their educational level, their source of drinking water, status of anaemia and number of pregnancies were not signi cantly associated with the PDQS quintiles (p>0.05). The corresponding demographic and socio-economic factors associated with PDQS were associated with marital status and visits to ANC (p<0.05).

Multivariate analysis
In multivariate analysis of the PDQS index, the tted models and the estimated effects are presented in table 4. We found a relationship between the independent variables and the PDQS index. All demographic, socio-economic and behavioral variables that were statistically (or directional) signi cantly associated with index were included in the multivariate analysis using a multinomial logistic regression in

Discussion
This study is novel in its examination of the demographic and socioeconomic predictors of nutrient and food intakes, as well as of the overall diet quality among a representative sample of pregnant women aged 15-49 years in Tanzania.
Dietary diversity scores have been developed to assess nutrient and energy intake adequacy among women and small children in under-resourced settings [20,21] [22] and, have been proposed as indicators for monitoring the implementation of the UN's Sustainable Development Goals (SDGs) [23]. The PDQS is an alternative index, recently developed as a diet quality index and was found to be associated with a lower risk of CHD in a large population in USA [19]. The present study found that diet quality among pregnant women in Mbeya measured by the PDQS was largely characterized by low intakes of nutrient-rich fruits and vegetables, low-fat dairy, sh, and nuts and seeds, and high intakes of re ned grains, baked goods, and sugar sweetened beverages. We also observed that the diet quality of the women varied signi cantly by marital status, and by the number of visits to ANCs, but contrary to our initial hypothesis, did not vary by age group, education level, household wealth income and occupational status. This nding was in line with other studies conducted among adults in Bosnia and Herzegovina in which demographic and socio-economic predictors of diet quality was associated with the BMI, type 2 diabetes status and marital status [11] These factors are not easily modi ed, while maternal diet diversity and quality are.

Strengths and limitations
There are several strengths of this study. The study had a large population sample and measured diet during the pregnancy using both the Prime Diet Quality Score tool and 24-hour dietary recall. This study provides information , in a LMIC, on the association of Socio-demographic variations with PDQS, a simple tool to incorporate analysis of diet quality that can be easily incorporated into programs. In addition, this study measured diet quality early in pregnancy, which is important for fetal development, given rapid cell growth, development of immune cells and organs in the rst trimester [37]. Among the strengths of the present study is that it links standard socio-economic data with dietary data and allows analysis of diet quality by SES indicators. However, there were several limitations of the study. First, we inevitably have some level of measurement error in both dietary and socio-economic variables, as both were based on self-report. This source of error is, however, expected to be largely random, producing valid estimates for the population as a whole just with somewhat higher SE. Second, we derived PDQS scores from 24-hour recalls, and there were limited precedents in published literature for converting these scores to equivalent scores for the food frequency questionnaire (FFQ), except for a study conducted in Bosnia and Herzegovina [11]. The validity of using the PDQS score assessed using 24 hour recall is an area of active research. Notably, 24 hour recall method is used widely in developing countries, and our ndings provide support to the use of this metric for deriving PDQS in these settings. Our ndings may not be generalizable to populations where dietary patterns and determinants of birth outcomes differ from rural Tanzania. Associations may be stronger in populations with more prevalent micronutrient and other de ciencies in pregnant women. Finally, the PDQS has not been validated in Tanzania or other lowincome settings. Further research is required to better understand the applicability of the PDQS in LMIC settings.

Conclusion
Pregnant women aged 15-49 years in the Mbeya region had low PDQS scores due to high intakes of re ned grains and red meats, and low intakes of healthy cruciferous vegetables, whole citrus fruits and poultry. These ndings suggest that public health action is needed to promote higher consumption of healthy cruciferous vegetables, whole citrus fruits and poultry, and a higher variety of fruits and vegetables. PDQS, as a measure of maternal diet quality, was inversely associated with marital status. PDQS was also positively associated with visits to ANC. These ndings suggest that diet quality should be considered as an important factor in understanding risk factors for poor birth outcomes. The results discussed in this paper can also be used to provide baseline data that can be sued to investigate dietary quality in Tanzania more generally. Such understandings could then be used to identify solutions to enable low-income pregnant women to afford and make healthful food choices. Moreover, the ndings presented here could contribute to policy decisions related to tackling the problem of socio-economic disparities in pregnant women nutrition at the national level. For more insights, further studies exploring these scoring systems in LMICs is warranted. Intervention trials should evaluate whether increasing diet diversity and quality can improve maternal and infant health outcomes. Ethical clearance was obtained from National Institute for Medical Research Ethical Committee-(NIMR/HQ/R.8a/Vol.IX/2589). All eligible subjects were informed on the purpose and nature of the survey and those who agreed to participate were asked to sign a written informed consent form.

Consent for publication
Not applicable

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

25.
Zerfu, T.A., M. Umeta, and K. Baye, Dietary diversity during pregnancy is associated with reduced risk of maternal anemia, preterm delivery, and low birth weight in a prospective cohort study in rural Ethiopia. The American journal of clinical nutrition, 2016. 103 (6)     The reference category is: Medium PDQS Quintile.