A diet quality index is a convenient and effective indicator of overall diet quality (1, 2). These indices capture information about a group of dietary components and are assumed to partially account for synergistic interactions between nutrients and/or food groups (3). Consequently, epidemiological studies are placing growing emphasis on exploring associations between diet quality indices and specific health outcomes, moving beyond a focus solely on single nutrients and foods. To illustrate, a meta-analysis on 18 studies concluded that a one unit higher maternal diet quality was associated with a one unit better child neurodevelopment score (1). Moreover, a low Healthy Eating Index 2015 score (<70) among 762 US pregnant women was associated with a higher risk of fetal growth restrictions and hypertensive disorders during pregnancy (RR = 0.33, 95%CI: 0.13–0.68 and RR = 0.46, 95%CI: 0.24–0.79) (4). In a study among 660 pregnant Mexican women, women in the highest tertile of the Maternal Diet Quality Score compared to lowest tertile had a reduced risk of a baby with low birth weight (OR = 0.22, 95%CI: 0.06–0.75) (2).

In addition to its application in epidemiological contexts, a diet quality index is a straightforward parameter to understand and apply by health care professionals, policy makers as well as the general population (3). Therefore, diet quality indices are valuable tools as part of mobile Health (mHealth)-applications to facilitate self-monitoring and providing more individualized feedback in clinical practices to improve diet quality (3, 5). In fact, diet quality indices improved dietary habits of users in 60% of previously reviewed apps (6). However, mHealth-applications using diet quality indices for pregnant women are currently not available.

A diet quality index can consist of several different components (e.g., food groups, macro or micronutrients), scoring methods (e.g., ratios, moderation, and optimum), weighting of components (e.g., dependent on health effects) and can be based on different dietary recommendations (5, 7). While several diet quality indices have been developed for pregnant women across the globe (8–11), no diet quality exists yet for Dutch pregnant women. However, there is a Dutch Healthy Diet 2015 (DHD-2015) index based on the Dutch dietary guidelines of 2015 for the general adult population, excluding pregnant and lactating women (5, 12). More recently (2021), the Health Council of the Netherlands developed dietary guidelines for pregnant women (13).

The main objective of this study was to refine the existing DHD-15 index, aligning it with the latest Dutch dietary recommendations for pregnant women (DHD-P), using data of the GLIMP-II study. This refinement serves as a crucial step in establishing the foundation for a personalized nutrition app tailored to the specific dietary needs of pregnant women. Subsequently, we studied the correlation between the DHD-P index calculated using food frequency questionnaire (FFQ) data and using 24-h recall (24hR) data, and examined the index’s ability to detect changes over time at 12 and 24 weeks of gestation.

Table 2 presents the baseline characteristics for the total population and stratified based on diet quality score. On average, women were 33 years old (SD: 3) with a BMI of 27 kg/m² (SD: 5). Most women were married (84%), had children (96%), identified as western ethnicity (98%), and had a high education level (73%). Only 5% (n = 2) smoked, and 9% (n = 4) had a history of gestational diabetes. Although not statistically significant, there were noticeable differences in education between women with low and high DHD-P scores (59% versus 86% high educated, p = 0.11).

Percentages of total energy from macronutrients were comparable for FFQ and 24hR data at 24 weeks (Table 3). FFQ data indicated higher intakes of vegetables, legumes, total fish, lean fish, fatty fish, liquid fat and iron compared to the recall data, while the opposite trend was observed for vitamin A and caffeine intake. A significant difference was also observed when comparing FFQ and recall data for the total diet quality score, i.e., 113 (IQR: 100–126) and 104 (IQR: 96–109) (p < 0.001), respectively. Examining individual component scores showed that vegetables, fruit, legumes, fish, fats, caffeine, vitamin A and iodine scored significantly higher in FFQ compared to 24hR data. The processed meat score was higher in 24hR data compared to FFQ data [4 (IQR: 0–10) versus 3 (IQR: 0–5)]. All other components did not significantly differ between FFQ and 24hRs.

Despite significant absolute differences in the total diet quality score for the FFQ and recall, the de-attenuated correlation coefficient between the two methods was 0.57 and classified as good. Similarly, correlations for individual component scores when comparing FFQ and 24hR data were moderate or good for most components (Table 3). Specifically, correlations between component scores based on FFQ and 24hR data were moderate for whole grains, nuts, dairy, processed meat, caffeine, unhealthy choices, vitamin A and iodine (rho = 0.25–0.49) (35). For fruit, vegetables, sugar-sweetened beverages, red meat and iron correlations were good (>0.50). A low correlation was observed for fish (rho = 0.09). The uncorrected correlation coefficients for the total diet quality score, whole grain products, nuts, dairy, sweetened beverages and fruit juices, caffeine, unhealthy choices, vitamin A and iodine were similar to the de-attenuated values. No de-attenuated correlation coefficients could be calculated for legumes (Spearman rho = 0.44) and fat (Spearman rho = 0.20) due to high within person variation. A de-attenuated correlation coefficient for alcoholic beverages and soy products could not be calculated due the high similarity in consumption between FFQ and 24hR data, resulting in minimal variance and within person variation.

