Several large, national studies have assessed energy and macronutrient intake patterns at the household level in Vietnam since the 1990s. However, these studies were based on household level data and none focused specifically on WRA. The generalizability of dietary intakes based on household level data is questionable especially for WRA. Previous work has shown differences in intra-household patterns of food consumption and the role of gender bias that do not favor WRA especially in South and South-East Asia . Moreover, studies using data from the Vietnam Household Living Standard Surveys relied on expenditure data to compute intakes, and did not specify how intakes were estimated for foods without information on amounts purchased . Other studies employed aggregate consumption data across food groups to estimate energy and nutrient consumption from Vietnamese  or international food composition tables [3, 22], and treated foods consumed outside the home as a single category. In contrast, this study presents data obtained at the individual level from WRA using a previously validated semi-quantitative FFQ and a food composition database compiled specifically for use in Vietnam, estimating energy and nutrient intakes from all foods, whether consumed inside or outside the home.
The mean energy intake in our study population was 2,196 kcal/day, similar to estimates from previous studies [1, 4, 5, 10]. We found that the mean energy intake was higher than the mean EER by 386 kcal, but the poor correlation between measured energy intakes and EER values suggests that comparison of means may have little value. Furthermore the estimation of energy requirements using the IOM formula  may be problematic since it is based on Caucasians and does not account for possible differences in body composition. It is therefore impossible for us to assess whether the actual intakes are low or excessive, because the biases in these estimates are unknown. However, we were still able to identify significant relationships between reported energy intakes with household food security and SES in the expected direction which is useful for policy makers.
The most recent NIN survey (2010) found that protein, fat, and carbohydrates constituted 14%, 18%, and 68% of total energy in the northern midlands and mountainous region  which are similar to our findings of 14.8%, 19.5%, and 65.5%. Although these results suggest improvements in the consumption of animal based foods and reduced consumption of staples compared to previous studies [1, 10, 23], there remain dietary imbalances mainly derived from a large proportion of total energy intake from carbohydrate. The results also suggest that food consumption patterns are not improving as fast as the country’s economy.
Nearly all our participants were within the optimal range for protein consumption, which is beneficial in maintaining lean muscle mass and reducing the incidence of chronic disease . However, only a third of protein came from animal source foods which have high bioavailability for several micronutrients. The percent of energy coming from animal sources is greater than reported in previous studies  and the percentage of energy from rice and other staples has decreased from an estimated 80% in 1998  to 66% in our study, implying an improvement in dietary diversity. However, the majority of participants still consume 66% of their total energy from carbohydrates, particularly from rice and other staples, indicating poor dietary diversity overall. Further improvements in dietary diversity could play an important role in decreasing undernutrition among WRA and improving micronutrient intake .
Fat intakes were low and more than half (56.5%) consumed <20% energy from fat which is the lower limit recommended by IOM . Increasing fat intakes could help increase energy intakes especially among food insecure households in developing countries, but ecological and observational data suggests that shifting from a low fat diet to a high fat diet as a percent of total energy is associated with an increase in unhealthy weight gain, and potentially contributes to obesity, diabetes, and other chronic conditions . We were able to examine the type of fat consumed using the Vietnamese food composition tables that were revised in 2007 and have reliable data on the fatty acid composition of foods that were determined using chromatographic methods . Mean saturated fatty acid and cholesterol intakes are within levels recommended by the IOM to prevent cardiovascular and other chronic diseases  but the intakes of n-3 and n-6 LCPUFAs are far below the recommended dietary intakes. This is a concern in light of recent evidence demonstrating the benefits of n-3 fatty acids for improving pregnancy outcomes and reducing the risk of cardiovascular disease [25–27]. Therefore, increasing fat intakes, especially n-3 fatty acids, may be beneficial if this is accompanied by reductions in carbohydrates, which is associated with increased risk of diabetes and cardiovascular disease .
We used the AMDR values recommended by the IOM to evaluate the adequacy of macronutrient intakes  instead of the Vietnamese recommendations  because of the lack of values for protein. The suggested ranges for fat are similar in both recommendations but the AMDR for carbohydrates are much higher and narrower in the Vietnamese recommendation (61-70%) compared to IOM (45-65%). The Vietnamese recommendation classifies an estimated 26% of women with deficient carbohydrate intakes, compared to only 0.6% using the IOM and to 2.2% using the WHO recommendations . It seems unlikely that around a quarter of Vietnamese women have deficient carbohydrate intakes. For these reasons we decided to use the ranges recommended by the IOM .
Ethnic minorities make up a disproportionately large proportion of the poor in Vietnam and they have significantly lower living standards, including less access to healthcare and education . Childhood malnutrition among rural minority populations did not decrease during the 1990’s as it did among the Kinh majority , and minority children under five were nearly twice as likely to be both moderately and severely malnourished in 2006 . Minorities have also been reported to be more likely to be anemic . Nearly 50% of our study population sample identifies itself as coming from an ethnic minority. Contrary to expectations, dietary energy intakes (overall and from protein) were greater among minorities compared to the majority Kinh in our study population, after adjusting for other factors. There were no differences in the amount of energy from fat and carbohydrate or the proportion having low protein and fat intakes. To our knowledge, no other study in Vietnam has examined food consumption pattern among ethnic minorities.
We found that food security and SES were the significantly and positively associated with energy intakes (overall and from protein and fat), and negatively associated with carbohydrate intakes. The odds of having energy intake lower than the EER, and insufficient intake of protein and fats were also higher among those from lower SES and food insecure households. These differences were not only statistically significant, but large and of public health significance. For example, energy intake was lower by 151 kcal among food insecure compared to food secured women and lower by 271 kcal in the lowest quintile compared to the highest quintile of SES. Similarly, insufficient fat and excessive carbohydrate intake were around 30% higher for food insecure compared to food secured women; and 50% higher for the lowest quintile compared to the highest quintile of SES. While increasing SES does not necessarily translate into improved nutritional status in developing countries , these findings suggest that overall intake, protein intake, and dietary diversity can be improved by targeting interventions to the poor and food insecure. Higher education was also associated with higher intakes of animal foods, which has been shown in previous studies and can be viewed as a marker of better diet quality in this population [33, 34]. We did not find major differences by occupation except that farmers were more likely to have lower fat intakes.
Key strengths of our findings include: 1) the large sample that is fairly representative of the northern mountainous areas of Vietnam, 2) the use of standardized methods that rely on a previously validated semi-quantitative food frequency questionnaire administered by well-trained interviewers, and 3) detailed information on sociodemographic characteristics that include measures of food security, ethnicity, socioeconomic status and education. Our findings can be generalized to the population of WRA in Thai Nguyen province and to the wider region of northern Vietnam with important policy implications for improving women’s health and nutritional status. The availability of detailed information on locally consumed foods with a food composition table designed specifically for the local diet are also important strengths of the study that allowed us to carefully examine the overall and detailed profile of intakes by nutrient and type of food. Nevertheless, there are some important limitations that call for caution in the interpretation of our findings. Most notably, the weak correlation between actual energy intakes and estimated energy requirements is a major concern which limits our ability to make conclusions on the adequacy of energy intakes in this population. This may be due to measurement error and/or the need for better methods to estimate energy requirement in this study population. For example, the use of repeated 24 hour recalls to assess intakes and/or the inclusion of measures of physical activity would have been useful. However, we are confident of our findings related to the patterns and determinants of macronutrient consumption as percentages of energy intake that have important implications for future work in this population.