Background: Diabetic nephropathy requires periodic monitoring, dietary modification, and early intervention to prevent the disease severity within limited resource settings. To emphasize the importance of continuous care for chronic diseases, various studies have focused on the association between continuity of care (COC) and common adverse outcomes. However, studies aimed at understanding the effect of COC on the incidence of chronic diseases, such as end-stage renal disease (ESRD), are few. The aim of this study was to determine whether there is an association between COC and the incidence of ESRD among patients with diabetic nephropathy. Moreover, we identified individual- and hospital-level factors associated with the incidence of ESRD among diabetic nephropathy patients. Methods: We conducted a retrospective cohort study using the administrative National Health Insurance claims data from 2005 to 2012 in the Republic of Korea. The dependent variable, a binary variable, was the incidence of ESRD due to diabetic renal complication. In addition, using the COC index as a binary variable with a cutoff point of 0.75, we divided patients into a ‘Good COC group’ (COC index≥0.75) and a ‘Bad COC group’ (COC index< 0.75). The survival analysis was performed using the Cox proportional hazards models. Results: Among 3565 diabetic renal complication patients, ESRD occurred among 83 diabetes mellitus patients (2.3%). Nephropathy patients with lower COC level (< 0.75) had 1.99 times higher risk of ESRD incidence (95% confidence interval [CI]:1.27–3.12). In addition, the lowest income level patients had higher hazard ratio (HR) of ESRD than the highest income level patients (HR: 1.69 95% CI: 0.95–2.98), while patients with disabilities had 2.70 higher HR of ESRD than patients without disabilities (95% CI: 0.64–43). Conclusions: Among patients with diabetic renal complication, higher continuity of care was associated with lower risk of ESRD. It is therefore recommended that continuous follow-up be encouraged to prevent ESRD among diabetic renal complication patients. Moreover, disparities in health outcomes between socially vulnerable groups including patients with disabilities and those in the lowest income level should be addressed. Keywords: Diabetic renal complication, Continuity of care, ESRD, Disparities in health outcomes * Correspondence: firstname.lastname@example.org Yun Jung Jang and Yoon Soo Choy contributed equally to this work. Institute of Health Services Research, Yonsei University, Seoul, Republic of Korea Department of Preventive Medicine & Institute of Health Services Research, Yonsei University College of Medicine, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-752, Republic of Korea Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Jang et al. BMC Nephrology (2018) 19:127 Page 2 of 12 Background Methods Nephropathy causes approximately 48.0% of end-stage Data renal disease (ESRD) cases, and accounts for an annual A retrospective cohort study was conducted using medical cost of more than 1 billion dollars in the United population-based data collated from the Korean NHI States . The major complications of nephropathy, database from 2005 to 2013. Data was obtained from the which dramatically increase medical costs in the Repub- Korean National Health Insurance Service-National lic of Korea, include kidney transplants, dialysis, percu- Sample Cohort (NHIS-NSC) claims database for 2002– taneous transluminal coronary angioplasty, coronary 2012, which includes information on approximately 1 artery bypass surgery, and leg amputation, according to million Koreans since 2002. The NHIS-NSC used a 2.5% the stage of diabetic retinopathy and nephropathy . (n = 1,025,340) stratified random sampling method, with Furthermore, in Korea, the number of patients with dia- the goal of providing representative, useful health insur- betes increased by 24.6% from 2010 to 2015, and among ance and health examination data to public health them, approximately 5.8% were diagnosed with diabetic researchers and policy makers. Of note, additional data renal complication [3, 4]. These diabetic complications handling to account for missing data is unnecessary, as need periodic monitoring, dietary modification, and early the data was already processed by the NHIS upon intervention. Therefore, to effectively manage chronic researcher request. diseases such as diabetes within limited resource Based on the NHI data, newly diagnosed diabetic renal settings, policymakers have developed clinical practice complication (International Classification of Diseases guidelines, with attention focused on the benefits of tenth edition diagnosis code (ICD-10): E11.2) patients continuity of care (COC) [5–7]. were identified. The data were also analyzed to deter- COC is an essential concept for high-quality patient mine the association between COC level and the inci- care, and it entails how patients’ experiences are linked dence of ESRD. The data were stratified according to with care over time or the connectedness of the discrete age, sex, region, health insurance type, disability status, elements of care [8, 9]. Furthermore, COC has positive residence area, income decile, insulin treatment, and the effects on various outcomes; for example, it encourages severity of complications. The data included participant patient satisfaction, treatment adherence [10, 11], or in- characteristics as well as medical institution variables creases the recognition rate of diabetes, and leads to bet- (hospital category, the number of beds, the number of ter glycemic control among diabetic patients [12–16]. doctors, and hospital location) as covariates. All individ- Moreover, higher COC with the usual provider for dia- ual- and hospital-level characteristics were measured at betes mellitus is associated with a lower risk of future or baseline in 2005. The diagnoses were based on the preventable hospitalizations for long-term diabetic com- ICD-10 codes. plications, and it might further decrease medical costs [17, 18]. Therefore, COC and management of the Participants physician-patient relationship should be reinforced in The total number of individuals with diabetic nephropa- the early stages of the disease condition for a more thy (ICD-10: E11.2) was 4019 from 2002 to 2013. Of effective management of diabetes . these, 3706 patients were diagnosed in 2005, and to spe- Higher provider COC is characterized