This study employed a cross-sectional design using nationally representative data from the KCHS, conducted annually by the Korea Disease Control and Prevention Agency (KDCA). The KCHS is the largest nationwide health survey in Korea, including approximately 230,000 participants each year and representing about 0.5% of the country’s adult population. It provides regionally disaggregated data that enable community-level health analyses. The sampling frame was established by linking resident registration data from the Ministry of the Interior and Safety with housing data from the Ministry of Land, Infrastructure and Transport to construct a complete list of households.

The survey used a multistage, stratified probability sampling method. In the first stage, primary sampling units (PSUs) were selected based on administrative districts (tong/ban/ri) with probabilities proportional to the number of households by housing type. In the second stage, households were systematically sampled within each PSU. The overall sample size was predetermined by the national survey design, and approximately 10 households per PSU in urban areas and 8 households per PSU in rural areas were selected to account for regional variations in health indicators and sampling adequacy.

Trained interviewers visited selected households and collected data through computer-assisted personal interviewing using tablet-based electronic questionnaires between August and October each year.

Since 2010, the KCHS has been organized into 4-year survey cycles, with items categorized according to importance, applicability, and sampling size. Accordingly, certain items are assessed every 1, 2, or 4 years. The total number of questionnaire items ranged from approximately 140 to 160 per year (142 in 2020, 163 in 2021, 138 in 2022, and 145 in 2023), with minor revisions made annually to reflect changes in public health priorities and regional conditions.

This study used merged KCHS data from 2020 to 2023, including 922,048 participants, to examine factors influencing subjective health by regional aging levels. Exclusions included 613,405 individuals aged 65 years or younger, 42,731 with missing or outlier values, 6048 with duplicated identifiers, and 6471 from strata with only 1 PSU, resulting in a final sample of 179,571. Regions were classified by linking each respondent’s location to aging rate data from Statistics Korea at the Si-Gun-Gu (city-county-district) level. Areas with aging rates of 14% or lower were defined as “Aging area” (n=19,759), and those with aging rates of 20% or higher were defined as “Super-aged area” (n=159,782), based on commonly used demographic thresholds that have been applied in national-level classifications as well as regional analyses in Korea [20]. The study design and analytical procedures are summarized in Figure 1.

Subjective health was assessed using a 5-point Likert scale ranging from “Very bad” to “Very good” (1=Very bad, 2=Bad, 3=Fair, 4=Good, 5=Very good). Respondents were asked the question, “In general, how would you rate your health?.” To identify the factors influencing subjective health, items from the KCHS related to demographic characteristics, health behavior characteristics, health status characteristics, and health literacy were used.

Demographic variables included gender, age (years), household type, marital status, education, economic activity, and household income (monthly). Specifically, household type was categorized into “living alone,” “with a spouse or other (single-generation household),” “two generations (couple with unmarried children, single parent with unmarried children, or other 2-generation households),” and “three generations.” Marital status was divided into “yes, cohabitation,” “separated or divorced,” “bereavement,” and “no.” Education was classified into “below elementary (including no formal education or traditional education),” “elementary,” “middle school,” “high school,” and “college or higher (2-y or 3-y college, 4-y university, graduate school or higher).” Monthly household income was reported as a continuous variable.

Health behavioral characteristics included smoking (yes), drinking frequency, active physical activity (weekly, min), moderate physical activity (weekly, min), walking (weekly, min), subjective body shape, and weight control experience. Smoking was categorized into “yes” and “nonsmoker (smoked fewer than 5 packs or never smoked).” Drinking frequency was classified into “almost never (no drinking experience or none within the past year),” “less than once a month,” “1‐4 times a month (once or 2‐4 times per month),” and “more than 2 times a week (2‐3 times or more than 4 times per week).” Physical activity (active, moderate, and walking) was calculated as a continuous variable by multiplying the average daily minutes spent on each activity by the number of days it was performed per week, resulting in total weekly minutes.

