Health literacy networks among older adults: central components and network differences by age and sex

Although previous studies have compared aggregate health literacy scores across groups, none have examined how individual competencies interrelate at the item level. Network analysis models conditional dependencies among items and identifies the most interconnected competencies. We aimed to construct and compare health literacy network models according to age and sex among older adults in South Korea. In this cross-sectional secondary analysis, we used data from the 2023–2024 Korea National Health and Nutrition Examination Survey and included 3,487 community-dwelling adults aged ≥ 65 years. Three hundred individuals with missing health literacy data were excluded; they were older, more often female, and had lower educational attainment. A 10-item health literacy instrument was dichotomised into experiencing difficulty (‘strongly agree’ or ‘agree’) and not experiencing difficulty (‘strongly disagree’ or ‘disagree’). Ising networks were estimated by age (young-old [65–74 years], n = 2,182; old-old [≥ 75 years], n = 1,305) and sex (male, n = 1,526; female, n = 1,961) using IsingFit (eLasso with EBIC model selection) without incorporating complex survey weights. Expected influence (EI) was calculated, and stability was assessed using case-dropping bootstrapping (1,000 samples). Network comparison tests were conducted using 1,000 permutations. Health literacy item distributions differed by age and sex. Judging which everyday behaviours are related to health (HL_4) showed the highest EI in the young-old (EI = 1.61) and female (EI = 1.67) groups, whereas understanding patient education materials (HL_8) had the highest EI in the old-old (EI = 1.58) and male (EI = 1.08) groups. Global network strength did not differ by age ( P = 0.972) or sex ( P = 0.051), and no age-specific edge-level differences were observed. Four uncorrected sex-specific edge-level differences were observed ( P = 0.015–0.045); these were considered exploratory because no adjustment for multiple comparisons was applied. EI correlation stability coefficients ranged from 0.361 to 0.517. The most central health literacy components varied descriptively across age and sex groups, while overall network connectivity did not differ significantly between groups. Exploratory analyses also suggested sex-specific edge-level associations. These findings provide a basis for generating hypotheses regarding age- and sex-tailored approaches to health literacy.