An overview of diet quality based on FFQ data at 12 and 24 gestational weeks is presented in Table 4, revealing significant absolute differences for the intakes of red meat [43 (IQR: 29–58) vs. 39 (IQR 23–47) g/day, p = 0.02], caffeine [2 (IQR: 1–2) versus 2 (IQR: 1–3) cups/day, p = 0.04] and supplemental folic acid [400 (IQR: 400–400) versus 400 (IQR: 0–400) mcg/day, p = 0.02]. No significant differences were observed for any other DHD-P components. Accordingly, evaluation of individual component scores did not significantly differ over time, except for red meat [diff = −0.9 (95%CI:−1.5; −0.3)], and the total diet quality score showed a correlation coefficient of 0.69 over time, with individual components correlations ranging from 0.27 for red meat to 0.84 for vegetables.

Consistent with the FFQ data, recall data showed a non-significant increase in energy intake over time (Table 5), with similar values for folic acid supplementation. In contrast to the FFQ data, recall data indicated lower refined grains [53 (IQR: 15–79) vs. 94 (IQR: 69–127)] and processed meat consumption at 24 weeks compared to 12 weeks gestation [24 (IQR: 0–49) and 41 (IQR: 14–70) g/d]. No significant differences in intakes were observed for any of the other DHD-P components. Accordingly, DHD-P scores did not substantially differ between 12 and 24 weeks, except for a higher score for grains at 24 weeks compared to 12 weeks [diff = 1.3 (95%CI: 1.0–2.5)]. The correlation coefficient for the total diet quality score between 12 and 24 weeks of pregnancy was 0.22 with correlation of individual components ranging between 0.02 for red meat and 0.69 for caffeine.

In this study, we developed the DHD-P index using the recommendations from the Dutch Health Council for pregnant women and evaluated it using GLIMP-II study data. Despite significant differences for several absolute intake estimates and individual component scores when comparing FFQ and 24hR data at 24 weeks gestation, de-attenuated correlation coefficients were moderate or good for almost all components. Moreover, FFQ and recall data both showed comparable dietary intake data for daily consumed food groups at 12 and 24 weeks, suggesting minimal changes during pregnancy. Further, this study demonstrated moderate or good correlations over time of the adjusted DHD score based on FFQ data compared to 24hRs data among Dutch pregnant women indicating good reproducibility.

Various diet quality indexes for pregnant women have been developed worldwide, using different dietary guidelines and scoring methods. For instance, the Diet Quality Index for Pregnancy (DQIP) including eight components was developed by Bodnar and Siega-Riz (10) based on FFQ data of 2,063 women in the second trimester while applying the US dietary guidelines. In contrast to the DHD-P’s different scoring for folic acid supplementation based on the specific stage of pregnancy, the DQIP maintained consistent scoring across trimesters. In terms of the number of food groups assessed in previously developed indices, our index markedly differs from a Brazilian diet quality score, developed by Crivellenti et al. (11). Whereas our index consists of 14 food groups and eight nutrient components, the Brazilian index is composed of five nutrient components, three food group components and one moderator component reflecting the percentage of total energy from ultra-processed foods (11). Kennedy et al. (8) developed an index based on EFSA, WHO and USA Institute of Medicine guidelines, which scored the adherence to the guidelines for each nutrient or food group with 1 point, weighting all food groups and nutrients the same. Further, the Healthy Food Intake Index containing 11 components based on the Nordic Nutrition Recommendations was scored ranging between 0 and 1 or 0 and 2 depending on the relevance for overall diet quality (9). Conversely, the DHD-P applies an equal weighting approach for all its components due to the challenges in assigning weights to components based on health effects. Variations in indices can be explained by variations in the underpinning recommendations, divergence in the employed dietary assessment methodologies during index development, and the prioritization choices made by developers (e.g., specific emphasis on certain food groups, nutrients, or health outcomes).

In this study, we observed median diet quality scores of 133 and 134 out of 190 based on FFQ data at 12 and 24 weeks gestation. In comparison, Nguyen et al. (38) observed a mean diet quality score of 7.6 out of 15 among 4,824 Dutch pregnant women who completed an FFQ between 2002 and 2006. They used a scoring system based on the 2015 Dutch dietary guidelines for adults, incorporating folic acid supplements, and assigning a maximum score for alcohol abstinence (38). Thus, scores in our study were higher than the scores observed by Nguyen et al. (38), which may potentially be explained by a higher percentage of high educated women in the GLIMP-II study (73% versus 47%). Further, we observed a good correlation between the diet quality index calculated with the FFQ and 24hRs, which is highly comparable to the results of the evaluation of the original DHD-index among 885 adults (5). Most component specific coefficients between the two methods were also comparable between our study and the original DHD-evaluation, including lower correlation coefficients for episodically consumed foods. Considering the absolute intake levels, the FFQ data of our study showed higher intake levels for 7 out of 28 nutrients or food groups compared to the average of the two 24hRs, particularly for some episodically consumed foods such as fish and legumes. A meta-analysis including 130 studies evaluating validity of the FFQ for epidemiological studies among adults (39), also showed higher intake levels for 17 out of 32 nutrients for FFQ data vs. recall data. Moreover, a validation study on the FFQ used in this study observed higher reported levels of 8 out of 30 nutrients compared to recall data (17). In light of the inherent attributes of the two methodologies, the observed outcome in our study is not unexpected. The FFQ is anticipated to more effectively capture habitual food intake when contrasted with two 24-h dietary recalls. As a result, the FFQ is expected to provide a more accurate estimation of the habitual intake pertaining to episodically consumed foods (5). Given that the total diet quality score and most individual component scores exhibited moderate or good correlation when comparing FFQ and 24hR data, we consider the DHD-P index a useful tool to compare scores between pregnant women and for self-monitoring (35).