by better cifically analyze newly diagnosed diabetic nephropathy approaches to the sharing of disease information, with- patients, we excluded 130 prior diabetic nephropathy out information asymmetry, while eliminating conflicts patients and 11 prior ESRD patients from 2002 to 2004, of interest between patient and practitioners . and included the first diabetic nephropathy Despite evidence that COC is associated with better (ICD-10:E11.2) patient from 2005. Because we focused patient outcomes, empirical studies of associations on the continuity of ambulatory care for 8 years among between COC and health outcomes in the Korean popu- surviving diabetic nephropathy patients, and observed lation are rare. Moreover, few studies exist on the effect the incidence of ESRD from 2005 to 2013, we excluded of COC on the incidence of other chronic diseases like 47 patients who died during this period. Moreover, in ESRD. Therefore, the aim of this study was to investigate calculating COC index, we limited included patients to the association between continuity of ambulatory care those on hospital outpatient visits due to diabetic and the incidence of chronic kidney disease among nephropathy; hence, 245 inpatients and 21 health center patients with renal diabetic complications, using the users were excluded. Using these criteria, the final study administrative claims data from the Republic of Korea’s sample included 3565 diabetic nephropathy patients. National Health Insurance (NHI) scheme. Moreover, we The flow diagram of the study participants’ selection is identified individual- and hospital-level factors associ- shown in Fig. 1. Furthermore, to perform survival ana- ated with the incidence of ESRD among diabetic lysis using a Cox proportional hazards models, we chose nephropathy patients. an appropriate follow-up period. In patients diagnosed Jang et al. BMC Nephrology (2018) 19:127 Page 3 of 12 Fig. 1 Flow diagram of study participants with ESRD, we set the follow-up period as the time from ambulatory visits with the same practitioner over the the start date to the date of ESRD occurrence. On the study period. On the other hand, patients who made less other hand, we set the end date as the death date or the than 75% of visits to the same doctor were classified as end of 2013 for the non-ESRD group and deceased having bad COC. This measure has been validated patients. extensively in previous studies. Therefore, we used the COC index because the caregivers are not determined in Independent variable: COC advance in Korea, and we calculated patients’ visits until COC was included as an independent variable, and the occurrence of ESRD. Furthermore, the number of included both prescription continuity and the clinical days from the first diagnosis to ESRD occurrence was management of diseases. Moreover, the measurement of calculated. The COC index was categorized as a binary COC considered the characteristics of the Korean medial variable into the ‘Good COC group’ (COC index≥0.75), delivery system, where patients are allowed to freely when diabetic patients most often visited the same med- choose their preferred primary care. Therefore, the COC ical institutions for the treatment of diabetic complica- index was chosen to measure consistency of care. tions, and the ‘Bad COC group’ (COC index< 0.75), According to the findings of previous studies, COC when diabetic patients visited different medical institu- can be determined in many ways, including the use of tions on most occasions, similar to the findings of previ- the Modified Continuity Index, and the Most Frequent ous studies [28, 29]. Provider Continuity [21–25]. As devised by Bice et al. (1977), the COC index can be Dependent variable: The incidence of ESRD calculated by the total patient contact times with the In this study, the dependent variable was the incidence of medical service providers and the number of healthcare ESRD. To measure this, we identified patients who devel- providers. The COC index ranges from 0 (when patients’ oped ESRD using ICD-10 codes such as hemodialysis visits to medical institutions occur with different pro- (O7020, O9991), peritoneal dialysis (O7062), or kidney viders at each visit) to 1 (when patients present to the transplant (R3280), among newly diagnosed patients with same medical institution for outpatient services on many diabetic renal complications from 2005 to 2012. occasions). Therefore, an index close to 1 implies that the patient frequently utilizes a particular medical Covariates service provider; hence, the COC level is high. We examined covariates that affected the occurrence of The formula for determining the COC index is shown ESRD and the COC of patients with diabetic renal below: complications. Furthermore, we included both individ- ual- and hospital-level characteristics as covariates. Individual-level characteristics adjusted for included sex, n −N age, residence area, health insurance type, income level, j¼1 COC ¼ insulin treatment, disability status, and disease severity. NNðÞ −1 Age was categorized into five groups (under 50, 50–59, Where N = the total number of visits, M = the number 60–69, 70–79, and over 80 years). Residence area was of available medical service providers, and n = the num- divided into three groups (metropolitan, urban, and rural th ber of visits to the j providers. areas). Health insurance type was categorized as health From previous studies [26, 27], we used the cutoff insurer and Medicaid user. Furthermore, we divided in- point of 0.75 for the COC index, whereby patients were come into four quartiles of 25% each, from the lowest classified as having good COC if they had 75% of their (Q1) to the highest (Q4). Jang et al. BMC Nephrology (2018) 19:127 Page 4 of 12 Disease severity was measured by the Korean log-rank test. The adjusted hazard ratios (HRs) for the Diagnosis-Related Group (KDRG) codes and comorbidi- incidence of ESRD by the Cox proportional hazards ties. Related with diabetes mellitus were six-digit KDRG models were calculated using the PROC PHREG proced- codes including K60000, K60001, K60002, and K60003. ure in SAS. Third, after adjusting for other individual- Among the numbers, the last digits (0, 1, 2, and 3) and hospital-level characteristics on the regression represent the Patient Clinical Complexity Level (PCCL). model, we conducted subgroup analysis to determine By definition, ‘0’ implies without severe