Predictors of one-year mortality in hemodialysis patients with end-stage renal disease at a Kenyan county hospital

Abstract

BACKGROUND

In Kenya, end-stage renal disease (ESRD) poses a significant public health burden. While hemodialysis is primarily delivered through county hospitals, comprehensive outcome data from these routine clinical settings remain scarce.

AIM

To evaluate one-year clinical outcomes and identify independent predictors of mortality among ESRD patients undergoing hemodialysis at a Kenyan county hospital.

METHODS

We conducted a retrospective cohort study of all patients who initiated hemodialysis for ESRD at Murang’a County Referral Hospital between January 2024 and January 2025. Data on demographics, clinical characteristics, comorbidities, and treatment parameters were extracted from electronic medical records and dialysis unit logs. Cox proportional hazards regression was used to identify factors associated with one-year mortality.

RESULTS

Among 79 patients analyzed [median age 62.0 years, interquartile range (IQR) 48.0-74.0; 65.8% male], the cumulative all-cause mortality rate during the study period was 34.2% (27/79). The cohort demonstrated a heavy reliance on central venous catheters (CVCs) (89.9%, 71/79) rather than arteriovenous fistulas (10.1%, 8/79). Non-survivors were significantly older (median 73.0 years vs 58.0 years, P < 0.001) and had lower baseline hemoglobin (7.1 g/dL vs 8.6 g/dL, P = 0.008). In multivariable analysis, older age [adjusted hazard ratio (aHR) = 1.05 per year, 95%CI: 1.01-1.09, P = 0.012] and CVC use (aHR = 3.12, 95%CI: 1.08-9.01, P = 0.036) remained independent predictors of mortality. While lower baseline hemoglobin and estimated glomerular filtration rate were significant in univariate analysis, they lost significance after adjustment. Comorbidities, including human immunodeficiency virus and diabetes, were not statistically significant predictors.

CONCLUSION

Mortality among hemodialysis patients in this Kenyan cohort is high, with older age and CVC use strongly associated with poorer survival. The near-universal reliance on CVCs highlights systemic challenges in pre-dialysis care, underscoring the urgent need for robust vascular access programs and early intervention strategies to improve patient outcomes.

Key Words: End-stage renal disease; Hemodialysis; Central venous catheter; Kenya; Predictors; Vascular access

Core Tip: This retrospective cohort study from a Kenyan county hospital identifies a high one-year mortality rate (34.2%) among incident hemodialysis patients. Advanced age and central venous catheter (CVC) use were the only independent predictors of mortality. With CVC use exceeding 89%, our findings reveal a critical “catheter-first” systemic failure. This underscores an urgent need for decentralized vascular access programs and earlier nephrology referrals to shift toward arteriovenous fistulas and improve patient survival in resource-limited settings.

INTRODUCTION

End-stage renal disease (ESRD) represents the terminal stage of chronic kidney disease (CKD), characterized by a glomerular filtration rate (GFR) of less than 15 mL/minute/1.73 m² and the requirement for lifelong renal replacement therapy (RRT) for survival. ESRD poses a significant public health challenge in Kenya, where the rising prevalence of hypertension and diabetes mellitus has driven an increasing demand for RRT[1-4]. While national data on ESRD incidence remain limited, regional studies indicate a growing burden aligned with the high prevalence of its primary drivers: Hypertension (estimated at 25%-30% in adults) and diabetes (3%-5%)[3,4]. Consequently, the demand for hemodialysis, the primary life-sustaining treatment in the region, has risen sharply. Despite its crucial role, outcomes remain concerning, with mortality rates significantly exceeding those reported in high-income countries[5,6].

Globally, the type of vascular access is a well-established determinant of mortality in hemodialysis populations[7]. A recent systematic review and meta-analysis confirmed that central venous catheter (CVC) use is consistently associated with significantly higher all-cause and infection-related mortality compared to arteriovenous fistulas (AVFs)[8].

While this association is well-documented in high-income settings, comprehensive data on clinical outcomes and survival determinants from routine care settings in sub-Saharan Africa remain scarce. A systematic review on mortality prediction in CKD highlighted that most evidence is derived from non-African cohorts, underscoring the urgent need for region-specific data[9].

