Risk factors for mortality in elderly haemodialysis patients: a systematic review and meta-analysis

Background Older haemodialysis patients accompany a high burden of functional impairment, limited life expectancy, and healthcare utilization. This meta-analysis aimed to evaluate how various risk factors influenced the prognosis of haemodialysis patients in late life, which might contribute to decision making by patients and care providers. Methods PubMed, Embase, and Cochrane Central were searched systematically for studies evaluating the risk factors for mortality in elderly haemodialysis patients. Twenty-eight studies were included in the present systematic review. The factors included age, cardiovascular disease, diabetes mellitus, type of vascular access, dialysis initiation time, nutritional status and geriatric impairments. Geriatric impairments included frailty, cognitive or functional impairment and falls. Relative risks with 95% confidence intervals were derived. Results Functional impairment (OR = 1.45, 95% CI: 1.20–1.75), cognitive impairment (OR = 1.46, 95% CI: 1.32–1.62) and falls (OR = 1.14, 95% CI: 1.06–1.23) were significantly and independently associated with increased mortality in elderly haemodialysis patients. Low body mass index conferred a mortality risk (OR = 1.43, 95% CI: 1.31–1.56) paralleling that of frailty as a marker of early death. The results also confirmed that the older (OR = 1.43, 95% CI: 1.22–1.68) and sicker (in terms of Charlson comorbidity index) (OR = 1.41, 95% CI: 1.35–1.50) elderly haemodialysis patients were, the more likely they were to die. In addition, increased mortality was associated with early-start dialysis (OR = 1.18, 95% CI: 1.01–1.37) and with the use of a central venous catheter (OR = 1.53, 95% CI: 1.44–1.62). Conclusions Multiple factors influence the risk of mortality in elderly patients undergoing haemodialysis. Geriatric impairment is related to poor outcome. Functional/cognitive impairment and falls in elderly dialysis patients are strongly and independently associated with mortality.


Background
Elderly end-stage renal disease patients constitute an increasing fraction of patients on renal replacement therapy worldwide [1]. The mortality rate of elderly dialysis patients remains confusingly high in spite of recent technical advances, especially in those with a high rate of multimorbidity, muscular functional impairment, cognitive defects or falls [2][3][4][5]. However, there is no consensus about the factors affecting the mortality of elderly haemodialysis patients [6]. In particular, survival is no longer the focus of care. The goal is to either improve the overall quality of life or at least meet some functional or emotional goals, which often entails successful living rather than mere survival. It is important for the need of estimating "geriatric syndromes," such as frailty and falls, to the risk-stratification of older dialysis patients and guiding treatment decisions [7].
This systematic review made an effort to achieve a broad-scale research of the accessible studies to recognize the risk factors for mortality in elderly haemodialysis patients. The goal was to evaluate the association of functional impairment, cognitive disfunction and falls with mortality in elderly haemodialysis patients, in addition to other known risk factors. Recognizing these associations might help improve treatment solutions or protective measures.

Methods
The study was written according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines and was displayed in keeping with the PRISMA-P checklist (Additional file 1).

Protocol and registration
No registered protocol.

Data extraction and risk-of-Bias assessment
Two authors, YHS and GYC, independently displayed the list of studies generated by the search, with disagreements resolved by a third author, YFX. Titles and abstracts of all studies were screened before acquiring full-text versions of relevant studies. Two authors extracted data from fulltext articles independently.
Inclusion criteria: (1) risk factors for mortality of elderly haemodialysis patients were the subject; (2) haemodialysis patients included an elderly population; and (3) study data included odds ratio(OR) values and 95% confidence intervals (CIs) or data that could be transformed to OR values and 95% CIs by statistical methods.
Exclusion criteria: (1) the abstract was not in English; (2) the study did not involve elderly haemodialysis patients; and (3) it was a case report, abstract, review, conference report or animal experiment.
The quality of articles was accessed using the Newcastle-Ottawa quality assessment scale [8]. Studies below 5 points were accounted to have a high risk of bias and were excluded.

Data collection and analysis
The data included authors, year of publication, number and mean age of participants, percentage of men, median duration of follow-up, survival or mortality, risk factors, and definition of ageing. The risk factors mainly included age, cardiovascular disease, diabetes mellitus, type of vascular access, dialysis initiation time, nutritional status and geriatric impairments. Geriatric impairments included frailty, cognitive or functional impairment and falls. Relative risks with 95% CIs were derived. We extracted the adjusted hazard ratios (HRs) and 95% CIs from all included studies.