Health-related characteristics included subjective stress, depression, hypertension, diabetes, subjective health, chewing discomfort, and subjective oral health. Specifically, chewing discomfort was classified into “uncomfortable (very uncomfortable, uncomfortable),” “neutral,” and “comfortable (slightly uncomfortable, not uncomfortable at all).”

Health literacy variables included nutrition label recognition, nutrition label reading, nutrition label utilization, unmet health care needs, and written health literacy. Nutrition label items were structured hierarchically to represent sequential levels of label literacy: recognition (awareness of labels on food packaging), reading (understanding the information displayed), and utilization (using label information in dietary decisions). Unmet health care needs were categorized into 2 groups: “yes” (excluding cases indicating no need for care) and “no.”

Statistical analyses were performed using RStudio (version 4.3.3; Posit, PBC) and IBM SPSS Statistics (version 25.0; IBM Corp.). A complex sample survey analysis method was applied considering the individual weights, strata, and cluster variables provided by the KCHS. These weights were calculated as the inverse of the selection probability and adjusted for response rates and population structure to represent the Korean adult population. Categorical variables were summarized as frequencies and percentages, and group differences were tested using the Rao-Scott chi-square test. Continuous variables were presented as mean (SD) and compared using design-based t tests. Statistical significance was defined as P<.05.

Propensity scores were calculated using 5 variables: gender, age, household type, education, and household income (monthly). Based on these propensity scores, data from aging and super-aged areas were matched using a 1:1 nearest-neighbor matching method. The covariate balance between the matched groups was confirmed by comparing standardized mean differences before and after matching. Postmatching weights were calculated using poststratification adjustment based on original KCHS sampling weights and applied in subsequent analyses [21].

Ordinal logistic regression analysis was conducted using subjective health as the dependent variable. The backward elimination method of stepwise selection was applied, starting with a full model containing all predictors and sequentially removing variables with the highest nonsignificant P values. The threshold for variable removal was set at P>.10. After each removal, the model was refitted and the P values were recalculated. This process was repeated until all the remaining variables reached statistical significance.

To identify significant regional differences in determinants of subjective health, subgroup analyses were performed separately for aging and super-aged areas using the same model structure. Odds ratios (OR) and 95% CI were estimated for each group. Subsequently, a z test was performed to compare the OR and SDs between the regions. The significance level used to identify trends in regional differences was set at 90% (P<.10).

This study was conducted using secondary data from KCHS provided by the KDCA. In accordance with Article 2(2) of the Enforcement Rule of the Bioethics and Safety Act [22], the study was exempt from institutional review board review because it did not involve direct human participation. The original data were collected under the Regional Public Health Act and its Enforcement Decree to support evidence-based community health programs. All participants in the KCHS provided written informed consent that included information about the purpose of data collection, the types of personal information gathered, and the period of data retention and use.

The dataset supplied by the KDCA was deidentified in accordance with the Personal Information Protection Act, using pseudonymization procedures in which identifiers such as sample point numbers and household numbers were replaced with numerical codes to prevent personal identification. Participants in the original survey received small incentives, including a gift voucher worth approximately KRW 10,000 (US $7) and additional daily necessities, provided by local community health centers. No identifiable images or materials that could reveal participants’ identities were used in this study.

Of the 159,782 respondents in super-aged areas, 19,759 were matched 1:1 to those in aging areas using nearest-neighbor matching. After excluding strata with a single PSU, the final matched dataset included 18,574 individuals. All covariates achieved standardized mean differences below 0.10 [23], including gender (<0.001), age (0.01), household type (0.009), education (<0.001), and household income (−0.01), confirming an adequate covariate balance between groups (Figure 2).

After matching, demographic characteristics were compared between aging and super-aged areas, including both the variables used for matching and additional variables not used in the matching process, such as age groups, marital status, and economic activity (Table 1). Comparisons of the same variables between aging and super-aged areas before propensity score matching are provided in Table S1 in Multimedia Appendix 1. Marital status and participation in economic activity differed significantly between regions (both P<.001).

