Prevalence and Risk Factors of Chronic Kidney Disease among Type 2 Diabetes Patients: A Cross-Sectional Study in Primary Care Practice

This cross-sectional study aimed to investigate the prevalence and risk factors of chronic kidney disease (CKD) among 1,096 primary care type 2 diabetes (T2DM) patients in northern Thailand between October 2016 and September 2017. CKD was defined as estimated glomerular rate filtration values of <60 mL/min/1.73 m2. Prevalence with confidence intervals across CKD advanced stages 3–5 were estimated. Factors associated with CKD were evaluated by multivariate logistic regression. The overall prevalence of CKD was 24.4% (21.9–27.0), with severities of 11.4% (9.7–13.4), 6.8% (5.5–8.5), 4.6% (3.5–6.0), and 1.6% (1.0–2.5) for stages 3 A, 3B, 4, and 5, respectively. Regarding age and glycaemic control, individuals older than 75 years and those with a haemoglobin A1c ≥ 8% had the highest prevalence of 61.3% (51.7–70.1) and 38.6% (34.3–43.2), respectively. The multivariable logistic regression model explained 87.3% of the probability of CKD. The six independent significant risk factors of CKD were older age, retinopathy, albuminuria, haemoglobin A1c ≥ 7%, anaemia, and uric acid>7.5 mg/dL. A relatively high prevalence of CKD, especially in older patients and those with diabetic complications-related to poor glycaemic control, was encountered in this primary care practice. Early identification may help to target optimise care and prevention programs for CKD among T2DM patients.

The multivariate logistic regression models identified six independent significant risk factors of CKD (  Fig. S2).
For risk factors associated with CKD, using the multiple imputation analysis, restricting the analysis by excluding patients with hyperfiltration (eGFR ≥120 mL/min/1.73 m 2 ), and re-analysed risk factors of CKD using the proposed different eGFR equations did not alter the risk factors model (c-statistic, 0.87-0.88; Supplementary  Tables S5, S6).

Discussion
This study examined the burden of CKD in adult T2DM patients in a suburban community in Thailand. We found that CKD is a common diabetes-related complication among T2DM patients. Within a primary care setting, the estimated prevalence of CKD stages 3-5 (eGFR <60 mL/min/1.73 m 2 ) in T2DM patients was 24.4% (95% CI, 21.9-27.0), with substantial variation by age and glycaemic control status. From a clinical perspective, risk factors for the development of CKD in our study can help inform the clinical decision-making process and the formation of the appropriate care strategy for T2DM patients. As such, our study can lay the foundation for routine surveillance for T2DM patients who are at high risk of CKD in the primary care setting.
The treatment of diabetes generally differs by CKD status because individuals without CKD are treated with oral antidiabetic drugs, while those with CKD receive insulin therapy. According to strategies targeting kidney-specific disease, T2DM patients in our study were more commonly prescribed renin-angiotensin system (RAS) inhibitors (59.0%), whereas the utilisation of these agents varied across diabetes care practices worldwide as 29.6-56.0% [22][23][24][25] . Despite an improvement in diabetes care over time, suboptimal glycaemic control remains observed in our study, with only 36.1% meeting the glycaemic goal of haemoglobin A1c < 7%, particularly those with CKD. We also found that T2DM patients with CKD were more likely to have diabetes-related complications including ischaemic heart disease, cerebrovascular disease, diabetic retinopathy, and albuminuria than those without CKD. Taken together, these figures are in line with previous nationwide reports in Thailand 26 .
Recently, large randomised controlled trials suggest that the use of sodium-glucose cotransporter 2 (SGLT-2) inhibitors or glucagon-like peptide 1 (GLP-1) receptor agonists shown to reduce the risk of CKD progression and improve kidney outcomes [27][28][29][30] . However, during the study period, the novelty of the new drug class of SGLT-2 inhibitors and GLP-1 receptor agonists were not available in the National Medicines Formulary in Thailand under the health benefits package. As such, further studies are needed on treatments modifying the risk of development of CKD among T2DM in the real-world primary care settings.