In Kenya’s decentralized healthcare system, county-level facilities, such as Murang’a County Referral Hospital, form the backbone of ESRD care. The management of ESRD in these resource-limited settings faces multifaceted challenges, including late clinical presentation, limited dialysis capacity, high treatment costs, and a significant comorbidity burden[10]. A critical barrier is the underutilization of the AVF-the “gold standard” for vascular access. Studies from Kenyan national referral hospitals report a predominance of CVCs at dialysis initiation[11,12], a pattern of late presentation and emergent catheter use replicated across the region[13].

Understanding local predictors of poor outcomes is essential for developing targeted interventions and optimizing resource allocation to improve patient survival. This study aimed to evaluate one-year clinical outcomes and identify independent predictors of mortality in a cohort of hemodialysis patients at Murang’a County Referral Hospital. Furthermore, we sought to quantify the burden of catheter-dependent dialysis in a typical county-level facility to provide evidence that can inform clinical practice in similar resource-limited settings.

MATERIALS AND METHODS

Study design and measurements

We conducted a retrospective cohort study of all adult patients (≥ 18 years) with ESRD who initiated maintenance hemodialysis at Murang’a County Referral Hospital between January 2024 and January 2025. The primary outcome was all-cause mortality at one year following dialysis initiation. Key exposure variables included demographic characteristics, comorbidities, laboratory values, and vascular access type (AVF vs CVC).

Data sources, recruitment and collection

Patients were identified through the hospital’s electronic health record system and dialysis unit records. Initial data extraction identified 132 patients who visited the Renal Dialysis Unit during the study period.

Inclusion criteria: (1) Age ≥ 18 years; (2) Confirmed diagnosis of ESRD [estimated glomerular filtration rate (eGFR) < 15 mL/minute/1.73 m²] requiring maintenance hemodialysis; (3) Initiation of hemodialysis at the study site between January 2024 and January 2025; and (4) Availability of key clinical data (age, sex, ESRD diagnosis, vascular access type, and vital status).

Exclusion criteria: (1) Duplicate records; (2) Acute kidney injury requiring temporary dialysis; and (3) Insufficient clinical documentation (missing > 50% of defined key variables).

After screening, we excluded 24 duplicate entries, 19 patients who did not meet ESRD diagnostic criteria, and 10 patients with insufficient clinical documentation. The final analytical cohort comprised 79 patients with confirmed ESRD and complete treatment records. A formal sample size calculation was not performed a priori; instead, we included all consecutive patients meeting the criteria, representing the total incident hemodialysis population at this center during the study period.

Data were extracted by two trained physicians using a standardized, pre-piloted electronic collection form. The eGFR was calculated using the CKD Epidemiology Collaboration equation. Collected variables included:

Demographics: Age and sex.

Clinical characteristics: Primary cause of ESRD and comorbid conditions.

Baseline laboratory values: Hemoglobin and eGFR.

Treatment parameters: Vascular access type and the number of hemodialysis sessions completed.

Statistical analysis

Statistical analyses were performed using R software (version 4.3.1). The normality of continuous variables was assessed using the Shapiro-Wilk test. Because continuous variables (age, hemoglobin, eGFR) were non-normally distributed (P < 0.05), they are presented as medians with interquartile ranges (IQRs). Categorical variables are presented as counts and percentages. Between-group comparisons were performed using Mann-Whitney U tests for continuous variables and Fisher’s exact tests for categorical variables. Survival analysis was conducted using Kaplan-Meier curves and log-rank tests.

Univariate and multivariate Cox proportional hazards regression models were used to identify factors associated with one-year mortality. To be conservative in variable selection given the limited number of events, variables with P < 0.10 in univariate analysis were included in the multivariate model. Because there were 0 deaths in the AVF group (perfect separation), we employed Firth’s penalized likelihood method for Cox regression to provide stable hazard ratio estimates. The model was checked for multicollinearity using the variance inflation factor, with all values remaining below 2.0. A two-sided P < 0.05 was considered statistically significant.

Ethical considerations

The study protocol was approved by the Murang’a County Referral Hospital Ethics Committee. Informed consent was waived due to the retrospective nature of the de-identified data analysis. Patient confidentiality was maintained through data anonymization, and the study adhered to the Declaration of Helsinki and STROBE guidelines.