status or comor- the association between COC and the incidence of ESRD bidity, ‘1’ indicates the presence of accompanying by residence area, income level, and disability to determine mid-level complications or comorbidities, ‘2’ refers to the the effect of socio-economic status on patients’ COC. presence of more severe complications or comorbidities, while ‘3’ implies the most severe complications or Results comorbidities. The increasing severity implies that the Demographic characteristics of the study population larger the number, the higher the severity of the disease. A total of 3565 patients with the diagnosis of Furthermore, PCCL was defined as the severity value non-insulin-dependent diabetes mellitus with renal com- according to the number of complications related to dia- plications were included in this retrospective cohort betes mellitus using the KDRG codes. study from 2005 to 2012. Table 1 presents the demo- Moreover, we adjusted for both hospital- and graphic characteristics of the study population at base- individual-level characteristics. We included hospital line (2005). To calculate the COC index, we defined the classifications (general hospital, hospital, and clinic), time period as the time from when a diabetic nephropa- categorized according to the Korean medical law. Based thy patient started using hospital ambulatory care to the on hospital organization type, acute care hospitals were event time, whether ESRD was diagnosed or not, by the included in ‘hospitals’, while ‘general hospitals’ were end of follow-up. During the 8-year follow-up from 2005 defined by The Korean Hospital Association as hospitals to 2012, 83 patients were diagnosed with ESRD with a with 80 or more beds and at least eight major clinical mean ± standard deviation (SD) COC level of 0.398 ± departments. Such clinical departments include internal 0.492. Among the 3482 patients with no ESRD, the medicine, general surgery, pediatrics, obstetrics and mean ± SD COC level was 0.612 ± 0.487. gynecology, radiology, emergency medicine, and path- Until 2012, among the 83 of 3565 diabetic nephropathy ology. Moreover, number of beds, number of doctors, patients who developed ESRD, the overall incidence of and hospital location (metropolitan, urban, rural) were ESRD was 520.4 per 100,000 person-years (95% confidence included as covariates. The capital and largest city interval (CI) 419.7–645.4 per 100,000 person-years), and (“Seoul”) was included in “Metropolitan area,” while 332.4 and 830.4 per 100,000 person-years for the Good and “Urban” was defined as cities classified by administrative Bad COC, respectively. The comparison of subject’sbase- districts (such as Busan, Daegu, Incheon, Gwangju, line characteristics showed that ‘Bad COC’ had higher inci- Daejeon, and Ulsan). For both the number of beds and dence of ESRD than Good COC. the number of doctors, the medians were reported; all of these covariates and all diagnostic information were Survival analysis for the incidence of ESRD using Kaplan- collected at baseline (2005). Meier curve Fig. 2 shows the Kaplan-Meier curve of the incidence Statistical analysis: Survival analysis probability of ESRD by independent variables among the For this study, we used SAS version 9.4 (SAS institute ‘Good’ and ‘Bad’ COC groups. According to the curve, Inc. Cary NC, USA) statistical software and the signifi- those in the ‘Good COC group’ had lower incidence cance level was set at 5%. First, the distribution of demo- probability than those in the ‘Bad COC group’, and, with graphic characteristics among diabetic patients with time, the incidence probabilities in both groups became renal complications were assessed at baseline. Baseline sustained. Moreover, the median time of survival for categorical variables were expressed as numbers and individuals with Good and Bad COC were 4.58 and percentages and were compared using the χ test. 4.30 years, respectively, in the Kaplan-Meier analysis. Second, survival analysis was performed to determine the effect of COC on the incidence of ESRD among Cox proportional hazard model showing association diabetic patients with renal complications using the Cox between continuity of care and the incidence of end- proportional hazards models. Furthermore, the stage renal disease Kaplan-Meier curve was plotted to demonstrate the es- Table 2 shows the results of the Cox proportional haz- sential assumption of proportional hazard regression ards models showing the association between COC and model. Moreover, we calculated the mean time to the the incidence of ESRD among diabetic patients with diagnosis of ESRD by each categorical variable using the renal complications. After adjusting for all covariates Jang et al. BMC Nephrology (2018) 19:127 Page 5 of 12 Table 1 Distribution of subject characteristics by ESRD occurrence at baseline 2005 Total ESRD occurrence Person-year Incidence Rate p-value −5 (×10 ) Yes No Patient level Sex Male 1937 44 (2.3) 1893 (97.7) 8597.9 511.8 0.81 Female 1628 39 (2.4) 1589 (97.6) 7350.2 530.6 Age (years) Under 50 738 9 (1.2) 729 (95.0) 3625.3 248.3 0.16 50–59 926 24 (2.6) 902 (94.1) 4169.8 575.6 60–69 1097 26 (2.4) 1071 (93.4) 4958.0 524.4 70–79 666 20 (3.0) 646 (93.2) 2665.8 750.2 80 and more 138 4 (2.9) 134 (94.2) 529.1 755.9 Residence area Metropolitan 829 18 (2.2) 811 (97.8) 3849.9 467.5 0.52 Urban 883 17 (1.9) 866 (98.1) 4105.7 414.1 Rural 1853 48 (2.6) 1805 (97.4) 7992.4 600.6 Health insurance type Health insurance 3471 80 (2.3) 3391 (97.7) 15,694.6 509.7 0.59 Medical aid 94 3 (3.2) 91 (96.8) 253.5 1183.2 Income Q1 (Low) 816 27 (3.3) 789 (96.7) 3360.8 803.4 0.16 Q2 747 12 (1.6) 735 (98.4) 3361.4 357.0 Q3 787 18 (2.3) 769 (97.7) 3694.2 487.2 Q4 (High) 1215 26 (2.1) 1189 (97.9) 5531.7 470.0 Insulin treatment Yes 160 3 (1.9) 157 (98.1) 15,312.6 522.4 0.69 No 3405 80 (2.4) 3325 (97.7) 635.5 472.1 Disabled type Yes 482 28 (5.8) 454 (94.2) 1919.1 1459.0 <.0001 No 3083 55 (1.8) 3028 (98.2) 14,028.9 392.0 PCCL index 0 3469 80 (2.4) 3389 (97.6) 15,722.7 521.5 0.23 1 61 2 (3.3) 59 (96.7) 131.5 1521.4 ≥2 35 1 (4.2) 34 (97.1) 93.9 1064.9 Continuity of care Good (0.75 ≤ COC index) 2164 33 (1.5) 2131 (98.5) 9926.7 332.4 <.0001 Bad (COC index< 0.75) 1401 50 (3.6) 1351 (96.4) 6021.4 830.4 Hospital level Hospital classification General hospital 2040 52 (3.5) 1968 (96.5) 9266.5 777.0 <.0001 Hospital 97 11 (11.0) 86 (89.0) 408.1 245.1 Clinic 1428 20 (1.4) 1408 (98.6) 6273.5 159.4 Number of