A hospital-based cross-sectional study was conducted at Ayder Comprehensive Specialized Hospital from 1 December 2023 to 4 April 2024. The hospital is located in Mekelle, in the Tigray Region, 783 km north of Addis Ababa, Ethiopia's capital. The hospital has more than 500 inpatient beds with 36 Intensive care units and more than 3,600 employees in 2022. It offers a wide range of medical treatments, such as dialysis, cancer treatment, and neonatal care, to both inpatients and outpatients of all ages. In addition, it acts as a research and teaching hospital for medicine and other health sciences. During the time of data collection, there were about 2,300 diabetic patients receiving care at the hospital's diabetic clinic.

The study included adult patients (≥18 years) who had been previously diagnosed with diabetes mellitus and attended the hospital's follow-up diabetes clinic during the study period. Patients with diabetes who were pregnant, critically ill, had incomplete clinical and laboratory data, or had a short follow-up period (<6 months) were not included in the study.

A minimum sample size of 328 was determined using a single population proportion formula, n = 1/d² × [(Z_α/2)² × P (1 − P)], where n is the sample size, Z = 1.96 [95% confidence interval (CI)], P = 0.263, the proportion of chronic kidney disease among DM patients from a previous study (15), and d is the assumed margin of error (5%), and a 10% non-response rate was considering. Sample size was also calculated for selected variables and found to be below 328.

According to the hospital report, 2,300 diabetic patients were in regular follow-up at the diabetic clinic. By taking their medical registration book as a sampling frame, interval size (k) was calculated by dividing the total number of diabetic patients on a follow-up by the desired sample size (i.e., k = 2,300/328 = 7). By using a systematic random sampling method, the participants were selected every 7 intervals from the list of medical registration book. If the kth or selected patients were not willing to participate in the study or did not fulfill the inclusion criteria, the next diabetic patient was recruited to the study.

Data on sociodemographic, economic, and behavioral/lifestyle factors were collected using pretested, structured questionnaires. Medical records of the patients were also reviewed for clinical history including types of DM, blood glucose levels, and a history of hypertension. The average blood pressure and blood glucose over the previous 3 months were utilized to calculate the participants’ blood pressure and glycemic status.

Blood pressure measurements were carried out using a manual sphygmomanometer after patients had been comfortably resting for at least 5 min. About 5 ml of venous blood specimens were drawn and processed, and the serum was separated for biochemical analysis. Serum creatinine analysis was carried out using kinetic alkaline picrate method on the Cobas C311 clinical chemistry analyzer. Participants with abnormal creatinine measurements were checked again for persistence after 3 months.

Fresh 10 ml urine specimen was collected from each participant to detect proteinuria using dipsticks. Participants with positive proteinuria were reexamined for persistence after 3 months.

The eGFR was calculated by using the CKD-EPI 2021 equation (16):eGFR=142×min(Scr/k,1)α×max(Scr/k,1)−1.200×0.9938age×1.012(iffemale)where GFR is the glomerular filtration rate (ml/min/1.73 m²), min is the minimum of Scr/k or 1, max is the maximum of Scr/k or 1, Scr is the serum creatinine level (in mg/dl), k = 0.7 (females) or 0.9 (males), and α = −0.241 (females) and −0.302 (males).

Data collectors and a supervisor received 1-day training on data collection instruments and study objectives to guarantee that the data quality was not compromised. Questionnaire was pretested using 5% of the total participants (17 patients) at the Mekelle General Hospital before data collection, and appropriate modifications were made. Close follow-up and supervision were carried out during the data collection period jointly by the first author and the supervisor. The collected data were reviewed and checked for its completeness before data entry. For creatinine analysis, the reaction was monitored every working shift by using both normal and pathological control samples. Quality control of the urine dipstick testing was also done in accordance with the standard operating procedure.