Table 2 summarizes regional differences in health behaviors, health status, and health literacy. Among health behaviors, drinking frequency differed significantly between regions, with respondents in the super-aged areas showing more polarized drinking patterns and slightly higher rates of frequent drinking (P=.02). Smoking prevalence was similar across regions and was not significantly different. Regarding chronic conditions, the prevalence of depression was significantly lower in the super-aged areas (P=.003), whereas hypertension and diabetes did not differ notably between regions. Subjective health status and subjective oral health were poorer among respondents in the super-aged areas (P<.001 and P=.004, respectively). For health literacy, recognition of nutrition labels was higher in the super-aged areas (P=.002), but reading and utilization rates were lower (P<.001 and P=.03, respectively), suggesting difficulties in applying nutrition information in daily practice.

Ordinal logistic regression was first performed on the overall matched data to identify factors associated with subjective health. The results, presented in Table S2 in Multimedia Appendix 2, showed consistent patterns with prior evidence. Positive associations were observed for male gender, younger age, higher socioeconomic status, nonsmoking, and greater physical activity (moderate), whereas higher stress, depression, and chronic conditions were negatively associated with subjective health. Subsequently, subgroup analyses were conducted for the aging and super-aged areas using identical models.

Using matched datasets for aging and super-aged areas, ordinal logistic regression was conducted with identical independent variables and ORs were derived to compare the factors influencing subjective health between the 2 groups (Table 3). Most variables showed significant associations consistent with the overall dataset, and the direction of these associations was largely similar across both regions. However, some differences emerged. Smoking status was not significantly associated with subjective health in the super-aged area (P=.06). Regarding nutrition label variables, label recognition was significant only in the aging area (P=.04), whereas label reading was significant only in the super-aged area (P=.05).

To assess regional differences in the impact of variables influencing subjective health, z tests comparing ORs between regions showed that perceiving oneself as “very obese” had a significantly weaker negative association with subjective health in the super-aged area (OR 0.64) than in the aging area (OR 0.40; P=.005). Hypertension (P=.09) and diabetes (P=.10) also showed marginally significant trends, with both conditions presenting lower ORs in the super-aged area.

In this study, the proportion of respondents from super-aged areas who positively evaluated their subjective health (“very good” or “good”) was 28.9%. Furthermore, the proportion of positive subjective health evaluations in aging areas was 31.5%, higher than that in super-aged areas.

For drinking frequency, those consuming alcohol more frequently (“more than 2 times a week”) reported higher odds of positive health assessments compared to those drinking infrequently. Although the proportion of smokers was similar across regions, smoking status was not significantly associated with subjective health in super-aged areas.

Nutrition label recognition and reading exhibited distinct patterns across regions. In aging areas, nutrition label recognition was significantly associated with positive subjective health, whereas nutrition label reading was not. Conversely, in super-aged areas, despite higher recognition rates, recognition alone was not significantly related to subjective health, whereas nutrition label reading significantly influenced positive evaluations.

Hypertension and diabetes did not differ significantly between the 2 regions; both conditions tended to reduce the likelihood of positive subjective health evaluations more strongly in super-aged areas than in aging areas [24].

The regional disparities observed in this study are consistent with prior research showing that rural regions tend to report lower subjective health than urban areas [25,26], likely due to the higher proportion of rural communities within super-aged areas [27].

Demographic characteristics, health behaviors, health status, and unmet health care needs were significantly associated with subjective health, consistent with previous studies [11,15,28-30]. Similar patterns were also observed for chronic conditions such as hypertension and diabetes, consistent with prior findings [31].

Although research investigating the association between nutrition labels and subjective health is limited, Dumotier et al [32] reported that recognizing and using nutrition labels significantly influenced dietary habits and obesity. Similarly, Jang’s study [33] indicated that health literacy significantly improved subjective health status.

