, Introduction
Chronic kidney disease (CKD) remains a major global public health concern, and identifying modifiable risk factors for CKD progression is essential for prevention and early intervention [1]. Although traditional clinical predictors such as diabetes mellitus and hypertension are well established, increasing attention has been directed toward structural and environmental factors, including geographic residence, that may influence long-term renal outcomes [1–3].
In South Korea, the population distribution has become increasingly concentrated in urban areas, whereas rural regions are facing continuous depopulation and demographic aging. Despite occupying a substantial portion of the national territory, rural areas are often characterized by lower healthcare service density, reduced access to specialists, and longer travel times to medical facilities [4–6]. However, the potential link between these structural disparities and decline in renal function has not been fully elucidated in population-based studies.
In this study, we investigated whether place of residence is associated with the risk of a rapid decline in the estimated glomerular filtration rate (eGFR) using cohort data that included both urban and rural populations in Korea.
Methods
Ethics considerations
This study was approved by the Institutional Review Board (IRB) of the Korea University Guro Hospital (No. 2020GR0151) and complied with the principles of the Declaration of Helsinki. The requirement for informed consent was waived by the IRB, given the use of publicly available, de-identified data.
Study setting and participants
This study was based on urban and rural cohorts from the Korean Genome and Epidemiology Study (KoGES). Urban participants were health screening examinees recruited from medical centers located in major cities and medium-sized urban areas, including Seoul, Busan, Incheon, Daegu, Gwangju, Ulsan, Anyang, Goyang, Seongnam, Chuncheon, Cheonan, Gwangju (Jeonnam), Hwasun, and Changwon. Rural participants were residents of farming communities recruited from predefined rural regions, including Yangpyeong (Gyeonggi-do), Goryeong (Gyeongsangbuk-do), Namwon (Jeollanam-do), Wonju and Pyeongchang (Gangwon-do), and Ganghwa (Incheon). Participants with at least one follow-up eGFR measurement were included. Exclusion criteria were as follows: missing baseline eGFR values; baseline eGFR <60 mL/min/1.73 m²; and proteinuria ≥1+ on dipstick analysis. After exclusion, 60,487 and 14,776 participants from the urban and rural cohorts, respectively, were included in the analysis (Fig. 1). For comparison, a 1:1 propensity score matching was performed based on age, sex, baseline eGFR, and dipstick proteinuria.
To calculate the annual eGFR decline rate, we used the difference between baseline and the earliest available follow-up eGFR measurement divided by the time interval in years. In the urban cohort, eGFR measurements at baseline (2004–2013) and first follow-up (2012–2016) were used. In the rural cohort, eGFR values from baseline (2005–2011) and the first available follow-up among up to four follow-up visits (2007–2016) were used.
Data collection and processing
Baseline demographic, clinical, and laboratory data were collected from all the participants. The variables included age, sex, comorbidities (hypertension, diabetes mellitus, cardiovascular disease, and dyslipidemia), nutritional status based on food frequency questionnaires, and laboratory parameters such as albumin level. Variables with >10% missing values (e.g., cancer history, income, smoking status, calcium, high-sensitivity C-reactive protein, and glycated hemoglobin) were excluded from the analysis. For the remaining variables with missing values below 10%, multiple imputation using chained equations was performed to reduce potential bias and improve robustness. This process generated five imputed datasets, each of which was subsequently used in downstream propensity score matching and analysis.
Statistical analysis
Baseline characteristics are summarized as number (%) for categorical variables and as mean ± standard deviation for continuous variables. Comparisons between the baseline characteristics of the groups were conducted using the t-test for continuous variables and the chi-squared test for categorical variables. To ensure comparability between cohorts, 1:1 propensity score matching was performed using the matchit function from the R MatchIt package, based on age, sex, baseline eGFR, and dipstick proteinuria, with a caliper of 0.1. This matching process was conducted separately within each of the five imputed datasets generated during data preprocessing. To evaluate the association between geographic residency and the risk of rapid eGFR progression, logistic regression models were applied across the five matched datasets. Odds ratios (ORs) with 95% confidence intervals (CIs) were pooled for the unadjusted, backward selection, and fully adjusted models using Rubin’s rules. Mediation analysis was performed using structural equation modeling with the lavaan package in R. Given the inclusion of both categorical (diabetes mellitus, hypertension, cardiovascular disease, education levels) and continuous mediators (serum albumin, eGFR, dietary protein and sodium intake), the analysis employed a weighted least squares mean and variance adjusted estimator, which is appropriate for models involving categorical dependent variables. Binary exposure (rural vs. urban residence) and binary outcome (rapid progression) were modeled with a logit link. The indirect and direct effects were estimated separately for each imputed and propensity-matched dataset, and the results were pooled using Rubin’s rules. All statistical analyses were performed using R software version 4.4.3 (R Foundation for Statistical Computing), and a p-value <0.05 was considered statistically significant.
