Reliability and validity of a General Nutrition Knowledge Questionnaire for adults in a Romanian population

In Romania, the profession of dietitian was only legally recognized in 2015, with application norms published in early 2019. Prior to this, nutrition-related guidance was provided mainly by medical doctors. Literature suggests higher nutrition knowledge is generally associated with healthier lifestyles and food choices, though the strength and consistency of this relationship are debated. Multiple factors—including socioeconomic status, traditions, taste preferences, and genetics—also shape dietary habits, and prior research indicates Romanian food choices are strongly influenced by economic considerations. Given the lack of validated instruments to assess general nutrition knowledge in Romanian adults, this study aimed to translate, culturally adapt, and evaluate the reliability and validity of a General Nutrition Knowledge Questionnaire (GNKQ) for use in Romania.

The study builds on established GNKQ instruments. Parmenter and Wardle originally developed an English-language GNKQ widely used and adapted across countries and purposes. Kliemann et al. later updated the GNKQ to reflect current expert recommendations. Prior research links higher nutrition knowledge to healthier behaviors, although evidence is mixed and influenced by other determinants (e.g., socioeconomic status, traditions, taste, genetics). In Romania, economic factors have been shown to substantially drive food choices, underscoring the need for a validated, context-appropriate knowledge assessment tool.

Design: Translation, cultural adaptation, and psychometric evaluation of Kliemann’s updated General Nutrition Knowledge Questionnaire. Adaptation steps: - Step 1: Forward translation (English→Romanian) and back-translation by independent translators; discrepancies resolved with minor corrections. - Step 2: Expert panel (3 experts in human nutrition/dietetics, nutritional epidemiology, nutrigenomics/nutrigenetics) reviewed all items for cultural relevance; English-specific foods replaced with Romanian equivalents. - Step 3: Pretesting with 25 volunteers (including a Romanian language specialist, medical doctors, and medical students) to refine wording; minor improvements made. Psychometric components and samples: - Component 1 (Internal reliability): General population sample n₁ = 412. - Component 2 (External reliability/test–retest): Subsample n₂ = 46 from Component 1, retested ≥30 days later. - Component 3 (Construct validity): Known-groups method with n₃ = 96 (48 dietetics specialists: senior undergraduates, master’s students, recent graduates ≤3 years; 48 nonspecialists: mathematics/informatics undergraduates). Sample size powered for large effect size (≥80% power, one-sided tests). - Component 4 (Convergent validity): n₄ = 508 combining participants from Components 1 and 3; powered at ≥80% with 5% margin of error to examine associations between sociodemographic characteristics and nutrition knowledge. Questionnaire content: Four sections—(1) Dietary recommendations; (2) Food groups; (3) Healthy food choices; (4) Diet, disease, and weight associations—plus demographics (gender; self-rated health, 5 levels; marital status, 6 categories; number of children; presence of minors in household; highest education, 8 levels; prior nutrition training, yes/no; self-reported height and weight for BMI calculation). BMI categorized as normal vs overweight/obese. Ethics: Approved by the Institutional Review Board at Victor Babes University of Medicine and Pharmacy Timisoara; informed consent obtained from all participants. Data collection: Paper-based through Step 4; online for subsequent steps under supervised conditions to minimize information sharing and external resource use. Scoring systems: - Type A: +1 for correct; 0 for incorrect or “I don’t know.” - Type B: +1 for correct; −1 for incorrect; 0 for “I don’t know.” Type B used for Component 4; Type A used for Components 2, 3, and 4. Achievement scores computed per section and overall, relative to maximum possible scores. For Component 4, the difference between Type A and Type B was used to indicate tendency to provide incorrect answers. Statistical analysis: IBM SPSS v21. Descriptive statistics (means ± SD, medians [IQR], percentages). Group comparisons: chi-square for proportions; t-tests with Levene’s test for two-group parametric comparisons; Mann–Whitney for nonparametric/ordinal two-group comparisons; Kruskal–Wallis for ≥3 groups with Bonferroni-adjusted post hoc tests. Wilcoxon signed-ranks test compared scoring systems (Type A vs Type B). Linear regression modeled the difference between Type A and Type B as the dependent variable with demographic factors as predictors.

Reliability: - Internal consistency (Cronbach’s alpha) in general population (n=412): overall α = 0.879 (>0.70). By section: Section 1 (Expert recommendations) α = 0.53; Section 2 (Food groups) α = 0.82; Section 3 (Healthy food choices) α = 0.53; Section 4 (Diet, disease, and weight associations) α = 0.72. - External consistency (test–retest ICC) in subsample (n=46): overall ICC = 0.96. By section: Section 1 ICC = 0.94; Section 2 ICC = 0.96; Section 3 ICC = 0.88; Section 4 ICC = 0.86. Construct validity (known-groups, n=96): Specialists scored significantly higher than nonspecialists across all sections and total score with very large effect sizes (Cohen’s d): - Section 1 (max 18): 13.9 ± 2.0 vs 10.6 ± 1.9; d = 1.68; p < 0.001. - Section 2 (max 36): 28.3 ± 3.4 vs 18.6 ± 4.6; d = 3.39; p < 0.001. - Section 3 (max 13): 10.6 ± 1.7 vs 8.3 ± 2.1; d = 1.18; p < 0.001. - Section 4 (max 21): 17.2 ± 2.5 vs 12.1 ± 3.2; d = 1.77; p < 0.001. - Total (max 88): 70.0 ± 7.3 vs 49.7 ± 9.6; d = 2.38; p < 0.001. Convergent validity (n=508): - Type B scoring yielded significantly lower median achievement scores than Type A across all sections and overall (p < 0.001). - Demographic associations: Females scored higher than males; middle-aged and older adults scored higher than young adults; higher education associated with higher knowledge scores. Predictors of giving incorrect answers (regression using difference Type A − Type B): - Male sex: beta = 0.170; approximately 2.6× more likely to give incorrect answers. - Lower education (high school or less): beta = 0.167; approximately 2.3× more likely. - No prior nutrition training: beta = 0.154; approximately 3.2× more likely.

The study addressed the absence of a validated Romanian instrument to measure general nutrition knowledge by translating, culturally adapting, and rigorously evaluating the updated GNKQ. Internal consistency was acceptable to high overall and comparable to international adaptations, while section-level values reflected known variability in item heterogeneity for dietary recommendations and healthy choices. Test–retest reliability was strong across all sections and overall, indicating temporal stability. Construct validity was demonstrated by markedly higher scores among nutrition specialists compared with nonspecialists, with very large effect sizes across all sections and total score, confirming the instrument’s ability to discriminate between known groups. Convergent validity analyses showed expected associations between knowledge and sociodemographic factors, including higher scores among females, older age groups, and those with higher education. Regression analyses further highlighted groups more prone to incorrect responses (males, lower education, no nutrition training), offering actionable insights for targeted education. Collectively, these findings support the questionnaire’s reliability and validity for assessing general nutrition knowledge among Romanian adults.

The Romanian adaptation of the General Nutrition Knowledge Questionnaire demonstrates good internal consistency, excellent test–retest reliability, strong construct validity, and meaningful convergent validity with sociodemographic factors. It provides a robust tool for assessing nutrition knowledge in Romanian adults and can be used in future research, surveillance, and evaluation of nutrition education interventions and policies.