To our knowledge, our finding suggests a lower prevalence and is comparable to a national study of CKD in adult T2DM patients in Thailand found at 24.4% vs. 35.4%, respectively 17 . A similar trend in the prevalence of CKD was observed in elderly patients (>65 years) with T2DM-at 40.5% and 56.1% in our study and national level in Thailand, respectively 31 . Unlike urbanised areas, CKD rates among the T2DM patients in our study were comparable to those reported in previous studies of less urbanised communities or regional areas in Thailand 19,20 . According to the Global Burden Disease-CKD study, CKD due to diabetes accounted for 30.7% of CKD populations, in which T2DM was the only cause of CKD to illustration a substantial increase in the age-standardised rate (changed by 9.5% from 1990 to 2017) 32 . Globally, the overall prevalence of CKD among T2DM patients varied at 6.0-39.3% (our result found at 24.4%) 17,25,[33][34][35][36][37][38][39][40][41][42][43][44][45][46][47] . These discrepancies across different settings may be attributed to the variations in diagnostic methods used and ethnicities such as the black race, which is associated with a greater rate of GFR decline 48 . Overall, our result parallels the global rates of diabetes populations, which are expected to occur lower than in the rural areas or less urbanised community 1 , suggesting that our findings have general relevance. www.nature.com/scientificreports www.nature.com/scientificreports/ In this study, several diabetes-specific and general risk factors in the literature for CKD among T2DM patients were investigated (Supplementary Table S1). However, we did not find an association between hypertension or blood pressure and the risk of CKD among T2DM patients, which was reported in previous studies 18,21,[39][40][41][42] . The lack of this relationship could be attributable to in part to increasing usage of RAS inhibitors for protection against kidney disease and improved blood pressure control in our diabetes practice. Moreover, since most of our study patients were already receiving antihypertensive agents, the lack of association between blood pressure and the risk of CKD is not surprising. Consequently, six independent significant risk factors of CKD were identified including older age (>55 years), retinopathy, albuminuria, haemoglobin A1c ≥ 7%, anaemia (haemoglobin <12 g/dL in females or <13 g/dL in males), and uric acid>7.5 mg/dL.
With respect to non-modifiable risk factors, managing elderly patients with T2DM is challenging, as this population has a high rate of comorbid conditions as also associated with a greater risk of developing CKD. Our findings showed that T2DM patients aged 56-65, 66-75, and>75 years had more than 2.8-fold, 5.4-fold, and 27.4-fold higher adjusted ORs for CKD, respectively. This result reaffirms that of previous studies that older age was associated with a higher risk of CKD among T2DM patients 33,37,[39][40][41][42]44 . Cardiovascular disease, obesity, and multimorbidity via endothelial cell dysfunction and sympathetic nervous system activation resulting in increased atherosclerosis, hypertension, and progressive nephrosclerosis are believed to explain the mechanisms underlying older age and the risk of CKD 49,50 .
With respect to modifiable risk factors, glycaemic control was the most determinant of the development of diabetes-related complications and the risk of CKD in T2DM. Based on our findings, the presence of albuminuria, diabetic retinopathy, and poor glycaemic control (haemoglobin A1c ≥ 7%) are independent risk factors for   the development of CKD among T2DM patients. Indeed, albuminuria and diabetic retinopathy are components of diabetes-related microvascular complications, especially in those with poor glycaemic control. These factors have been previously recognised as risk factors for the development of CKD in T2DM patients 18,37,39,41,42 . In concordance with previous reports 42,51 , our study demonstrates that anaemia, defined as haemoglobin <12 g/dL in females or <13 g/dL in males, commonly occurs in T2DM patients (38.5%), particularly in the elderly and those with more comorbid conditions. As expected, a significant association was observed that T2DM patients with anaemia had more than a 3.0-fold higher risk of CKD. Our finding corresponds well with previous studies that hyperuricemia is a strong independent risk factor of the development of CKD [52][53][54][55] . Evidence illustrates that the GFR deterioration is associated with progressive impairment in uric acid excretion, resulting in insulin resistance and hypertension. Experimental studies also revealed that increased serum uric acid concentrations are associated with kidney damage via stimulating RAS activity and promoting endothelial damage along with oxidative stress [56][57][58] . This study was based on patient-level information by the retrieval and linking of routinely collecting data, which provide detailed primary care practice on diabetes and kidney care. Our study delivers previously   www.nature.com/scientificreports www.nature.com/scientificreports/ unrecognised data on the prevalence and risk factors of CKD among T2DM in a suburban community through a comprehensive process and rigorous statistical approaches. Moreover, the consistency of findings was observed based on our set of sensitivity analyses.