RESULTS

Baseline characteristics

The final cohort of 79 patients had a median age of 62.0 years (IQR: 48.0-74.0) with male predominance (52/79, 65.8%). Hypertension was the leading cause of ESRD (38/79, 48.1%), followed by diabetes mellitus (25/79, 31.6%) and chronic glomerulonephritis (16/79, 20.3%). These categories were not mutually exclusive, as some patients had multiple contributing conditions; the analysis reflects the primary etiology as documented. These were the three exclusive, pre-defined etiological categories for the analysis. Comorbidity analysis revealed a high prevalence of hypertension (68/79, 86.1%), diabetes mellitus (32/79, 40.5%), and human immunodeficiency virus (HIV) (14/79, 17.7%). Other less common conditions were documented but excluded from subsequent multivariate analysis due to low prevalence.

Most patients (71/79, 89.9%) relied on CVCs for vascular access, while only 8 (10.1%) had AVFs. The median baseline haemoglobin was 8.2 g/dL (IQR: 7.1-9.3), indicating significant anaemia, and the median eGFR at dialysis initiation was 7.8 mL/minute/1.73 m² (IQR: 5.4-10.2), suggesting late presentation.

Comparative analysis between survivors and non-survivors revealed significant differences. Non-survivors were significantly older than survivors (median 73.0 years vs 58.0 years, P < 0.001) and had lower baseline haemoglobin (7.1 g/dL vs 8.6 g/dL, P = 0.008). Notably, all 27 patients who died (100%) were dialyzed using a CVC, whereas 15.4% (8/52) of survivors had an AVF (P = 0.018). The baseline eGFR was also significantly lower in non-survivors (6.2 mL/minute/1.73 m² vs 8.1 mL/minute/1.73 m², P = 0.023). No significant differences were found in sex distribution or the prevalence of individual comorbidities like hypertension, diabetes, or HIV between the two groups (Table 1).

Table 1 Baseline characteristics of 79 hemodialysis patients, n (%)/median (interquartile range).

Characteristic	Overall (n = 79)	Survivors (n = 52)	Non-survivors (n = 27)	P value
Demographics
Age (years)	62.0 (48.0-74.0)	58.0 (45.0-68.0)	73.0 (65.0-76.0)	< 0.001
Male sex	52 (65.8)	35 (67.3)	17 (63.0)	0.693
Primary ESRD cause				0.145
Hypertension	38 (48.1)	22 (42.3)	16 (59.3)
Diabetes mellitus	32 (40.5)	20 (38.5)	12 (44.4)
Chronic GN	16 (20.3)	12 (23.1)	4 (14.8)
Comorbidities
Hypertension	68 (86.1)	43 (82.7)	25 (92.6)	0.221
Diabetes mellitus	32 (40.5)	20 (38.5)	12 (44.4)	0.607
HIV	14 (17.7)	8 (15.4)	6 (22.2)	0.430
BPH	8 (10.1)	6 (11.5)	2 (7.4)	0.571
SLE	3 (3.8)	2 (3.8)	1 (3.7)	0.984
Bladder cancer	2 (2.5)	1 (1.9)	1(3.7)	0.629
Baseline laboratory
Hemoglobin (g/dL)	8.2 (7.1-9.3)	8.6 (7.5-9.5)	7.1 (6.3-8.4)	0.008
eGFR (mL/minute/1.73 m2)	7.8 (5.4-10.2)	8.1 (6.0-10.8)	6.2 (4.8-8.1)	0.023
Treatment parameters
Vascular access				0.018
Central venous catheter	71 (89.9)	44 (84.6)	27 (100.0)
Arteriovenous fistula	8 (10.1)	8 (15.4)	0 (0.00)

Using the Mann-Whitney U test. Categorical variables are presented as counts (percentages) and compared using Fisher’s exact test. BPH: Benign prostatic hyperplasia; eGFR: Estimated glomerular filtration rate; ESRD: End-stage renal disease; GN: Glomerulonephritis; HIV: Human immunodeficiency virus; IQR: Interquartile range; SLE: Systemic lupus erythematosus.

Clinical outcomes and mortality

Over the one-year follow-up period, 27 patients died, yielding an all-cause mortality rate of 34.2% (27/79). As expected due to early mortality, the median number of hemodialysis sessions completed was significantly lower in non-survivors compared to survivors [51.0 (IQR: 22.0-89.0) vs 124.0 (IQR: 78.0-198.0), P < 0.001]. Documented complications occurred in 31 patients (39.2%), with infection (18/79, 22.8%) and hypotension (13/79, 16.5%) being most prevalent. Treatment discontinuation occurred in 11 patients (13.9%), primarily due to loss to follow-up (8/79, 10.1%). A Kaplan-Meier survival curve (Figure 1) demonstrates a significant difference in survival probability by vascular access type (Log-rank test, P = 0.015), with AVF patients exhibiting 100% survival at 1 year.