When comparing diet quality between 12 and 24 weeks gestation, we observed limited variation in absolute intake levels and diet quality component scores of commonly consumed food groups for both FFQ and 24hR data indicating minimal dietary changes during pregnancy. Particularly if differences existed these are expected to become evident using 24hR data as these provide more detailed dietary information compared to FFQs. It can be questioned whether diet remains completely similar during pregnancy trimesters as, e.g., women may be more motivated to eat healthy at the beginning of pregnancy (40, 41), but many other factors, such as physical complaints, could play a role in dietary habits during pregnancy as well (41). Although only a limited number of studies explored dietary intake at different stages during pregnancy, a study among about 2,500 UK pregnant women also observed minimal alterations in dietary patterns (42). In line, no changes in overall diet quality were observed over pregnancy trimesters among 79 Canadian pregnant women. However, in terms of individual components, intakes of fruit and vegetables decreased and ‘milk and alternatives’ increased during pregnancy among the Canadian women (41). Based on the FFQ data of our study, we observed a similar trend for dairy intake. In terms of reproducibility, FFQ data showed moderate to good correlations, while those based on 24hRs were categorized as low to moderate according to Lombard’s criteria (35, 43). It is essential to emphasize that correlation coefficients signify the relatedness between measurements over time, not the level of agreement (44). Consequently, this suggests that scores over time based on FFQ are more closely intertwined compared to those based on 24hR data. This difference may again be attributed to substantial day-to-day variation, particularly for episodically consumed foods like fish, legumes, and associated nutrients (e.g., iodine) when relying on only two 24hRs.

So far, a few diet quality scores exist for Dutch pregnant women, but these were not evaluated and are not based on the recent guidelines. We used data from the GLIMP-II study including intake levels based on both FFQ and two 24hRs at two time points during pregnancy to exploratively evaluate the DHD-P index. While correlation coefficients can assess reproducibility and indicate the validity of long-term dietary intake measures, it is crucial to recognize that high reproducibility does not necessarily ensure high validity (43). Limitations of this study are its limited sample size, and the inclusion of women with a predisposition for gestational diabetes, reducing external validity to the generally pregnant population in the Netherlands. Additionally, both 24hRs and FFQ have its limitations in assessing habitual dietary intake as they both rely on memory and a food composition table to calculate nutrient values. Furthermore, the 24hRs used in GLIMP-II were web-based, which likely resulted in lower reporting rates as no additional questions could be asked in comparison to when performed during a phone call or meeting in person. Additionally, the FFQ used in this study was not designed to inquire about nutrients or food groups specifically of interest during pregnancy, such as vitamin A consumption via liver products. Distinguishing between types of products (e.g., low or high fat milk) in FFQs depends on question specificity. Despite attempts to score similar food products in both 24hR and FFQ data, differences may have occurred from the FFQ’s broader product inquiries compared to the 24hRs. Another limitation of the current evaluation study is the inability to calculate the intake of all intended components, such as filtered coffee and salt intake. To note, the Eetscore tool, developed by researchers from Wageningen University, can be used as a diet screener to collect all this necessary dietary data to calculate DHD-15 scores (45). Therefore, using the updated Eetscore tool in combination with the DHD-P potentially provides the optimal assessment method for diet quality among Dutch pregnant women.

This study presents the development of the DHD-P index based on Dutch dietary guidelines for pregnant women issued by the Dutch Health Council in 2021. Additionally, we performed an explorative evaluation on the DHD-P to provide insight in the index’s usability and reproducibility. The DHD-P appears to be an appropriate index to assess dietary quality among pregnant women, and could serve as a foundation to provide dietary feedback toward healthier food choices. Further evaluations using more complete dietary data specifically relevant in respect to the dietary guidelines for pregnant women and the DHD-P, such as salt intake and type of coffee consumption, are recommended.

The data analyzed in this study is subject to the following licenses/restrictions: the raw data supporting the conclusions of this article will be made available by the authors, upon reasonable request. Requests to access these datasets should be directed to elske.brouwer-brolsma@wur.nl.

The studies involving humans were approved by Medical Ethics Committee of Wageningen University & Research. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

JF: Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing. EF: Data curation, Funding acquisition, Resources, Supervision, Writing – review & editing. EB-B: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Resources, Supervision, Writing – original draft, Writing – review & editing.