beds 787.5 ± 461.0 475.7 ± 508.6 <.0001 Number of doctors 220.2 ± 164.3 136.2 ± 194.0 <.0001 Jang et al. BMC Nephrology (2018) 19:127 Page 6 of 12 Table 1 Distribution of subject characteristics by ESRD occurrence at baseline 2005 (Continued) Total ESRD occurrence Person-year Incidence Rate p-value −5 (×10 ) Yes No Location Metropolitan 973 19 (2.0) 954 (7.5) 4608.6 412.3 0.63 Urban 1016 24 (2.4) 992 (6.5) 4681.2 512.7 Rural 1576 40 (2.5) 1536 (5.3) 6658.3 600.8 Survey Year 2005 587 15 (2.5) 572 (97.5) 4975.1 301.5 0.36 2006 440 5 (1.1) 435 (98.9) 2797.6 178.7 2007 493 10 (2.0) 483 (98.0) 2611.7 382.9 2008 526 12 (2.3) 514 (97.7) 2278.4 526.7 2009 544 17 (3.1) 527 (96.9) 1797.4 945.8 2010 457 14 (3.1) 443 (96.9) 1073.1 1304.6 2011 255 8 (3.1) 247 (96.9) 343.7 2327.4 2012 263 2 (3.8) 261 (96.2) 71.0 2817.8 Total 3565 83 (2.3) 3482 (97.7) COC index ranges from 0 to 1; 1 means that one patient has visited only one physician, and we could not put other continuity indices in this model at the same time because of multicollinearity among indices; Good is defined as COC score greater than 0.75, and Bad is defined as COC score less than 0.75 The larger value was used regardless of whether the number of complications was related to diabetes mellitus or PCCL (Patient Clinical Complexity Level) using the KDRG code (individual- and hospital-level characteristics), ne- In Additional file 1: Table S1, factors associated with phropathy patients with Bad COC had 1.99 times ESRD incidence are shown. Among individual-level higher risk of ESRD incidence (95% CI: 1.27–3.12). characteristics, some variables were associated with We also implemented sensitivity analysis, which ESRD incidence. The female sex showed a hazard ratio included 47 study participants who had died, and the (HR) of 1.06, which was higher than that of males, while results were similar to those in Table 2 (see patients above 50 had higher probability of ESRD than Additional file 1:Table S1). those under 50. Patients who lived in rural (HR: 0.84 Fig. 2 the Kaplan-Meier curve of the incidence probability of ESRD by divided COC groups and legend for exposure categories outcome with p- value for log-rank test Jang et al. BMC Nephrology (2018) 19:127 Page 7 of 12 Table 2 Association between continuity of care and ESRD areas, no significant association was observed between incidence* COC and the incidence of ESRD. ESRD incidence Furthermore, among patients with diabetic complica- tion in the lowest (Q1), second lowest (Q2), and the Hazard Ratio 95% CI highest income levels, hospital medical services were Continuity of care more poorly utilized, the incidence probability was Good (COC index≥0.75) 1.00 –– increased, and these associations were statistically sig- Bad (COC index< 0.75) 1.99 (1.27 – 3.12) nificant. Moreover, patients with disabilities in the *Adjusted for sex, age, residence area, health insurance type, disability type, ‘Good’ COC group had a lower probability of ESRD insulin treatment, PCCL index, hospital classification, number of beds, number of doctors, and hospital location compared with the ‘Bad’ COC group, with 1.86 higher incidence probability of ESRD (95% CI: 1.07–3.26). 95% CI: 0.41–1.70), or urban areas (HR: 0.50 95% CI: Among hospital factors, hospitals in rural areas had a 0.19–1.34) had lower incidence probability of ESRD than lower probability of ESRD for the ‘Good’ COC group metropolitan citizens. In addition, those in the lowest than the ‘Bad’ COC group, with 2.05 higher incidence income group had higher probability of developing renal probability of ESRD (95% CI: 1.09–3.85). In addition, failure than patients in the highest income group (HR: hospital classification was associated with the relation- 1.69 95% CI: 0.95–2.98). However, none of these findings ship between COC and the occurrence of ESRD. How- were statistically significant. Similarly, although patients ever, no significant association was observed between with disabilities had higher ESRD incidence (2.70) than COC and the incidence of ESRD in hospitals and clinics. patients without disabilities, no significant difference (95% CI: 0.64–43) was found. Discussion Among hospital-level characteristics, patients who According to findings of a previous study, COC consists utilized hospitals with either the highest number of beds of various elements not only the service provider and or doctors had the highest probability of ESRD. Further- patient relationship, but also continuous data accessibil- more, when the hospital was located in a rural or urban ity or total care management, which meant coherent area, the probability was also higher; however, these delivery of care from different doctors . Although findings were not statistically significant. In addition, COC is a complex concept, the core component is the primary care level attendance showed a lower hazard consistency of contacts between specific patients and the (0.48) of ESRD than with general hospital attendance, providers . Moreover, according to previous studies, with no statistically significant association (95% CI: the higher the severity or number of diabetic complica- 0.20–1.14). tions, the higher the probability of adverse patient out- comes, such as mortality and hospitalizations; therefore, there is need for consistency in the care of diabetic Subgroup analysis: Association between COC and the patients [32, 33]. In addition, newly diagnosed patients incidence of ESRD by residence area, income level, and with hypertension, diabetes, hyperlipidemia, and those disability with bad COC levels, have higher risks of heart attack In Table 3, the results of the subgroup analysis of the and cardiovascular heart diseases than patients with association between continuity of care with and occur- good COC levels . rence of ESRD occurrence by residence area, income, The results of this study demonstrate that good COC disability type, sex, and health insurance type are shown. in patients with diabetic renal complications is signifi- After building the model adjusted for other individual- cantly associated with a lower probability of ESRD. and hospital-level characteristics such as sex, age, health Subgroup analysis was performed to determine whether insurance type, hospital categories, number of doctors, there is an association between COC and ESRD by resi- number of beds, and hospital location, we conducted dence area, income level, and disability, revealing a