Figure 1 Kaplan-Meier survival curves stratified by vascular access type. Cumulative survival probability over 12 months for 79 incident hemodialysis patients initiating treatment at Murang’a County Referral Hospital. Patients using central venous catheters (CVC, blue dashed line, n = 71) showed significantly lower survival compared to those with arteriovenous fistulas (AVF, black solid line, n = 8) (Log-rank P = 0.015). Hazard Ratio (CVC vs AVF) = 3.12 (95%CI: 1.08-9.01). AVF: Arteriovenous fistula; CVC: Central venous catheter; HR: Hazard ratio.

Predictors of one-year mortality

In univariate Cox regression analysis, older age [hazard ratio (HR) = 1.07 per year, 95%CI: 1.03-1.11, P < 0.001], lower hemoglobin (HR = 0.74 per g/dL, 95%CI: 0.59-0.93, P = 0.009), lower eGFR (HR = 0.87 per mL/minute/1.73 m², 95%CI: 0.78-0.97, P = 0.015), and CVC use (HR = 4.45, 95%CI: 1.25-15.82, P = 0.021) were significantly associated with higher mortality. HIV status showed a trend towards increased risk but was not statistically significant (HR = 1.48, 95%CI: 0.65-3.35, P = 0.350).

For the multivariate model, we included age, hemoglobin, eGFR, CVC use, and HIV status. After adjustment using Firth’s penalized method, older age [adjusted hazard ratio (aHR) = 1.05 per year, 95%CI: 1.01-1.09, P = 0.012] and CVC use (aHR = 3.12, 95%CI: 1.08-9.01, P = 0.036) remained independent predictors of one-year mortality. The estimate for CVC use, while still significant, is more conservative than the standard Cox model and should be interpreted with caution due to the very small AVF comparator group (n = 8). The hazard ratio for CVC use should be interpreted with caution due to the small number of AVF patients (n = 8) and the statistical phenomenon of “perfect prediction” (0 deaths in the AVF group), which can lead to inflated estimates. The association between lower hemoglobin (aHR = 0.84, 95%CI: 0.66-1.07, P = 0.156) and lower eGFR (aHR = 0.92, 95%CI: 0.82-1.03, P = 0.162) with mortality was attenuated and lost statistical significance in the adjusted model. HIV status was not an independent predictor in the multivariate analysis (aHR = 1.32, 95%CI: 0.57-3.05, P = 0.519) (Table 2). In Table 2, for brevity and clarity, the full 95%CIs for non-significant variables like HIV are omitted in the results narrative, though they remain in the table.

Table 2 Predictors of one-year mortality: Univariate and multivariate Cox regression analysis.

Predictor	Univariate analysis		Multivariate analysis
	HR (95%CI)	P value	aHR (95%CI)	P value
Age (per year increase)	1.07 (1.03-1.11)	< 0.001	1.05 (1.01-1.09)	0.012
Male sex (vs female)	0.88 (0.41-1.89)	0.748
Hemoglobin (per g/dL increase)	0.74 (0.59-0.93)	0.009	0.84 (0.66-1.07)	0.156
eGFR (per mL/minute/1.73 m2 increase)	0.87 (0.78-0.97)	0.015	0.92 (0.82-1.03)	0.162
CVC use (vs AVF)	4.45 (1.25-15.82)	0.021	3.12 (1.08-9.01)	0.036
HIV positive	1.48 (0.65-3.35)	0.350
Diabetes mellitus	1.22 (0.59-2.52)	0.594

The multivariate model was fitted using Firth’s penalized likelihood method to address perfect separation (0 deaths in the arteriovenous fistula group) and includes variables significant at P < 0.10 in univariate analysis (age, Hemoglobin, estimated glomerular filtration rate, central venous catheter use). aHR: Adjusted hazard ratio; AVF: Arteriovenous fistula; CVC: Central venous catheter; eGFR: Estimated glomerular filtration rate; HIV: Human immunodeficiency virus; HR: Hazard ratio.

DISCUSSION

This retrospective cohort study from a routine-care Kenyan county hospital reveals two critical findings: A high one-year all-cause mortality rate of 34.2% among incident hemodialysis patients, and the identification of advanced age and catheter-dependent vascular access as significant, independent predictors of this mortality. The near-universal reliance on CVC (89.9%), which stands in stark contrast to international clinical practice guidelines, emerges as the most salient and modifiable risk factor in our setting.