The data were entered into SPSS version 29 for analysis. A descriptive statistic was used to describe the study participants and relevant variables. The associations between chronic kidney disease and all above mentioned variables were evaluated in bivariable logistic regressions. Those variables that showed a p-value of 0.25 or smaller in these bivariable regressions were further included in a multivariable logistic regression model. The final multivariable model only included variables that were significantly associated based on p-values ≤0.05. Adjusted odds ratios (AORs) and 95% CIs are reported to measure the strength of the associations. Hosmer and Lemeshow goodness of fit was typically utilized to assess model fitness.

A total of 328 adult diabetic patients participated in the study. Of them, 169 (51.5%) were in the age group of ≤45 years old, more than half [191 (58.2%)] were female, 243 (74.1%) were urban dwellers, 289 (88.1%) were ethnic Tigrayans, 241 (73.5%) were married, 125 (38.1%) had primary education, and 205 (62.5%) had less than 3,000 Ethiopian birr in monthly income (Table 1).

The proportion of patients with type 2 and type 1 diabetes mellitus was 181 (55.2%) and 147 (44.8%), respectively. Among the study participants, 97 (29.6%) had hypertension. The majority of the participants [200 (61%)] were classified as having normal BMI and 35 (10.7%) had family history of kidney disease.

Furthermore, 89 (27.1%) had a long duration of diabetes (>10 years), almost all [325 (99.1%)] never smoked, and 308 (93.9%) had no alcohol consumption habit. The majority [314 (95.7%)] had never engaged in regular physical exercise. About half of the participants 165 (50.3%) had normal systolic blood pressure (<120 mmHg), and the majority [239 (72.9%)] had normal diastolic blood pressure (60–79 mmHg) (Table 2).

The overall prevalence of CKD was 26.5% (95% CI: 21.8%–31.7%), of which, 29 (8.8%) had eGFR <60 ml/min/1.73 m² and 69 (21%) had proteinuria. From the total patients with CKD (87), 17 (5.2%), 41 (12.5%), 24 (7.3%), 3 (0.9%), and 2 (0.6%) were categorized as stage 1–5, respectively.

From the total patients with CKD (87), 11 (13%) had both impaired GFR (<60 ml/min/1.73 m²) and proteinuria, 13 (14%) had only impaired GFR (eGFR < 60 ml/min/1.73 m²), and 63 (74%) had only proteinuria.

When it was analyzed with multivariate logistic regression, hypertension, >10 years duration of diabetes, and family history of kidney disease were independently associated factors for chronic kidney disease (p < 0.05).

Patients with concomitant hypertension were 2.390 (AOR = 2.390; 95% CI, 1.394–4.099; p = 0.002) times more likely to develop CKD compared with non-hypertensive patients. The odds of CKD was 2.585 (AOR = 2.585; 95% CI, 2.321–5.807; p = 0.001) times higher among patients who have had a longer (>10 years) duration of diabetes compared with patients having a short duration. Patients with a family history of kidney diseases were 2.884 (AOR = 2.884; 95% CI, 1.338–6.218; p = 0.007) times more likely to have CKD than having no family history of kidney diseases (Table 3).

Our study had a number of limitations. First, since it was a facility-based study, it might not reflect the true prevalence of CKD in patients with diabetes in the community. Second, we used a cross-sectional study design. Therefore, we could not establish any causal relationship between CKD and its risk factors. This study, therefore, proposes that a longitudinal study with representative samples from health facilities and communities should be conducted to provide up-to-date data for policymakers and program planners. Furthermore, we considered a descriptive multivariable modeling approach without considering the causal dependencies between the variables in the model. Third, even though albumin creatinine ratio is a stronger indicator of CKD, we assessed proteinuria using semiquantitative methods (dipstick), which may affect the reliability of the results due to its low specificity and sensitivity. Another limitation of our study is that we used average blood glucose over 3 months, and this might not represent the true glycemic status. HbA1c would have been a better option to assess glycemic control. Hence, the results should be interpreted with caution.