Results
Baseline characteristics of age, sex, and estimated glomerular filtration rate-matched urban and rural cohort
A total of 60,487 and 14,776 eligible participants were identified in the urban and rural cohorts, respectively. The baseline characteristics of all the participants before matching are shown in Supplementary Table 1 (available online). Variables with >10% missing data were excluded, and the remaining variables were imputed using multiple imputation by chained equations, resulting in five imputed datasets. In each dataset, 1:1 propensity score matching was conducted based on age, sex, baseline eGFR, and dipstick proteinuria. As a result, 14,336 matched participants remained in each of the urban and rural cohorts.
All five matched datasets showed consistent distributions. Table 1 presents the baseline characteristics from one representative dataset. After matching, the mean eGFR was well-balanced between the two groups (81.1 ± 10.7 mL/min/1.73 m² vs. 81.6 ± 10.3 mL/min/1.73 m²). However, small residual differences in age (57.3 ± 7.9 years vs. 58.2 ± 9.1 years), male sex (44.8% vs. 38.9%), and trace proteinuria (7.5% vs. 8.3%) remained.
The urban cohort had a higher prevalence of hypertension and dyslipidemia compared to the rural cohort, whereas the rural cohort had a higher prevalence of diabetes mellitus and cardiovascular events. According to the food frequency questionnaire, a higher proportion of the urban cohort followed a low-sodium diet (Na <2 g/day), while in the rural cohort, more than 50% of participants reported adhering to a low-protein diet.
Prevalence and risk of rapid estimated glomerular filtration rate decline according to geographic residency
During a median follow-up of 3.8 years (interquartile range, 2.2–4.2 years), the prevalence of rapid progressors—defined as an eGFR decline exceeding 5 mL/min/1.73 m² per year—was 279 (1.9%) in the urban cohort and 1,305 (9.1%) in the rural cohort, respectively. Pooled logistic regression analysis across the five imputed and matched datasets was conducted to assess the effect of residential area on rapid progression. In the unadjusted, backward selection, and fully adjusted models, rural residency was associated with an odds ratio (OR) of 5.0 (95% CI, 4.42–5.76), 8.0 (95% CI, 7.17–9.00), and 5.4 (95% CI, 4.61–6.37) for rapid progression compared to urban residency, respectively. Fig. 2 presents the significant risk factors for rapid progression. Rural residency exhibited the highest OR, even when compared to diabetes mellitus or cardiovascular events.
Mediation analysis for evaluating direct and indirect effects of residency area on rapid estimated glomerular filtration rate decline
Regarding the significant increase in risk for rapid eGFR decline by residential area, we hypothesized that this association might be partially mediated by differences in clinical or socioeconomic factors rather than entirely attributable to a direct effect of residency itself. To explore this, we performed a mediation analysis using serum albumin level, educational attainment, dietary protein intake, diabetes mellitus, hypertension, and cardiovascular disease as potential mediators of the relationship between rural residency and rapid progression (Fig. 3).
The pooled OR for the total indirect effect was 1.12 (95% CI, 1.07–1.16; p < 0.001), indicating a significant contribution from the mediators. Among them, lower serum albumin showed the strongest mediating effect (OR, 1.07; 95% CI, 1.06–1.09; p < 0.001), followed by lower educational attainment (OR, 1.04; 95% CI, 1.01–1.06; p = 0.003) and low dietary protein intake (OR, 1.01; 95% CI, 1.01–1.02; p = 0.001). Nonetheless, a significant direct effect of rural residency remained after adjusting for these mediators, suggesting that additional, unmeasured factors such as environmental exposures, healthcare accessibility, or behavioral patterns may also play a role in the observed geographic disparity.
Discussion
In this study, we observed substantial geographic disparities in the risk of rapid eGFR decline among urban, rural, and emigrant populations. Specifically, individuals residing in rural areas of Korea exhibited a significantly higher risk of rapid kidney function decline than their urban counterparts, even after adjusting for age, sex, baseline kidney function, and proteinuria. These findings suggest that rural residency may independently contribute to accelerated renal deterioration, potentially due to unmeasured confounders, including reduced healthcare access, environmental stressors, or socioeconomic disadvantages.
According to Statistics Korea, approximately 9.2% of the national population resides in non-urban areas [9]. As urbanization continues and regional inequalities intensify, rural populations are increasingly affected by healthcare disparities, including limited service availability and geographic isolation [10,11]. Rural areas in Korea account for only 13% of healthcare institutions nationwide, and the number of private healthcare facilities per million people is 70% of that in urban areas [12]. Additionally, the spatial dispersion of rural residents results in longer travel times to medical facilities [12]. Such systemic limitations are compounded by unfavorable health behaviors and lower socioeconomic status among rural populations, which have been associated with a higher burden of chronic diseases, including CKD [13–18].
Our findings are consistent with those of previous international studies highlighting the challenges faced by patients with CKD in rural regions, including financial and logistical barriers to accessing nephrology care, increased caregiving burden, and social isolation. A global review of 18 studies emphasized the disproportionate difficulties encountered by rural residents in accessing CKD care services [19]. Delayed referral to nephrologists and late CKD diagnosis, often due to an asymptomatic disease course, can result in suboptimal outcomes [20]. Evidence indicates that early nephrology referral is critical for slowing CKD progression [21,22], and rural residents may have less access to specialty care, as demonstrated in a previous study [23].
Our mediation analysis provides insight into the mechanisms linking rural residency to rapid kidney function decline. While traditional risk factors such as diabetes mellitus, hypertension, and cardiovascular disease were included, nutritional status and educational attainment emerged as particularly influential. Specifically, lower serum albumin and reduced dietary protein intake highlighted the contribution of malnutrition and poor diet quality—likely reflecting broader inequities in nutritional access and health literacy. Likewise, lower educational levels may signal limited capacity for disease self-management.
These results suggest that geographic disparities in kidney health are driven not only by medical conditions but also by socioeconomic and behavioral determinants, shaped by systemic barriers in healthcare access and education. Nonetheless, a significant direct effect of rural residency persisted, indicating that additional unmeasured factors such as environmental exposures, healthcare availability, or cultural barriers may also play an important role.
Addressing these disparities requires a comprehensive public health approach. Evidence-based strategies such as CKD screening [24] in rural areas and better referral systems [21] can support early diagnosis and access to specialist care. In addition, community programs focused on nutrition and health education may help reduce the impact of poor diet and limited health knowledge. Together, these approaches are key to lowering the burden of CKD and improving kidney health in underserved regions.
This study has several strengths, including the use of a large, well-matched cohort and robust statistical methods to address confounding factors. Importantly, the analysis incorporated a wide range of clinical, dietary, familial, and socioeconomic variables, allowing for comprehensive adjustment for both medical and contextual risk factors. However, some limitations must be acknowledged. First, although matching was performed on key variables including age, sex, eGFR, and proteinuria, some residual imbalance remained, which may have affected the results. Unmeasured confounding factors may also persist due to limitations in the dataset. Second, although rapid eGFR decline was defined according to the 2012 KDIGO guideline as an annual decrease >5 mL/min/1.73 m², in a generally healthy population this may in some cases reflect acute, transient fluctuations or measurement variability rather than true CKD progression. This potential misclassification could attenuate the specificity of the outcome definition and should be considered when interpreting the findings. Third, the mean follow-up duration differed between groups (urban, 4.7 years; rural, 2.8 years), which may have introduced bias in estimating annual eGFR decline. Finally, a follow-up duration of 3.8 years may not be sufficient to fully characterize the long-term trajectory of kidney function, particularly in individuals with slowly progressive disease.
In conclusion, our findings highlighted the critical role of geographic residence in CKD risk stratification. In particular, rural Korean populations appear to be at elevated risk of rapid kidney function decline and may benefit from targeted interventions. Future research should aim to identify specific environmental and healthcare-related contributors to this increased risk and to develop effective preventive strategies tailored to rural settings.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement table 1
Print