However, our findings should be interpreted in the context of certain limitations. First, the causal inference and the chronicity of the observations must be considered because our findings were based on the observational cross-sectional nature of the analyses. Moreover, longitudinal data were not obtained in this study; thus, temporal trends in prevalence and dynamic risk prediction for CKD among T2DM patients cannot be established over time. Second, this study was conducted within a single centre and was limited by the unique organisation of the Sansai Hospital, the suburban community care protocol implemented throughout the primary care unit and village health volunteers of this community. Accordingly, the generalisability of our finding to other T2DM populations and healthcare settings other than in primary care practice in Thailand is uncertain and warrants further study. Third, although we performed a series of sensitivity analyses using different equations for estimating GFR, misclassification (potential errors relating to CKD staging) is possible because eGFR alone is insufficient to evaluate kidney function, particularly in cases of advanced CKD. Moreover, urinary protein tests were not routinely available in our primary care practice. Therefore, detection bias should be noticed as it was not considered in our definition of CKD. Finally, contextual factors related to diabetes control including, patient comorbidities, health behaviours (e.g. dietary intake and physical activity), mental health problems (e.g. depression, social support, and coping skills), and social determinants of health (education and literacy, income and social status, physical environments, employment status, and health inequity) were obtained. Moreover, novel biomarkers and relevant inflammatory markers were not available in our primary diabetic care practices. In this circumstance, the residual risk factors may also influence the prevalence and risk factors of CKD among T2DM patients. However, the risk factors for development CKD in our study illustrated an excellent performance of the model prediction in terms of discriminative ability, which explained 87.3% of the probability of CKD among T2DM patients.
Due to rapid urbanisation and the dramatic increase in the elderly population, our findings support the well-recognised fact that routine surveillance is mandatory to prevent the development of ESRD to decrease the healthcare burden and costs-related to RRT treatment. This study may also contribute to improved diabetes care management by the early identification and targeting of T2DM patients who are at high risk of developing CKD. Further studies are needed to assess the utility of integrating the clinical predictive factors of CKD among T2DM patients as a part of routine diabetes care and call for strategic goals and actions upon their recognition to reduce the CKD incidence or slow CKD progression. Ultimately, long-term holistic healthcare services in a primary care practice should be targeted based on multimorbidity concepts, particularly in the elderly, to reduce the prevalence of CKD and mitigate the large public health effect of CKD in T2DM patients.
In summary, here we found a relatively high prevalence of CKD among T2DM patients in a suburban community in Thailand, particularly in elderly patients and those with diabetes complications related to poor glycaemic control. Our study also underscores an important opportunity to identify T2DM patients who are at high risk of CKD through readily available and routinely obtained factors in the primary care setting. Early identification may help optimise care and prevention programs for these populations.

Methods
Study design and patient population. This retrospective cross-sectional study used a cohort of T2DM patients from the Sansai Hospital suburban community in northern Thailand from October 1, 2016 through September 30, 2017. Data were obtained from the electronic health records from the Sansai Hospital database along with the routine medical records from the Sansai primary care settings. The datasets were linked and merged comprising: (i) outpatient and inpatient data; (ii) administrative data on pharmacy dispensing and laboratory support system; (iii) primary care practice on diabetes and kidney care, with patient-level detail on socio-demographic factors, clinical characteristics, and routine diabetes clinical examination findings. An external consensus panel of two health information professionals reviewed, verified, and validated the datasets for high-quality data collection system and to limit the quantity of missing data.
The study was approved by the institutional review boards of Chiang Mai Provincial Public Health Office and the Hospital Authority of the Sansai Hospital. Informed consent in this study was waived owing to the retrospective nature of our study and de-identification of the patient information which has been accepted and allowed by the institutional review boards of Chiang Mai Provincial Public Health Office. The study protocol was conducted according to the Declaration of Helsinki and reported in line with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement guidelines for cross-sectional studies (Appendix in the Supplement) 59 .
Patients eligible for inclusion in this study included those: (i) who were aged 18 years or older; (ii) who were diagnosed with T2DM; (iii) for whom eGFR values had been documented more than twice from their index date. The exclusion criteria were: (i) having received chronic dialysis treatment or kidney transplantation; (ii) incomplete data on glycaemic control; and (iii) currently pregnancy or breastfeeding.
