Prevalence and factors associated with chronic kidney disease among medical inpatients at the Kenyatta National Hospital, Kenya, 2018: a cross-sectional study

Introduction The burden of chronic kidney disease (CKD) is increasing worldwide. Few studies in low and low-middle income countries have estimated the prevalence of CKD. We aimed to estimate prevalence and factors associated with CKD among medical inpatients at the largest referral hospital in Kenya. Methods We conducted a cross-sectional study among medical inpatients at the Kenyatta National Hospital. We used systematic sampling and collected demographic information, behavioural risk factors, medical history, underlying conditions, laboratory and imaging workup using a structured questionnaire. We estimated glomerular filtration rate (GFR) in ml/min/1.73m2 classified into 5 stages; G1 (≥ 90), G2 (60-89), G3a (45-59), G3b (30-44), G4 (15-29) and G5 (<15, or treated by dialysis/renal transplant). Ethical approval was obtained from Kenyatta National Hospital-University of Nairobi Ethics and Research Committee (KNH-UoN ERC), approval number P510/09/2017. We estimated prevalence of CKD and used logistic regression to determine factors independently associated with CKD diagnosis. Results We interviewed 306 inpatients; median age 40.0 years (IQR 24.0), 162 (52.9%) were male, 155 (50.7%) rural residents. CKD prevalence was 118 patients (38.6%, 95% CI 33.3-44.1); median age 42.5 years (IQR 28.0), 74 (62.7%) were male, 64 (54.2%) rural residents. Respondents with CKD were older than those without (difference 4.4 years, 95% CI 3.7-8.4 years, P = 0.032). Fifty-six (47.5%) of the patients had either stage G1 or G2, 17 (14.4%) had end-stage renal disease; 64 (54.2%) had haemoglobin below 10g/dl while 33 (28.0%) had sodium levels below 135 mmol/l. ). History of unexplained anaemia (aOR 1.80, 95% CI 1.02-3.19), proteinuria (aOR 5.16, 95% CI 2.09-12.74), hematuria (aOR 7.68, 95% CI 2.37-24.86); hypertension (aOR 2.71, 95% CI 1.53-4.80) and herbal medications use (aOR 1.97, 95% CI 1.07-3.64) were independently associated with CKD. Conclusion Burden of CKD was high among this inpatient population. Haematuria and proteinuria can aid CKD diagnosis. Public awareness on health hazards of herbal medication use is necessary.


Introduction
Chronic kidney disease (CKD) is defined as decreased glomerular filtration rate (GFR) of less than 60 mL/min per 1.73 m 2 for more than 3 months, with or without kidney damage; or functional or structural abnormalities of the kidneys with or without decreased GFR [1,2]. It is classified into five stages, with the severest form being end-stage renal disease which requires renal replacement therapy either in the form of dialysis or renal transplantation [2]. The main risk factors for CKD globally are diabetes mellitus, hypertension and glomerulonephritis; other associations include genetics, family history, gender, increasing age smoking and nephrotoxins [3][4][5]. The global burden of chronic kidney disease is estimated to be 11 to 13% [6]. The prevalence of renal disease in Africa is not known, though estimates point to a substantial burden especially in the middle-aged [7]. Prevalence of CKD in specific health conditions has previously been estimated in Kenya, including in HIV, rheumatoid arthritis, heart failure and type 2 diabetes [8][9][10][11][12]. However, little is known on overall prevalence of CKD in the Kenyan population; such data would be invaluable in informing public health investment in treatment facilities and prevention. The gold standard for determining CKD burden is population surveys; however these are difficult due to time and financial constraints. Prevalence of CKD in medical inpatients at tertiary/referral health facilities has been used in several studies in different countries and settings to estimate the overall disease burden and its complications. These include Uganda, Botswana and the United States [13][14][15]. Our study aimed to estimate the prevalence and identify factors associated with CKD among medical inpatients at the largest tertiary hospital in Kenya.

Methods
Study design: this was a cross-sectional study among medical inpatients at a tertiary health facility.

CKD diagnosis
A diagnosis of CKD was defined as presence of decreased glomerular filtration rate (GFR) below 60 mL/min per 1.73m 2 , or GFR above 60 mL/min per 1.73 m 2 but with markers of renal damage, as determined by the Chronic Kidney Disease Epidemiology collaboration (CKD-Epi) equation [16] and at least one of the following: 1) Contracted kidneys, hypo-echoic kidneys or loss of corticomedullary differentiation on renal imaging; 2) Serum phosphate levels above 1.4mmol/l; 3) Serum calcium level below 2.2mmol/l; 4) History of kidney transplantation; 5) History of documented kidney dysfunction for more than 3 months; 6) Anaemia of chronic disease (normochromic, normocytic, hypoproliferative picture), with haemoglobin level of <10/dl [17].