Although previous studies generally suggest that higher drinking frequency is negatively associated with health [14], some studies have found that moderate drinking frequency is positively linked to subjective health [34], possibly due to the social interactions accompanying drinking [35]. The lack of association between smoking and subjective health may indicate that factors beyond individual smoking behavior, such as government tobacco control policies or the socioeconomic characteristics of smokers in different regions, could have influenced self-reported health, potentially contributing to regional differences [36,37].

In Korea, approximately 84% of older adults have at least one chronic condition, many of which require dietary modifications, making it particularly important for this population to utilize information on food labels [38]. However, a high proportion of older adults report difficulties in reading and interpreting food labels [39]. Such challenges may stem from age-related declines in numeracy and attention, which can reduce the ability to process and apply nutritional information effectively [40].

Nutrition knowledge plays a key role as a predictor of both comprehension and utilization of nutrition labels [41]. In Korea, nutrition education for older adults is mainly provided through public health centers and senior welfare centers, yet access to such programs remains limited in super-aged areas. These areas are disproportionately rural, where older residents often face longer travel distances to health care facilities and have fewer community-based resources, leading to reduced participation in preventive programs and health education initiatives.

This study has several limitations. First, although the KCHS is a nationally representative dataset, its cross-sectional design allows for the assessment of associations but not causality between subjective health and its related factors. Longitudinal studies are needed to establish causal relationships and to address potential reverse causation. Second, regional classification based solely on aging rates may not fully reflect the complex social and environmental characteristics of each community. However, by applying this classification at the district level rather than using broad administrative boundaries, the study was able to capture substantial variations in aging rates and subjective health. This finer granularity likely improved the sensitivity of the analysis and better reflected Korea’s ongoing demographic transitions.

Improving subjective health is essential for increasing healthy life expectancy and enhancing quality of life in South Korea, a nation experiencing rapid population aging. It has also been designated as a specific indicator within Korea’s 5th National Health Plan (HP2030), highlighting it as a national priority requiring focused intervention [42]. The lower proportion of older adults reporting positive subjective health in super-aged areas, compared with both the 2023 KCHS national average (47.6%) and the HP2030 target (34.7%), underscores the need for targeted public health interventions.

Given that 77% of households with older adults in Korea report that older adults are the primary purchasers of food [43], the ability to accurately read and interpret nutrition labels is particularly important for this population. In Korea, front-of-package (FOP) nutrition labeling has not yet been implemented, whereas countries such as Canada have adopted standardized FOP symbols to help consumers quickly identify foods high in sodium, sugar, or saturated fat [44]. Such government-endorsed labeling has been shown to improve the accuracy and speed of consumers’ food choices and is particularly beneficial for populations with lower health literacy, including older adults [45]. Introducing a similar labeling policy in Korea could therefore promote informed food selection and facilitate the effective use of nutrition labels in daily dietary decisions.

Complementary to such policy measures, nutrition education remains essential, particularly in super-aged areas where accessibility to educational programs remains limited. Previous research has shown that when older adults focus on specific nutrients, their accuracy in interpreting nutrition labels can be comparable to that of younger adults [40]. This suggests that targeted and nutrient-specific education can effectively enhance comprehension among older adults. Notably, in this study, hypertension and diabetes, although similarly prevalent across regions, showed a stronger negative association with subjective health in super-aged areas than in aging areas [24], indicating that nutrition label education emphasizing disease-related nutrients may be particularly beneficial in super-aged areas, as supported by previous evidence that tailored nutrition education significantly improved dietary knowledge and behaviors among the elderly [46].

When combined with FOP labeling and community-based education through local health centers, such tailored nutrition programs could strengthen dietary awareness, promote active label use, and help reduce health disparities among older adults [47].

This study revealed regional differences in factors influencing subjective health among older adults in Korea. In super-aged areas, reading and understanding nutrition labels, rather than merely recognizing them, were more strongly associated with positive subjective health, while the negative effects of hypertension and diabetes were also more pronounced. These findings highlight the need for region-specific public health strategies that include tailored nutrition education focusing on disease-related nutrients to improve subjective health among older adults.