Outcome: Kidney function measures. According to the Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group, serum creatinine measurements were used to calculate eGFR based on the CKD-EPI equation 60,61 . Theoretically, eGFR alone is insufficient to indicate the presence of CKD, particularly in less advanced stages. Thus, we considered only those with advanced CKD stages (stages [3][4][5]: eGFR values of <60 mL/min/1.73 m 2 on more than days apart from their index date. Patients were classified as having CKD stage 3 A, 3B, 4, or 5 if they had an eGFR value of 45-59, 30-44, 15-29, or <15 mL/min/1.73 m 2 , respectively, since urinary protein tests were not available in the primary care practice. However, to be comprehensive and reflect routine clinical practice, patients for whom urine dipstick measurements or urine albumin-creatinine ratios (UACR) were available were evaluated for albuminuria (≥1+ in dipstick studies or>30 mg/g in UACR studies). www.nature.com/scientificreports www.nature.com/scientificreports/ Candidate risk factors. Potential risk factors of CKD among T2DM patients were recognised based on a comprehensive review and the list of risk factors that are routinely and readily available at primary care practice 62,63 . Candidate risk factors included: (i) socio-demographic characteristics (age, sex, BMI, smoking status, alcohol consumption, insurance status, medical history [hypertension, coronary artery disease, cerebrovascular disease, retinopathy, and albuminuria], blood pressure, and duration of diabetes); (ii) laboratory values (serum creatinine, fasting plasma glucose, haemoglobin A1c, haemoglobin, uric acid, and lipid profiles); and (iii) medication treatment (glycaemic control, antihypertensive agents, lipid-lowering agents, antiplatelet agents, and anti-gout agents).
Statistical analysis. Descriptive data are summarised as the number with percentage for categorical variables and mean ± SD or medians with IQR as appropriate. Difference between CKD status (eGFR<60 vs. ≥60 mL/ min/1.73 m 2 ) were assessed using Fisher's exact test and unpaired t-test or Wilcoxon rank-sum test for categorical and continuous data, respectively.
Prevalence rates estimated with 95% CIs of CKD (eGFR <60 mL/min/1.73 m 2 ), as well as CKD stage (stage 3-5), were analysed according to sociodemographic (age and sex) and glycaemic control (haemoglobin A1c). To identify the candidate risk factors, the crude association between patient characteristics and CKD was assessed through the univariable logistic regression models. Subsequently, risk factors with a P-value less than 0.100 were then included in the multivariate logistic regression analysis with the stepwise backward method. The final model was also determined for multicollinearity by investigation of the variance inflation factors of the risk factors within the multivariable model.
The effect estimates of final risk factors model for CKD among T2DM patients were expressed as ORs with corresponding 95% CI. Moreover, the c-statistic or the AuROC curve was performed to indicate the ability of a final model to distinguish patients with or without CKD (eGFR <60 vs. ≥60 mL/min/1.73 m 2 ). A c-statistic more than 0.7 indicate acceptable discriminative of the model 64 . Variables with more than 20% of the values were excluded from the primary analysis; however, a multiple imputation method was performed in the sensitivity analysis. All analyses were performed using Stata version 14.0 (StataCorp LP, TX). Two-tailed tests with values of P < 0.05 were considered statistically significant. Sensitivity analyses. Additional analyses were further assessed to address the robustness of our findings.
For the prevalence of CKD, the different equations for estimating GFR < 60 mL/min/1.73 m 2 were performed using the CKD-EPI equation for Asian population 65 , MDRD equation 66 , and Thai GFR equation 67 The agreement of prevalence of CKD using the CKD-EPI equation and other proposed study equations for estimating GFR was estimated using the κ statistic (>0.8 indicates almost perfect agreement) 68 .
For risk factors associated with CKD, sensitivity analyses were conducted by (i) using the multiple imputation analysis to account for missing values; (ii) restricting the analysis by excluding patients with hyperfiltration (eGFR ≥120 mL/min/1.73 m 2 ) that may contribute to the progression of kidney disease among diabetes patients 69 ; and (iii) re-analysing the risk factors of CKD ( < 60 mL/min/1.73 m 2 ) using the three different eGFR equations as described above.

Data availability
Data are available from the authors upon reasonable request and with permission of the Hospital Authority of the Sansai Hospital, Chiang Mai Province, Thailand.