Statistical analysis
We evaluated the pooled relative risk and the 95% CI of the included articles through the inverse variance method. ORs of retrospective studies were considered as approximate risk ratios (RRs). We used the I 2 statistic and Q test to account the heterogeneity among the included studies. No significant heterogeneity was present if the I 2 statistic value was < 50%. Then we used a fixedeffect model to calculate the pooled 95% CI. If significant heterogeneity was showed (I 2 statistic value was ≥50%), we used the random-effect model. Review Manager 5.3 was used for statistical analyses.

Publication Bias
We evaluated Publication bias by Egger's and Begg's tests at the 5% significance level. Point prevalence with 95% CIs was showed in the forest plot pattern. A funnel plot was used to evaluate the publication bias.

Sensitivity analysis
We accessed the OR value and 95% CI of each risk factor under the fixed-effect model and the random-effect model separately. If the difference between the two results was small, the combined results had low sensitivity and stability.

PRISMA flow chart
The overall literature search generated 6785 articles. Of these, 291 articles were selected on the basis of the inclusion and exclusion criterion of the literature. We excluded 211 irrelevant studies after reading their titles and abstracts. Thus, 83 latent full-text articles were appraised for qualification, which brought about further exclusion of 55 articles because the result of interest was not demonstrated. After all, 28 articles were in line with the suitability criteria and were included in this metaanalysis. The flow diagram of the study screening procedure was showed in Fig. 1. The features of the 28 articles were summarized in Table 1.

Risk factors for mortality in elderly haemodialysis patients Age
Eleven studies assessed the association between age and mortality in elderly haemodialysis patients [10,11,14,19,21,22,25,27,32,34,35]. A random-effect model was used to analyse these eleven studies because there was heterogeneity between them (p < 0.001, I 2 = 97%). The results showed that age was a risk factor for mortality in elderly haemodialysis population (OR = 1.43, 95% CI: 1.22-1.68), as shown in Fig. 2.

Cardiovascular disease (CVD)
Five studies appraised the association between CVD and mortality in elderly haemodialysis patients [9,11,22,35,36]. Was used a random-effect model to analyse these 5 studies because of the heterogeneity between them (p = 0.05, I 2 = 58%). The results showed that CVD was not a

Diabetes mellitus (DM)
We screened five studies assessing the association between DM and mortality in elderly haemodialysis patients [9,11,22,35,36]. The results showed no heterogeneity between studies (p = 0.72, I 2 = 0%), so a fixed-effect model was used. Analysis of these 5 studies revealed that DM was a risk factor for mortality in aged haemodialysis population (OR = 1.19, 95% CI: 1.06-1.33).

Central venous catheter dialysis
Four studies assessed the association between central venous catheters and mortality in elderly haemodialysis patients [20,22,25,36]. The results showed heterogeneity between studies (p = 0.07, I 2 = 58%), so the randomeffect model was used to analyse these data. Central venous catheter dialysis was a risk factor for mortality in aged haemodialysis population (OR = 1.55, 95% CI: 1.38-1.75).

Early-start dialysis
Three studies assessed the association between dialysis initiation time and mortality in elderly haemodialysis patients [19,23,28]. The results showed no heterogeneity between studies (p = 0.26, I 2 = 26%), so we used a fixedeffect model. Analysis of these 3 research studies showed that early dialysis was an infuencing factor for mortality in elderly haemodialysis population (OR = 1.11, 95% CI: 1.08-1.14).

Frailty
Five studies assessed the association between frailty and mortality in elderly haemodialysis patients [16,18,19,26,30]. A fixed-effect model was used to analyse these five studies because there was no heterogeneity between them (p < 0.00001, I 2 = 32%). The results showed that frailty was a risk factor for mortality in elderly haemodialysis population (OR = 1.43, 95% CI: 1.31-1.56), as shown in Fig. 3.

Functional impairment
Seven studies assessed the association between functional impairment and mortality in elderly haemodialysis patients [9,11,16,18,22,31,35]. A random-effect model was used to analyse these data because there was heterogeneity between the studies (p = 0.0006, I 2 = 75%). The results showed that functional impairment was a risk factor for mortality in elderly haemodialysis population (OR = 1.45, 95% CI: 1.20-1.75), as shown in Fig. 4.