subgroup analysis to determine the association between higher probability of ESRD among patients with bad COC and the incidence of ESRD by residence area, COC index (< 0.75) who lived in metropolitan and rural income level, disability type, sex, and health insurance areas. Moreover, diabetic patients (who either had type (Table 3). disabilities or were in the lowest income level) with bad Among patients with renal diabetic complications liv- COC level had higher probability of ESRD, but there ing in metropolitan areas, those with ‘Bad’ COC index was no statistically significant association with these (< 0.75) had 2.11 times higher probability of developing variables. ESRD (95% CI: 1.04–5.30), and this was similar to the According to the findings of this study, the association finding among patients living in rural areas (HR: 1.92, between COC for diabetic complications and the prob- 95% CI: 1.06–3.48) or urban areas. However, in urban ability of developing ESRD is consistent with the Jang et al. BMC Nephrology (2018) 19:127 Page 8 of 12 Table 3 Subgroup analysis for the association continuity of care with occurrence of secondary diabetic complication by patient- level and hospital-level factors Continuity of care p-value Good (0.75 ≤ COC index) Bad (COC index < 0.75) Hazard ratio (Ref.) Hazard ratio 95% CI ESRD incidence Patient-level Sex Male 1.00 2.04 (1.09 – 3.82) 0.03 Female 1.00 1.86 (0.96 – 3.60) 0.07 Age Under 50 1.00 3.00 (0.75 – 11.98) 0.12 50–59 1.00 3.83 (0.36 – 1.95) 0.83 60–69 1.00 4.24 (1.78 – 10.09) 0.00 70–79 1.00 4.39 (1.59 – 12.07) 0.00 80 and more 1.00 1.41 (0.20 – 10.01) 0.73 Residence area Metropolitan 1.00 2.11 (1.04 – 5.30) 0.04 Urban 1.00 1.70 (0.62 – 4.65) 0.24 Rural 1.00 1.92 (1.06 – 3.48) 0.04 Health insurance type Health insurance 1.00 1.91 (1.21 – 3.03) 0.01 Medical aid 1.00 1.93 (0.17 – 21.57) 0.59 Income Q1 (Low) 1.00 2.04 (0.92 – 4.52) 0.08 Q2 1.00 1.70 (1.38 – 4.82) 0.02 Q3 1.00 0.94 (0.36 – 2.45) 0.91 Q4 (High) 1.00 2.27 (1.01 – 5.14) 0.05 Disabled type Yes 1.00 2.08 (0.96 – 4.48) 0.06 No 1.00 2.23 (1.43 – 3.49) 0.00 Hospital-level Hospital classification General hospital 1.00 1.86 (1.15 – 2.99) 0.01 Hospital 1.00 2.11 (0.64 – 6.91) 0.22 Clinic 1.00 2.73 (0.79 – 9.42) 0.11 Number of beds Q1 (Low) 1.00 2.60 (1.03 – 6.59) 0.04 Q2 1.00 1.39 (0.54 – 3.60) 0.50 Q3 (High) 1.00 2.00 (1.11 – 3.61) 0.02 Number of doctors Q1 (Low) 1.00 1.99 (0.72 – 5.48) 0.18 Q2 1.00 1.87 (0.67 – 5.25) 0.24 Q3 (High) 1.00 1.86 (1.06 – 3.27) 0.03 Location Metropolitan 1.00 1.96 (0.79 – 4.86) 0.15 Jang et al. BMC Nephrology (2018) 19:127 Page 9 of 12 Table 3 Subgroup analysis for the association continuity of care with occurrence of secondary diabetic complication by patient- level and hospital-level factors (Continued) Continuity of care p-value Good (0.75 ≤ COC index) Bad (COC index < 0.75) Hazard ratio (Ref.) Hazard ratio 95% CI Urban 1.00 3.46 (1.52 – 7.91) 0.00 Rural 1.00 2.05 (1.09 – 3.85) 0.03 findings of previous studies, which emphasized that the disabilities. However, there was no statistically significant better the COC, the better the outcome of various difference in the COC among patients with disabilities. chronic conditions. Liao et al. revealed that diabetes As shown in the subgroup analysis, patients who lived mellitus patients with a high medical care-seeking in rural areas with worse COC had higher probability of consistency with a physician had a lower risk of diabetic ESRD. Although there was no difference by location in complications compared with patients having a medium the subgroup analysis, in both regions the continuity of or low medical care-seeking consistency . Christakis diabetic care among nephropathy patients was important et al. further emphasized that children with a medium to prevent future ESRD. Additionally, except for or high COC were less likely to be hospitalized for dia- mid-high income level (Q3), improved COC level af- betic ketoacidosis . Another study revealed that fected the probability of ESRD, and although this was consistency in diabetic care increases patient satisfaction not statistically significant, patients with bad COC levels and decreases the risk of other chronic diseases . had higher probability of ESRD. A similar trend was also The benefits of COC are enhanced among chronic dis- observed for patients with disabilities and other covari- ease patients because they tend to demand treatment ates. With disability status, the proportion of ESRD was more consistently than others without chronic diseases, higher in patients with disabilities and those with the who tend to engage in passive medical services lowest income level. Although the socioeconomic status utilization. Thus, the physician-patient relationship showed no direct effect on the probability of ESRD, and might be improved. Furthermore, improved COC was there was no significant gap in the consistency of associated with a reduction in all-cause mortality, and diabetic care, there might be underlying disparities in prevented hospitalizations as reported in previous the care of patients who are socially vulnerable, such as studies [38, 39]. patients with disabilities or those in the lowest income The findings of our study indicate that improved COC level. Hence, these subgroups of patients should be among patients with nephropathy is associated with a aware of how to prevent the occurrence of ESRD. This lower probability of ESRD. Based on the patients’ socio- is similar to the explanations from previous studies, economic statuses, some distinctions were observed which emphasized that poorer access to health care among the variables. Specifically, with patients who lived among black patients might explain their excess risk of in rural areas, the COC level was lower than that of ESRD, beyond the excess risk explained by demographic, those living in urban areas. In addition, based on the socioeconomic, lifestyle, and clinical factors [42–45]. hospital characteristics, patients tend to visit clinics However, our study demonstrated the effect of COC on rather than general hospitals or hospitals, and so the the probability of ESRD, and showed that patients with COC index was higher among clinic users. The gap in nephropathy (patients with renal complication) with bad the COC level influenced the probability of ESRD. Ac- COC level had higher probability of ESRD. In future re- cording to our findings, patients who used medical ser- search, the