The Impact of Chinese Adult's Food Literacy on Healthy Eating Intentions Based on the Planned Behaviour Theory

Unhealthy diets are major contributors to obesity and chronic diseases. In 2023, 50.7% of Chinese adults were overweight or obese, underscoring the need to strengthen healthy-eating intentions.

We analysed a cross-sectional online survey of 1145 adults (18-64 years) from Henan and Shandong. Moderation was tested using multiple linear regression with mean-centred interaction terms between each Theory of Planned Behaviour (TPB) construct (attitude, subjective norms, and perceived behavioural control [PBC]) and each food-literacy component (production, choices, preparation and cooking, intake, disposal). Models were adjusted for age, occupation, marital status, alcohol use, physician-diagnosed chronic disease, and living with family. To address multicollinearity, we performed a ridge-regression robustness check (L2-regularised linear model; λ = 0.02 selected by 10-fold cross-validation; CV-RMSE = 0.483; CV-R² = 0.631). We report B, SE, β, p-values, and R²/adjusted R².

The overall food-literacy score did not significantly moderate the associations between attitude, subjective norms, or PBC and healthy-eating intention (p = 0.328, 0.671, 0.985). In component-wise analyses, only intake (intake) significantly moderated the PBC-intention association (B = 0.002, SE = 0.001, t = 2.497, p = 0.013); in the ridge model, the effect remained positive (β = 0.182; λ = 0.02). PBC (β = 0.459) and subjective norms (β = 0.169) were the strongest main-effect predictors. The best-fitting model explained R² = 0.663 of the variance in intention (adjusted R² = 0.663).

Among adults in Henan and Shandong, the intake component of food literacy strengthened the association between PBC and healthy-eating intention, whereas overall food literacy showed no general moderating effect. Interventions should prioritise intake-related skills (e.g., portion planning, lower-sodium choices and nutrition label use) to enhance perceived behavioural control and, in turn, intention. Given the cross-sectional design, causal inference is limited; longitudinal, capability-building evaluations are warranted.