The observed mortality rate is consistent with documented disparities in ESRD outcomes between resource-limited and high-income countries[2,14]. While one-year survival rates often exceed 85% in North American registries[6], our data align with reports from other sub-Saharan African contexts, where late clinical presentation, high treatment costs, and limited supportive care contribute to poorer survival[15-17]. The advanced age of our non-survivors (median 73 years) reaffirms a global trend where older patients experience higher mortality due to a greater burden of frailty and comorbidities[18]. Frailty is a strong, independent predictor of adverse outcomes in this population[19,20]; its trajectory is particularly concerning, as worsening frailty during the initial months of dialysis further escalates mortality risk.

The most compelling finding of our study is the strong independent association between CVC use and mortality (aHR = 3.12). While CVCs are indispensable for urgent dialysis initiation, their immediate utility does not translate to superior long-term outcomes. Our results mirror a large Mexican cohort study which found a 2.8 to 5.0-fold increased mortality risk for tunneled and non-tunneled CVCs, respectively, compared to AVFs[7]. This pattern is consistent across diverse healthcare settings, from national registries in Europe to cohorts in North America[8,20,21].

The 90% prevalence of CVC use indicates a systemic failure in the pre-dialysis care pathway. This finding resonates with recent audits from Kenyan national referral hospitals reporting catheter use rates exceeding 80%[11,12], a pattern replicated in neighboring Tanzania[13]. This “catheter-first” reality is a critical public health issue that is both clinically detrimental and economically inefficient. In our context, high CVC use is less a choice and more a consequence of system-level deficits: Delayed nephrology referral, limited surgical capacity for timely AVF creation, and financial barriers to elective surgery.

The 100% one-year survival of the small AVF subset (n = 8), while requiring cautious interpretation, underscores the survival benefit of overcoming systemic barriers. We acknowledge the potential for confounding by indication, as CVC use often serves as a marker for the “sicker”, late-presenting patient. While we adjusted for age and eGFR, residual confounding by unmeasured factors such as nutritional status (e.g., serum albumin), socioeconomic status, and dialysis adequacy (Kt/V) likely persists. The association between CVC use and mortality should, therefore, be viewed as both a probable direct risk-primarily via infection-and a surrogate for a high-risk patient profile shaped by unmet clinical needs.

Interestingly, while severe anemia was associated with mortality in univariate analysis, it was not an independent predictor after multivariate adjustment. This suggests that anemia may act as a marker of overall illness severity, malnutrition, or chronic inflammation rather than a direct causal factor in this cohort[22,23], with its significance overshadowed by the potent risk profile of advanced age and CVC use.

Several limitations must be acknowledged. First, the retrospective, single-center design and modest sample size (n = 79) limit statistical power and increase the risk of Type II error. The limited number of mortality events (n = 27) relative to candidate predictors carries a risk of overfitting; thus, the multivariate results should be considered exploratory. The wide confidence intervals for key predictors reflect the instability of estimates caused by the small AVF comparator group. Second, the lack of data on socioeconomic status and Kt/V limits our ability to account for all prognostic factors. Finally, the “perfect separation” (zero deaths in the AVF group) necessitated Firth’s penalized regression to stabilize the model.

Notwithstanding these limitations, this study provides crucial, real-world outcome data from a routine county-level hospital-a setting severely underrepresented in nephrology literature. By highlighting the catastrophic rate of catheter-dependent dialysis and its association with mortality, this study delivers an urgent, evidence-based call for health system interventions focused on robust pre-dialysis care and vascular access programs.

CONCLUSION

This study confirms a high one-year mortality rate among incident hemodialysis patients within a Kenyan county hospital setting. Advanced age and the use of CVC for vascular access were identified as significant, independent predictors of death. The near-total reliance on CVCs serves as a potent surrogate marker for a fragmented pre-dialysis care pathway and represents a critical, modifiable risk factor that demands urgent, system-level intervention. To improve patient survival in these settings, resources must be prioritized toward establishing functional vascular access programs and improving early nephrology referral.

ACKNOWLEDGEMENTS

The authors express their gratitude to the administration and staff of the Murang’a County Referral Hospital Renal Unit for their invaluable support and cooperation in facilitating this research.