effect of COC on the actual change in clinical vices consistently had lower probability of ESRD than parameters, such as glycemic control, should be imple- patients with lower COC index. This is consistent with mented to emphasize the importance of continuous care. the findings of previous studies, which emphasized on Moreover, policy development to encourage COC the importance of continuous diabetic care to prevent among patients with diabetic complications, aimed at future hospitalizations, mortality, and excessive medical preventing severe chronic disease and managing diabetic expenses. Furthermore, patients who used clinics for the complications more effectively in the future, should be care of diabetes had a lower probability of ESRD than implemented. There is need for further research aimed patients who visited general hospitals [40, 41]. at minimizing the probability of ESRD among vulnerable Among other characteristics, including the lowest classes of patients, among whom access to care is more income, and elderly patients, higher probability of ESRD difficult . was reported; in particular, patients with disabilities had Our study has some limitations. First, we only used much higher probability of ESRD than those without the procedure codes and ICD-10 codes from NHI claim Jang et al. BMC Nephrology (2018) 19:127 Page 10 of 12 data for defining ESRD. The procedures included other indices such as the Usual Provider Care index and hemodialysis, peritoneal dialysis, and kidney transplant. the Sequential Continuity of Care index; hence, we were However, ESRD could be divided into several stages unable to explain which COC index might explain the based on the disease severity while other clinical proce- probability of ESRD better. Moreover, despite our inclu- dures could be used for defining ESRD. Hence, we might sion of the patients’ individual- and hospital-level char- have underestimated the number of ESRD patients in acteristics as covariates in the main analysis, we failed to our study. However, we defined ESRD by ICD-10 codes test for correlations within hospitals. Since we only that were implemented for improving kidney functions; determined the association of COC among diabetic hence, the findings of our study showed that better nephropathy patients, further research should be imple- consistency in diabetic nephropathy care resulted in a mented to also assess the correlation within hospitals lower probability of ESRD. among the care episodes. Second, since we conducted a retrospective cohort Furthermore, because of inappropriate handling of study, with the exposure categorized as a dichotomous missing data, there is a potential of the effect estimate variable, misclassification was inherent. In cohort studies, (HR) to be biased towards the null. Since we determined exposure misclassification commonly occurs. If the assess- the association between COC and the incidence of ESRD ment of exposure is implemented independently of the among survivors during the 8-year follow-up, the exclu- disease diagnosis, misclassification tends to result in spuri- sion of patients who died tends to underestimate the ous conclusion. Furthermore, not only with the exposure number of diabetic nephropathy patients. Furthermore, but also with certain covariates, using administrative data we excluded inpatients or health center users. To assess usually leads to misclassifications. Minimizing misclassifi- the effect of the COC on the incidence of ESRD, all out- cation in future research should be ensured. patient medical use history should be included. To The third limitation is the lack of information on minimize potential selection bias, more refined sampling health behaviors such as perceived health status, smok- methods are needed, such as matching, or including the ing, depression, and drinking status, which might affect appropriate control group. the incidence of ESRD. Health behaviors and blood Despite these limitations, our study has several glucose control are highly correlated, as revealed by strengths. First, we analyzed a representative sample of pa- findings of previous studies. Due to unmeasured or un- tients with diabetic renal complications using nationwide known factors, which might act as effect modifiers for claims data in Korea. Secondly, unlike most previous stud- the outcome (ESRD incidence), residual confounding ies, we included each patient’s socioeconomic status, might exist in this study, and this might have resulted in which could define the difference in COC between socially an imperfect adjustment, or could have led to spurious vulnerable patients, including patients with lower income conclusions. To minimize this limitation, propensity level, or those with or without disabilities. score methods for determining the appropriate covari- Our results support the hypothesis that reducing frag- ates to adjust for in the analysis should be applied in mented care and improving COC among diabetic com- future research. plication patients can decrease the probability of ESRD. Moreover, we did not include hospital or doctor char- We therefore encourage policymakers to recognize the acteristics that might have had an influence on the prob- need for an effective healthcare delivery system that pro- ability of ESRD. In addition, we could not explain how motes COC. Furthermore, to enhance accessibility, we the severity of diabetic renal complications affected the suggest the need to undertake studies aimed at examin- probability of ESRD, and with the limited data, we were ing the relationship between medical service providers unable to investigate whether study patients used out- and patients with diabetic complications. In addition, to patient prescriptions that might also have had an effect prevent the progression of ESRD among patients with on the progression of ESRD. Furthermore, since the type 2 diabetes mellitus with renal complications, it is cause of death was not included in the health insurance necessary to undertake measures aimed at improving the qualification database, we could not clarify whether pa- sustainability of patients with disabilities. Continuous tients died as a result of diabetic complication or not. efforts of medical staff and changes in the national pol- Since our focus was on the association between continu- icy for chronic disease management are necessary to im- ity of ambulatory care and the incidence of ESRD