Sample size assumptions and calculations
The sample size was calculated using the Cochrane formula [18], with the following assumptions: Z statistic of 1.96 for 95% confidence level, expected prevalence of CKD of 27% in a study done in similar settings and a precision of 0.05. The minimum sample size required was 303.

Sampling methods
A pre-visit was conducted before the commencement of the study to ascertain the patients on admission per ward at the time of study. A number between the first and the K th patient was selected as the first participant to be sampled and thereafter sampled very K th patient. This was repeated till the required sample size from that ward was achieved; the procedure was then repeated in all the eight wards. Sampling with replacement was carried out by obtaining consent from the next sequential inpatient meeting the case definition in case the one selected opted out of the interview.

Data collection
Data on socio-demographic information, behavioural, family and medical history were collected using structured paper questionnaires administered through face to face interviewing of the study participants. The questionnaire was pretested on 30 patients selected from the general medical outpatient clinic. Data on laboratory and imaging investigations were collected from the inpatient medical files.

Data management and statistical analysis
The data from the paper questionnaires were entered into a computer, cleaned, coded and loaded it into statistical software for analysis. We derived means for continuous variables and proportions to describe the socio-demographic characteristics of the study participants as well as the prevalence of CKD. After calculating GFR by the CKD-Epi equation to diagnose CKD, the patients were classified into five stages using GFR in ml/min/1. Factors that became insignificant at 0.05 level when first introduced into the model were removed; the model was run several times till the best fit was achieved.

Ethical considerations
We obtained written informed consent form every study participant, removed personal identifiers and maintained strict confidentiality of the collected data using lockable cabinets for paper questionnaires and password-protected computer for the electronic database.

Socio-demographic characteristics
We interviewed 306 medical inpatients, originating from 33 of the 47 severe forms are higher in men [19]. One suggested mechanism is role of testosterone and protective function of oestrogen in women [20]. Majority of the CKD cases were middle-aged, and were significantly older than their non-CKD counterparts. Older age is also a recognized risk factor for CKD [5,21]. One explanation is that renal function generally decreases with age; hence older individuals are more prone to CKD after renal injury. In our study, most of the cases were rural residents. A systematic review and meta-analysis of studies on CKD in Sub-Saharan Africa did not find any difference in prevalence between rural and urban populations [22].

Study limitations
Some signs (blood/proteins in urine) could have occurred after CKD process. To ensure we captured only symptoms experienced prior to diagnosis, we corroborated the information from the interviews with the medical history captured in the inpatient records. Interpretation of our results does not intend to infer causality; rather the association between the two early symptoms and CKD may have a role in early diagnosis. There was also possibility of misclassifying AKI as CKD; however, our study only considered those with reduced GFR and presence of at least one other feature to help differentiate it from AKI, we believe this increased the specificity of the diagnostic criteria.

Conclusion
There was high burden of CKD in medical inpatients at this Kenyan What is known about this topic  Prevalence of CKD in various medical conditions, including hypertension, diabetes type 2 and HIV;  CKD risk factor prevalence in the Kenyan general population is known.

What this study adds
 An estimate of the population CKD burden, in the absence of population-based survey data;  Associations amenable to public health intervention.

Competing interests
The authors declare no competing interests.

Authors' contributions
VM conceived the study idea, developed the protocol, carried out the study and prepared the first draft of the manuscript; JG, GG and WOO provided guidance from protocol development to study completion, BMW offered guidance on focusing the research question and reviewed the manuscript, JG provided reviews during manuscript drafting, ZG provided technical advice for conduction of the study and arranged the financial support for the study. All the authors approved the final version of the manuscript.

Acknowledgments
We wish to thank the head of clinical medicine department, nursing and medical staff at the Kenyatta National Hospital for their support during the study. We thank the patients who took time to respond to our interviews. We also thank Vincent Were for his assistance in statistical analysis. Last, we thank the Kenya Ministry of Health and the US Centers for Disease Control and Prevention for financial support for the study. Table 1: socio-demographic characteristics of the selected medical inpatients, Kenyatta National Hospital, Kenya, 2018 (n=306)