Cognitive impairment
Three studies assessed the association between cognitive impairment and mortality in elderly haemodialysis patients [12,15,31]. A fixed-effect model was used to analyse these data because there was no heterogeneity between studies (p < 0.00001, I 2 = 0%). The results showed that cognitive impairment was a risk factor for Fig. 2 Forest plot of the relationship between age and mortality in elderly hemodialysis patients

Falls
Only two studies assessed the association between falls and mortality in elderly haemodialysis patients [13,33].
The results showed that falls were a risk factor for death in elderly haemodialysis population (OR = 1.14, 95% CI: 1.06-1.23).

Sensitivity analysis and summary of the meta-analysis results of risk factors for mortality in elderly haemodialysis patients
From the 28 selected studies , a summary of the meta-analysis results of risk factors for mortality in elderly haemodialysis patients was shown in Table 2. The OR value and 95% CIs of each risk factor were assessed under the fixed-effect model and the random-effect model separately. The difference between the two results was small, indicating that the combined results had low sensitivity and stability.

Publication bias
The funnel plots expressed symmetric patterns for each outcome, as shown in Figs. 6 and 7. We conducted Begg's test to evaluate the publication bias using Stata software because the sample sizes of the outcomes included in this meta-analysis were small, which demonstrated no significant heterogeneity among the 28 studies.

Discussion
We evaluated 28 studies that composed of risk factors for mortality in elderly haemodialysis population in this study. This study supported the opinion that evaluation of geriatric senescence might promote to making decisions for dialysis by illustrating that multiple impaired factors are relevant to poor consequence [37][38][39]. Functional impairment is considered to be a contributor to subsequent disability, recurrent hospitalization, and decreased survival rate [18,31,40]. Loss of independent functioning has been recognized in geriatric dialysis patients [41][42][43]. Sensorial degeneration and sight defect are also not uncommon [44]. We found that functional impairment was a powerful, independently coherent predictor of mortality in elderly dialysis person. There is a requirement for early recognition of elderly haemodialysis patients who might get help from involvements in order to prevent or decrease geriatric impairment.
Cognitive impairment is not uncommon among dialysis patients [45]. This review showed that cognitive impairment in older haemodialysis patients is positively correlated to mortality. Older haemodialysis patients are also at potential risk of being befalled with Alzheimer's disease, and receiving this diagnose is associated with an increased mortality [46,47]. Another study found that dementia was associated with an increased risk of death and dialysis drop out in adults aged over 75 years on dialysis [31]. Elderly dialysis patients should be considered to establish routine screening for cognitive impairment so as to recognize those at risk for related adverse consequences [12]. Large-scale studies to clear the vintage methods for detection, treatment and prevention of cognitive impairment are of critical necessity in this highrisk groups [48].
Other age-related comorbidities, such as falls, consult an independent and significant mortality risk for geriatric dialysis population. Approximately 40% of elderly dialysis patients encounter one or more unexpected falls within one-year phase [13]. Multiple mediations have been performed to decrease fall rates and/or prevent damage associated with falls. These consist of multivariate evaluation and intervention, exercise moderating and the use of hip protectors in specific populations [13]. We also found that older age and more combined conditions (such as diabetes mellitus or hypertension) were correlated with higher mortality, which is well known in the general population. The observation that low BMI conferred a mortality risk paralleled the finding of frailty as a marker of early death. Increased mortality was also associated with early-start dialysis and with the use of a central venous catheter. The latter two points are well understood in the renal literature.
However, our meta-analysis had several limitations. First, the sample size of the included studies was too different and may have amplified the impact of individual studies on our results. Second, follow-up time affected the mortality of haemodialysis patients, which affected the accuracy of our meta-analysis. Third, this evidence is derived from a heterogeneous cohort of studies and the definitions of old age were different between our studies. Some of the research only included a small number of elderly patients aged over 80 years, which may have increased the mortality of haemodialysis patients and reduced the accuracy of our results. Additionally, the quality of this metaanalysis might be affected by the limitations at the review level (e.g., reporting bias) and at the outcome level (e.g., risk of bias).

Conclusions
This review described the impact of various characteristics on the risk of mortality in elderly patients undergoing haemodialysis. The mortality is high in
Additional file 2. Example search strategy using PubMed.
Abbreviations CKD: Chronic kidney disease; CVD: Cardiovascular disease; BMI: Body mass index; DM: Diabetes mellitus; OR: Odds ratio; HR: Hazard ratio; RR: Risk ratio; PRISMA: Preferred Reporting Items for Systematic reviews and Meta-Analyses