during prove outpatient care for the sustainable use by patients. the 8-year follow-up, we limited the analysis to survivals during follow-up period. Therefore, the number of Conclusion patients might have been underestimated in this study. In conclusion, we measured COC among newly diag- The fourth limitation is that we did not evaluate the nosed nephropathy patients during the entire study COC in multiple ways. We focused on the COC based period and analyzed the relationship between the COC on the COC index alone. However, there are several and the probability of ESRD. According to our results, Jang et al. BMC Nephrology (2018) 19:127 Page 11 of 12 higher continuity of care was associated with lower risk of Author details The Health Insurance Dispute Mediation Committee, Ministry of Health & ESRD among patients with diabetic renal complication. Welfare, Sejong Government Complex, Sejong City, Republic of Korea. To prevent ESRD among diabetic renal complication pa- 2 Department of Health Policy and Management, Graduate School of Public tients, continuous follow-up should be encouraged. More- Health, Yonsei University, Seoul, Republic of Korea. Department of Public Health, Graduate School, Yonsei University, Seoul, Republic of Korea. over, there were disparities in health outcomes between Institute of Health Services Research, Yonsei University, Seoul, Republic of socially vulnerable groups. To bridge the gap in health 5 Korea. Department of Preventive Medicine & Institute of Health Services outcomes among the different socio-economic patient Research, Yonsei University College of Medicine, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-752, Republic of Korea. National Evidence-Based groups, accessibility of care without disparities should be Healthcare Collaborating Agency, Seoul, Republic of Korea. guaranteed. Further research is needed on the associations between long-term COC and various healthcare out- Received: 2 August 2017 Accepted: 25 May 2018 comes, including medical expenses or patient’squality of life. Moreover, to ensure that the study findings could be References generalized to the larger population of patients, long-term 1. O'Brien JA, Patrick AR, Caro J. Estimates of direct medical costs for follow-up of patients with diabetic complications would microvascular and macrovascular complications resulting from type 2 be needed. Health behavior, changes in the socioeconomic diabetes mellitus in the United States in 2000. Clin Ther. 2003;25(3):1017–38. 2. Hwang JA, Park TC, Jung SH, Kim HJ, Kim DJ, Kim SH, Nam MS, Kim TH, Lee status, and disease progression mechanisms, which might MK, Lee KW. Direct medical costs of type 2 diabetic patients in the tertiary mediate the utilization of medical services and health out- hospital. Korean Diabetes J. 2008;32(3):259–68. comes, should be considered. 3. Song SO, Jung CH, Song YD, Park C-Y, Kwon H-S, Cha BS, Park J-Y, Lee K-U, Ko KS, Lee B-W. Background and data configuration process of a nationwide population-based study using the Korean National Health Insurance System. Additional file Diabetes Metab J. 2014;38(5):395–403. 4. Parfrey PS, Griffiths SM, Barrett BJ, Paul MD, Genge M, Withers J, Farid N, McManamon PJ. Contrast material-induced renal failure in patients Additional file 1: Table S1. Sensitivity analysis of the association between with diabetes mellitus, renal insufficiency, or both. N Engl J Med. 1989; continuity of care and ESRD incidence (including deaths). (DOC 31 kb) 320(3):143–9. 5. Kim DJ. The epidemiology of diabetes in Korea. Diabetes Metab J. 2011; Abbreviations 35(4):303–8. CI: Confidence interval; COC: Continuity of care; ESRD: End-stage renal 6. Davis TM, Millns H, Stratton IM, Holman RR, Turner RC. Risk factors for stroke disease; HR: Hazard ratio; ICD-10: International Classification of Diseases tenth in type 2 diabetes mellitus: United Kingdom prospective diabetes study edition diagnosis code; KDRG: Korean Diagnosis-Related Group; NHI: National (UKPDS) 29. Arch Intern Med. 1999;159(10):1097–103. Health Insurance; PCCL: Patient Clinical Complexity Level; Q: Quartile; 7. group DEr. Effect of intensive diabetes treatment on albuminuria in type 1 SD: Standard deviation diabetes: long-term follow-up of the diabetes control and complications trial and epidemiology of diabetes interventions and complications study. Acknowledgements Lancet Diabetes Endocrinol. 2014;2(10):793–800. The authors would like to thank the associate managing editor of Editage for 8. Starfield B, Simborg D, Johns C, Horn S. Coordination of care and its reviewing and translating an earlier and final version of this article. relationship to continuity and medical records. Med Care. 1977;15:929–38. 9. Van Walraven C, Oake N, Jennings A, Forster AJ. The association between Funding continuity of care and outcomes: a systematic and critical review. J Eval Clin This research received no specific grant from any funding agency in the Pract. 2010;16(5):947–56. public, commercial, or not-for-profit sectors. 10. Knight JC, Dowden JJ, Worrall GJ, Gadag VG, Murphy MM. Does higher continuity of family physician care reduce hospitalizations in elderly people with diabetes? Popul Health Manag. 2009;12(2):81–6. Availability of data and materials The datasets used and analyzed during the current study are available from 11. Lin W, Huang I-C, Wang S-L, Yang M-C, Yaung C-L. Continuity of diabetes the corresponding author on reasonable request. care is associated with avoidable hospitalizations: evidence from Taiwan’s National Health Insurance scheme. Int J Qual Health Care. 2009;22(1):3–8. 12. Hong JS, Kang HC, Kim J. Continuity of care for elderly patients with Authors’ contributions diabetes mellitus, hypertension, asthma, and chronic obstructive pulmonary YJJ and YSC participated in the conception of the study, data collection and disease in Korea. J Korean Med Sci. 2010;25(9):1259–71. analysis, and drafted the manuscript. YJJ, CMN, KTM, and ECP contributed to 13. Liu CW, Einstadter D, Cebul RD. Care fragmentation and emergency drafting, critical reviewing, and finalization of the manuscript. All authors department use among complex patients with diabetes. Am J Manag Care. contributed to development of the manuscript and approved the final 2010;16(6):413–20. manuscript. 