, Introduction
Chronic kidney disease (CKD) is a global health problem strongly associated with increased cardiovascular risk, progression to end-stage kidney disease, and significant healthcare expenditures [1–3]. The Kidney Disease: Improving Global Outcomes (KDIGO) guidelines classify CKD based on its cause, the glomerular filtration rate (GFR) category (G1–G5), and albuminuria category (A1–A3). These guidelines also recommend the frequency of kidney function monitoring per year based on estimated GFR (eGFR) and albuminuria [4]. However, the rationale for these monitoring frequencies is unclear, as much of the guidance is based on expert opinion rather than empirical evidence [5]. To date, only a few observational studies have investigated the optimal visit frequency for patients with CKD [6].
In clinical practice, the optimal timing and frequency of follow-up visits, not only for CKD but also for other chronic conditions, are poorly defined. Many clinicians determine monitoring intervals based on personal habits, institutional policies, or logistical constraints rather than evidence-based guidelines [7]. Specifically, in CKD management, although patients with more comorbidities make follow-up visits more frequently, laboratory test results appear to have minimal impact on the follow-up interval, except in cases involving erythropoietin use [8,9]. This discrepancy suggests that the recommended interval may not be consistently followed, potentially compromising the quality of care and straining healthcare resources.
Assessing the extent to which clinical practice aligns with KDIGO recommendations and identifying factors that influence the follow-up interval are important to inform tailored, evidence-based, and resource-efficient care strategies for CKD management. Therefore, the objective of the present study was to evaluate adherence to the KDIGO-recommended outpatient follow-up interval among Korean patients with CKD in a tertiary care center and to identify the factors that influence these intervals.
Methods
Study design and participants
This retrospective observational study used electronic medical record (EMR) data from Severance Hospital, a tertiary care center in the Republic of Korea. The study population consisted of patients diagnosed with CKD who visited the Department of Nephrology between November 1, 2005, and December 31, 2021. CKD was defined as a decline in kidney function persisting for more than 3 months, with an eGFR less than 60 mL/min/1.73 m2, corresponding to stage G3a or higher [4].
Patients were included if they had made at least two visits to the Department of Nephrology between November 1, 2005, and December 31, 2021. Patients were excluded if their visits were unrelated to kidney disease or if they occurred after surgery for advanced kidney dysfunction. Additionally, patients with visits whose CKD stage could not be determined or who were not diagnosed with CKD were excluded. Patients with a gap of 365 days or more between their CKD diagnosis and the next nephrology visit as well as those with fewer than two nephrology visits after their CKD diagnosis were also excluded. Furthermore, patients without address information were excluded because the distance to the hospital was considered a potential factor influencing the outpatient visit interval in this study. The final cohort consisted of 13,621 patients who met all the eligibility criteria (Fig. 1).
This study complied with the Declaration of Helsinki and was approved by the Yonsei University Health System (No. 4-2021-1532). The requirement for written informed consent was waived because the study used de-identified administrative data.
Potential factors influencing the outpatient visit interval
To investigate the clinical and nonclinical factors influencing the outpatient visit interval, we examined patient demographics, healthcare utilization patterns, clinical markers, and physician-specific variability.
Demographic factors, including age and sex, were analyzed to assess their influence on visit patterns. The driving distance to the hospital was calculated using Naver Directions Application Programming Interface (https://www.ncloud.com/product/applicationService/maps). If only city-level address data were available, the median distance of the patients from the same city was used as an estimate. The CKD stage was assessed using the eGFR and albuminuria test results from the hospital where the study was conducted. If the test results were unavailable for a specific visit date, the most recent results within the preceding 2 weeks were used. Comorbidity data were extracted from the study hospital’s EMR and included data on hypertension, diabetes mellitus (DM), dyslipidemia, cancer, myocardial infarction, ischemic stroke, hemorrhagic stroke, peripheral arterial disease, chronic obstructive pulmonary disease, hyperthyroidism, atrial fibrillation, heart failure, and dementia (Supplementary Table 1, available online). Healthcare utilization was assessed based on the number of non-nephrology outpatient visits made to the study hospital in the preceding 12 months. This variable was included to account for the influence of comorbid conditions and cross-specialty management on follow-up patterns, given that patients with CKD in tertiary hospitals are often co-managed by other specialties such as cardiology, neurology, or endocrinology. Physician-related variability was accounted for by defining the primary physician as the nephrologist responsible for most patient visits. In cases in which the number of visits was the same for multiple nephrologists, the nephrologist who initially treated the patient was designated as the primary physician.
Outcome variable
The primary outcome variable in this study was the interval between nephrology outpatient visits, which was defined as the number of days between two consecutive nephrology outpatient visits.
Statistical analysis
When presenting summary statistics, the normality of continuous variables was assessed using the Shapiro-Wilk test. Variables that followed a normal distribution were presented as means ± standard deviations, whereas those that did not follow a normal distribution were presented as medians with interquartile ranges (IQRs). For the histogram depicting the outpatient visit interval across CKD stages, the median, IQR, and adherence rates to the KDIGO guideline-recommended follow-up interval were calculated and annotated to facilitate visual interpretation.
A linear regression model was used to evaluate the potential factors influencing visit interval and was chosen for its simplicity, interpretability, and ability to quantify the magnitude and direction of the variable effects. As the visit interval data were not normally distributed, a logarithmic transformation was applied to meet the linear regression assumptions. Shapley Additive exPlanations (SHAP) values were derived from a trained linear regression model to assess the contribution of each factor to the outpatient visit interval [10]. A Kruskal-Wallis test was conducted to investigate potential significant differences in the visit interval across CKD stages for each primary physician. All data were analyzed using STATA (version 17.0, StataCorp LLC) and Python (version 3.11.2, Python Software Foundation). Statistical significance was set at p < 0.05.
Results
Baseline characteristics of study participants
The baseline characteristics of 13,621 patients with CKD are shown in Table 1. The baseline date was defined as the first nephrology outpatient visit date after the patient’s kidney function had remained in stage G3a or worse (eGFR <60 mL/min/1.73 m2) for at least 90 consecutive days. Most patients had baseline kidney function at stage G3a or worse. Due to transient improvements after meeting the inclusion criteria, a small number of patients were classified as G1 or G2 at baseline.
The median age was 68 years (IQR, 57–76 years), and 60.6% of the patients were male. Patients tended to be older as the CKD stages progressed; however, this trend was reversed at stage G5, where the median age decreased. The median number of non-nephrology outpatient visits in the past 12 months was 5 (IQR, 2–11). Hypertension was the most common comorbidity, affecting 75.8% of the cohort, with the highest prevalence observed among patients in stage G3b (78.4%). Other prevalent conditions included DM (44.4%) and dyslipidemia (40.3%).
The median number of visit intervals from the baseline date to the end of the study period was 13 (IQR, 6–15), and a higher CKD stage was associated with more frequent visits. Similarly, the mean interval between nephrology visits decreased from a median of 140 days for patients in stage G2 to 46 days for those in stage G5, reflecting increased clinical monitoring in advanced CKD.
Distribution of the outpatient visit intervals across the chronic kidney disease stages
Fig. 2 illustrates the distribution of the outpatient visit intervals across the CKD stages, with the median visit intervals and IQRs shown in the first row of the text within each histogram. The median visit interval progressively decreased as the CKD stage and severity increased, reflecting the increased need for clinical monitoring. Specifically, the median visit interval decreased from 115 days in patients in stage G3a to 84 days in patients in stage G4 and 42 days in patients in stage G5. This trend was consistent across all albuminuria categories (A1, A2, and A3), with the shortest median interval observed in G5–A3 (37 days). The IQR also narrowed with CKD progression; for example, the IQR for G3a–A1 was 91 to 181 days, compared with 28 to 63 days for G5–A3.
The red lines in Fig. 2 represent the KDIGO-recommended visit intervals, whereas the percentages in the second row of the text within each histogram represent the proportion of visits that adhered to these guidelines. Adherence was the highest in earlier CKD stages but declined in advanced stages, particularly within the A3 category. Notably, in the most advanced stage (G5–A3), 90.0% of the visits adhered to the recommended visit interval, in contrast to the general trend of decreasing adherence observed for other CKD stages.
Analysis of potential factors influencing the outpatient visit interval
Table 2 and Supplementary Table 2 (available online) show the results of the linear regression analysis examining the potential factors influencing the outpatient visit interval. Although our study focused on patients in stage G3a or higher for at least 3 months, some visits were made when they were in stage G1 or G2, because of temporary kidney function improvement after CKD diagnosis.
The analysis showed that the visit interval decreased progressively as CKD severity increased. As the outcome variable was log-transformed, the regression coefficients represented the approximate percentage change in the visit interval. For example, the transition from G1–A1 to G5–A3 was associated with an approximately 66% reduction in the visit interval (p < 0.001).
The influence of demographic factors was observed, suggesting their potential role in shaping the outpatient visit interval. For example, each additional year of age was associated with a 0.6% increase in the visit interval (p < 0.001). Distance to the hospital had a modest but significant effect, with a slight increase in the visit interval for each additional kilometer (p < 0.001). Among the comorbidities, hypertension, dyslipidemia, and cancer were associated with longer visit intervals, with increases of 3.7%, 7.9%, and 4.3%, respectively (p < 0.001). Conversely, DM, dementia, and hemorrhagic stroke were associated with 1.8%, 4.9%, and 4.8% shorter intervals, respectively (p < 0.001). Physician assignment also influenced the visit interval. Compared with patients treated by Physician A, those treated by Physician L had 85.9% shorter visit intervals (p < 0.001), whereas those treated by Physician J had 13.9% longer intervals (p < 0.001).
The SHAP analysis identified the CKD stage as the most influential predictor of the visit interval (Fig. 3). In addition, nonclinical factors, such as age and number of outpatient visits to other departments in the previous 12 months, were identified as the next most significant contributors, followed by physician assignment and the presence of dyslipidemia. Supplementary Fig. 1 (available online) shows the disaggregated SHAP results derived from one-hot encoding of the categorical variables. Within the CKD stage category, subcategories such as G5–A3, G4–A3, and G5–A1 showed elevated SHAP values, highlighting their strong influence on shorter visit intervals. Similarly, contributions varied across physicians, revealing physician-specific patterns in outpatient scheduling.
Physician-induced variation in the outpatient visit interval
To evaluate physician-specific variations in the outpatient visit interval, we performed subgroup analyses stratifying the patients by their primary physician. Table 3 shows the median visit intervals across CKD stages for 13 nephrologists. The Kruskal-Wallis test revealed significant differences in the median visit intervals at all CKD stages. Physician A consistently adhered to shorter follow-up intervals, with a median of 91 days for patients in stage G3a–A1 and 53 days for patients in stage G5–A3. Patients treated by Physician F had the shortest visit intervals overall, with a median of 7 days for patients in stage G3a–A1 and 19 days for patients in stage G5–A3.
The physician-specific linear regression results (Supplementary Table 3, available online) show the variability in how different factors were associated with the visit interval across physicians. For example, patients in advanced CKD stages (G4 and G5) treated by Physicians A, B, and C had shorter visit intervals. Conversely, patients at advanced stages treated by Physicians I and K had longer visit intervals, suggesting different clinical practices or interpretations of patient needs (p < 0.001). The SHAP analysis provided further insight into the factors driving these differences. Fig. 4 presents the SHAP summary plots for the four selected physicians (A, B, I, and J), where the CKD stage was shown to be the most influential factor across all four physicians. For Physicians A and B, age and prior outpatient visits were highly influential, with Physician A ranking age as the second most important factor and Physician B prioritizing prior outpatient visits. For Physicians I and J, comorbidities, such as dyslipidemia (Physician I) and hypertension (Physician J), were the second most important factors, suggesting a greater focus on the presence of comorbidities. The SHAP results for the remaining nine physicians are shown in Supplementary Fig. 2 (available online).
Discussion
This study examined the potential factors influencing the outpatient visit interval in patients with CKD, with a particular focus on adherence to the KDIGO guidelines and the role of patient-specific and clinical characteristics. Understanding these patterns is essential for optimizing follow-up care, ensuring timely monitoring, and improving resource allocation in CKD management.
Adherence to KDIGO-recommended follow-up intervals was relatively high in early CKD stages, but declined in the later stages, with a notable rebound in the most advanced stage (G5–A3), where 90.0% of visits aligned with guidelines. This U-shaped pattern suggests that clinicians tend to follow guidelines more closely at the initial stages of the disease, transition to a more flexible practice in the middle stages, and subsequently resume strict adherence as patients approach highest risk.
Our analysis identified CKD stage as the primary factor influencing the visit interval, with more advanced stages associated with increased visit frequency. This finding is consistent with the findings of previous studies and the KDIGO guidelines, which recommend closer monitoring of patients with worsening kidney function [4]. Previous research has also highlighted that CKD progression requires intensive clinical management because of its association with adverse health outcomes, including cardiovascular disease and mortality [11,12]. In addition to cardiovascular diseases, CKD is associated with an increased risk of chronic infections and cancer, requiring a more integrated approach in public health policies [13,14]. Given the high prevalence of kidney dysfunction in patients with certain malignancies and the nephrotoxicity associated with many cancer treatments, regular monitoring of kidney function is critical in oncology care [15].
Demographic factors, such as age and sex, were also associated with the visit interval. Prior studies have shown that these variables significantly impact healthcare utilization, with older patients and women typically using outpatient services more frequently, owing to greater healthcare needs [16]. In this study, older patients had longer visit intervals, which may reflect a perception of clinical stability or barriers, such as mobility limitations. In addition, the slower progression of CKD in older patients may have contributed to the longer visit intervals. Studies have shown that CKD progresses more slowly in older populations, possibly because of less structural kidney damage, lower levels of proteinuria, and natural aging of the kidneys [17]. For instance, one study reported a median annual eGFR decline of only 0.8 mL/min/1.73 m2 in patients aged 80 years or older, with only 8.1% progressing to end-stage kidney disease during the study period, supporting the notion that CKD progression is typically slower in older adults [18]. Similarly, sex-based differences in the visit interval may be related to health-seeking behaviors, as previous research suggests that women use healthcare services more frequently than men do [19].
Healthcare accessibility was another important factor, with a greater distance to the hospital associated with a longer visit interval. This finding is consistent with previous studies showing that geographic barriers can reduce specialist consultations and delay the need for medical care [20]. However, notably, in this study, the distance to the hospital reflected geographic accessibility to a specific tertiary care center rather than the overall accessibility to the healthcare system. Healthcare utilization patterns were analyzed based on the frequency of outpatient visits to non-nephrology departments. The results showed that patients with higher engagement in healthcare services at the study hospital had shorter nephrology outpatient visit intervals, suggesting that prior interaction with healthcare services within the same institution was associated with more frequent nephrology follow-ups.
Physician-specific patterns also emerged, suggesting that the visit interval is influenced not only by patient characteristics but also by provider practices. Previous research has suggested that physician preferences, care coordination, and workload may contribute to variability in care delivery [21]. This variability underscores the need for the consistent application of clinical guidelines across providers to ensure standardized follow-up care.
The SHAP analysis provided an interpretable framework for quantifying the relative contribution of multiple factors to the outpatient visit interval. Among all variables, CKD stage consistently showed the highest SHAP values, confirming its dominant influence on follow-up intensity. Nonclinical factors such as age, prior healthcare utilization, and distance to the hospital also had meaningful contributions, indicating that visit scheduling reflects both disease severity and patient engagement with the healthcare system. Physician-specific SHAP profiles revealed heterogeneity in clinical decision-making, suggesting that individual practice styles or institutional routines may influence visit intervals beyond objective clinical indicators. However, it is worth noting that SHAP values capture the magnitude of each predictor’s contribution to model predictions, whereas statistical significance reflects the uncertainty around an effect estimate [10,22,23]. Thus, a variable may appear highly influential in the SHAP plot even if its coefficient is not statistically significant. This distinction highlights that SHAP should be interpreted as explaining predictive impact rather than inferential strength.
Taken together, our findings indicate that although the KDIGO guidelines provide a valuable framework for CKD management, real-world practice is shaped by patient demographics, comorbidities, healthcare accessibility, and physician-specific decisions. These results highlight the need for follow-up recommendations that are more adaptable across CKD stages and that integrate both clinical and nonclinical factors, such as age, comorbidity burden, and access to care. A more individualized approach may improve adherence, optimize healthcare resource use, and better align monitoring intensity with patient needs.
This study has several limitations. First, the results reflect the specific healthcare environment in South Korea, which limits their generalizability to other global healthcare settings. Second, the data were derived from a single hospital’s EMR system and may not capture the full spectrum of patient health records, including interactions with external healthcare providers. Third, patients’ comorbidities were identified based on International Classification of Diseases codes recorded within the same hospital. Therefore, comorbid conditions that could have affected outpatient visit intervals may have been overlooked if they were not documented in the hospital records. Fourth, the log-linear regression model assumed linear relationships between predictors and the logarithm of visit interval, which may have missed complex nonlinear dynamics. Fifth, although we included the number of non-nephrology outpatient visits as a healthcare utilization variable to account for cross-specialty care, our analysis was restricted to nephrology outpatient visits. We did not evaluate the proportion of nephrology visits relative to total hospital visits or designate the primary managing specialty, which may have influenced the observed follow-up intervals. Finally, we did not assess patient outcomes associated with adherence or non-adherence to the KDIGO recommendations. Future research should focus on linking visit interval patterns with outcomes such as CKD progression, hospitalization, and mortality, which would provide stronger evidence for whether tailoring guideline-based follow-up intervals by patient- and system-level factors can improve efficiency and patient prognosis.
In conclusion, CKD stage was the most important determinant of outpatient follow-up intervals in patients with CKD. However, adherence to KDIGO recommendations varied across stages, and visit schedules were further influenced by patient characteristics, healthcare utilization, accessibility, and physician practice patterns. These findings emphasize the need for tailored follow-up strategies that account for disease severity as well as the broader clinical and healthcare context in the management of patients with CKD.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement table 1
Print