14. Saultz JW. Defining and measuring interpersonal continuity of care. Ann Fam Med. 2003;1(3):134–43. Ethics approval and consent to participate 15. Donaldson MS. Continuity of care: a reconceptualization. Med Care Res Rev. The ethical committee of the Graduate School of Public Health in Yonsei 2001;58(3):255–90. University approved the study(2–1,040,939-AB-N-01-2016-313). This study did 16. Kerse N, Buetow S, Mainous AG, Young G, Coster G, Arroll B. Physician- not require informed consent from patients since the datasets were patient relationship and medication compliance: a primary care completely anonymized and therefore contained no personal information on investigation. Ann Fam Med. 2004;2(5):455–61. participants. 17. Chen C-C, Chen S. Better continuity of care reduces costs for diabetic patients. Am J Manag Care. 2011;17(6):420–7. Competing interests 18. Hänninen J, Takala J, Keinänen-Kiukaanniemi S. Good continuity of care The authors declare that they have no competing interests. may improve quality of life in type 2 diabetes. Diabetes Res Clin Pract. 2001;51(1):21–7. Publisher’sNote 19. Koopman RJ, Mainous AG III, Baker R, Gill JM, Gilbert GE. Continuity of care Springer Nature remains neutral with regard to jurisdictional claims in and recognition of diabetes, hypertension, and hypercholesterolemia. Arch published maps and institutional affiliations. Intern Med. 2003;163(11):1357–61. Jang et al. BMC Nephrology (2018) 19:127 Page 12 of 12 20. Gill JM, Mainous AG, Diamond JJ, Lenhard MJ. Impact of provider 44. Li S, McAlpine DD, Liu J, Li S, Collins AJ. Differences between blacks and continuity on quality of care for persons with diabetes mellitus. Ann whites in the incidence of end-stage renal disease and associated risk Fam Med. 2003;1(3):162–70. factors. Adv Ren Replace Ther. 2004;11(1):5–13. 21. Bice TW, Boxerman SB. A quantitative measure of continuity of care. Med 45. Perneger TV, Whelton PK, Klag MJ. Race and end-stage renal disease: Care. 1977;15(4):347–9. socioeconomic status and access to health care as mediating factors. Arch Intern Med. 1995;155(11):1201–8. 22. Cyr M-C, Martens AC, Berbiche D, Perreault S, Blais L. Continuity of care in 46. Krop JS, Coresh J, Chambless LE, Shahar E, Watson RL, Szklo M, Brancati FL. the ambulatory treatment of adolescents with asthma. J Adolesc Health. A community-based study of explanatory factors for the excess risk for early 2006;39(6):926. e911–7. renal function decline in blacks vs whites with diabetes: the atherosclerosis 23. Yoon CH, Lee SJ, Choo S, Moon OR, Park JH. Continuity of care of patient risk in communities study. Arch Intern Med. 1999;159(15):1777–83. with diabetes and its affecting factors in Korea. J Prev Med Public Health. 2007;40(1):51–8. 24. Hong J-S, Kim J-Y, Kang H-C. Continuity of ambulatory care among adult patients with type 2 diabetes and its associated factors in Korea. Health Policy Manag. 2009;19(2):51–70. 25. Son K, Kim Y-S, Hong M-H, Jeong M-A. Analysis of the continuity of outpatient among adult patients with hypertension and its influential factors in Korea. J Korea Academia-Industrial Cooperation Society. 2010; 11(6):2161–8. 26. Gill JM. Can hospitalizations be avoided by having a regular source of care? Fam Med. 1997;29(3):166–71. 27. Menec V, Black C, Roos N, Bogdanovic B. What is the potential for formal patient registration in Canadian primary care? The scale of 'informal registration’ in Manitoba. J Health Serv Res Policy. 2001;6(4):202–6. 28. Cho KH, Kim YS, Nam CM, Kim TH, Kim SJ, Han K-T, Park E-C. The association between continuity of care and all-cause mortality in patients with newly diagnosed obstructive pulmonary disease: a population-based retrospective cohort study, 2005-2012. PLoS One. 2015;10(11):e0141465. 29. Menec VH, Sirski M, Attawar D, Katz A. Does continuity of care with a family physician reduce hospitalizations among older adults? J Health Serv Res Policy. 2006;11(4):196–201. 30. Information CIfH, Services ACoH, McKendry R, Foundation CHSR, Haggerty J: defusing the confusion: concepts and measures of continuity of health care: Canadian Health Services Research Foundation= Fondation canadienne de la recherche Sur les services de santé; 2002. 31. Van Walraven C, Mamdani M, Fang J, Austin PC. Continuity of care and patient outcomes after hospital discharge. J Gen Intern Med. 2004;19(6):624–31. 32. Hong JS, Kang HC. Relationship between oral antihyperglycemic medication adherence and hospitalization, mortality, and healthcare costs in adult ambulatory care patients with type 2 diabetes in South Korea. Med Care. 2011;49(4):378–84. 33. Jee SH, Cabana MD. Indices for continuity of care: a systematic review of the literature. Med Care Res Rev. 2006;63(2):158–88. 34. Breslau N, Reeb KG. Continuity of care in a university-based practice. Acad Med. 1975;50(10):965–9. 35. Liao P-J, Lin Z-Y, Huang J-C, Hsu K-H. The relationship between type 2 diabetic patients’ early medical care–seeking consistency to the same clinician and health care system and their clinical outcomes. Medicine. 2015;94(7):e554. 36. Christakis DA, Feudtner C, Pihoker C, Connell FA. Continuity and quality of care for children with diabetes who are covered by Medicaid. Ambul Pediatr. 2001;1(2):99–103. 37. Al Shahrani A, Baraja M. Patient satisfaction and it's relation to diabetic control in a primary care setting. J Family Med Prim Care. 2014;3(1):5. 38. Worrall G, Knight J. Continuity of care is good for elderly people with diabetes. Can Fam Physician. 2011;57(1):e16–20. 39. Saultz JW, Lochner J. Interpersonal continuity of care and care outcomes: a critical review. Ann Fam Med. 2005;3(2):159–66. 40. Cabana MD, Jee SH. Does continuity of care improve patient outcomes. J Fam Pract. 2004;53(12):974–80. 41. O'keefe JH, et al. Strategies for optimizing glycemic control and cardiovascular prognosis in patients with type 2 diabetes mellitus. Mayo Clin Proc. 2011;86(2). Elsevier, 2011. 42. Aso Y, Hara K, Ozeki N, Yatsuka C, Nakano T, Matsumoto S, Suetsugu M, Nakamachi T, Takebayashi K, Haruki K. Low-dose pioglitazone increases serum high molecular weight adiponectin and improves glycemic control in Japanese patients with poorly controlled type 2 diabetes. Diabetes Res Clin Pract. 2009;85(2):147–52. 43. Evans K, Coresh J, Bash LD, Gary-Webb T, Köttgen A, Carson K, Boulware LE. Race differences in access to health care and disparities in incident chronic kidney disease in the US. Nephrol Dial Transplant. 2010;26(3):899–908.
BMC Nephrology – Springer Journals
Published: Jun 5, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud