The Impact of Anaemia on Outcomes, Admissions, and Costs in Patients with Chronic Kidney Disease in Two Public Nephrology Practices in Queensland: A CKD.QLD Registry Study

Aim Anaemia among patients with chronic kidney disease (CKD) leads to poor overall outcomes. This study explores anaemia and its impact on nondialysis CKD (NDD-CKD) patients. Methods 2,303 adults with CKD from two CKD.QLD Registry sites were characterised at consent and followed until start of kidney replacement therapy (KRT), death, or censor date. Mean follow-up was 3.9 (SD 2.1) years. Analysis explored the impact of anaemia on death, KRT start, cardiovascular events (CVE), admissions, and costs in these NDD-CKD patients. Results At consent, 45.6% patients were anaemic. Males were more often anaemic (53.6%) than females, and anaemia was significantly more common over the age of 65 years. The prevalence of anaemia was highest among CKD patients with diabetic nephropathy (27.4%) and renovascular disease (29.2%) and lowest in patients with genetic renal disease (3.3%). Patients with admissions for gastrointestinal bleeding had more severe anaemia, but accounted for only the minority of cases overall. Administration of ESAs, iron infusions, and blood transfusions were all correlated with more severe degrees of anaemia. The number of hospital admissions, length of stay, and hospital costs were all strikingly higher with more severe degrees of anaemia. Adjusted hazard ratios (CI 95%) of patients with moderate and severe anaemia vs. no anaemia for subsequent CVE, KRT, and death without KRT were 1.7 (1.4–2.0), 2.0 (1.4–2.9), and 1.8 (1.5–2.3), respectively. Conclusion Anaemia is associated with higher rates of CVE, progression to KRT and death in NDD- CKD patients, and with greater hospital utilisation and costs. Preventing and treating anaemia should improve clinical and economic outcomes.


Introduction
Anaemia is common among patients with chronic kidney disease (CKD). Studies indicate approximately 16% of Australians with reduced kidney function have low haemoglobin levels compared to 3.1% among the background population [1]. It is estimated that among the 20 million people with CKD in the United States, approximately 2-4 million (15.4%) have anaemia [2] which is twice the Management of anaemia in CKD has been through treatment of the underlying cause and the use of erythropoietin-stimulating agents (ESAs). Improved haemoglobin levels with use of ESAs have shown to improve outcomes such as quality of life [11] and requirement for blood transfusions; however, the majority of these data comes from studies conducted among patients on dialysis compared to patients with CKD not on dialysis (nondialysisdependent chronic kidney disease: NDD-CKD). Importantly, some clinical trials have indicated increased risks for thrombotic events and adverse cardiovascular outcomes with correction of haemoglobin to normal levels, which have raised concerns regarding targets for treatment. Anaemia in CKD tends to be overlooked and goes undertreated in a number of health settings even though efective treatments are clinically available [2,8], with prescribed treatment reported as low as 30% [3]. CKDopps showed that overall only 48% of patients in the US were treated within 3 months of detecting a Hb < 10 g/dL whilst proportions with any treatment for anaemia in stage 4 and stage 5 NDD-CKD were 24% and 32%, respectively [12].
Tis study provides more evidence on the characteristics of anaemia in people with NDD CKD; its efect on hospital admissions; and specifc outcomes such as cardiovascular events (CVE), conversion to kidney replacement therapy (KRT), all cause deaths, and costs. Results could inform clinical decision-making in NDD-CKD patients.

Study Design and Setting.
Tis is a linkage study on people who had previously given informed consent to participation in the Queensland Chronic Kidney Disease (CKD.QLD) Registry [13]. Specifc health clinical and healthcare data were extracted from local and state repositories.

Study
Participants. Te CKD.QLD Registry is described in detail in a previous publication [14]. In brief, people over the age of 18 years who had been referred to public specialist kidney services in Queensland Health in Australia were deemed by a consultant nephrologist to have established CKD and who were not on KRT were invited to consent to having a defned subset of their clinical data deposited iteratively into the CKD.QLD Registry and be followed over time. Recruitment started in May 2011, and participants were followed until they either reached an endpoint of conversion to KRT or death, or until a censor date of June 2018. Tis anaemia study includes recruits from two separate geographically and demographically distinct health services run by Queensland Health. Tese are labelled site A, a metropolitan health service, and site B, a nonmetropolitan health service.

Data Sources.
All clinically relevant data were extracted from the CKD.QLD Registry. Linked data on individual patients' health service utilisation and costs between May 2011 and June 2018 were provided by the Queensland Health Statistical Branch from the a Queensland Hospital Admitted Patient Data Collection (QHAPDC) [15], the Queensland Death Registry and the Activity-Based Funding (ABF) Model Output datasets [17].

Study
Outcomes. Te evaluated outcomes were the prevalence of anaemia in the study cohort and its association with hospital admissions, length of stay in hospital, hospital admission costs, cardiovascular events (CVE), gastrointestinal (GI) bleeding, neoplasms, initiation of KRT, and all cause deaths.
CVE were all admissions related to a CVE diagnosis (either as a principal or other reason) and/or admissions related to cardiovascular death. We further classifed CVE into three groups: CVE any as identifed from both principal diagnosis and/or other diagnosis code sets, CVE principal as identifed only from principal diagnosis code set, and CVE associated as identifed only from other diagnosis code sets. Death was defned as death from major cardiovascular disease/s according to the International Classifcation of Diseases, Tenth Revision (ICD-10) I00-I78 [21]. GI bleeding and neoplasms were each similarly classifed as any, principal, and associated. was used for data analysis. Cox regression was employed to explore the relationship between anaemia severity and primary outcomes, adjusting for potential confounders, which included age, gender, CKD stage, BMI category, proteinuria/albuminuria, site of recruitment, and primary renal diagnosis.  Table 1 summarises the baseline characteristics of the study cohort. A total of 2,303 CKD patients from the two (2) sites were eligible for the study. Tere were 1,305 eligible study participants at site A, 52% males, with a mean age of 65.8 years, and with a mean follow-up of 4.2 (SD 2.2) years. Tere were 998 eligible study participants in site B, 55.6% males, with mean age at consent of 64.2 (SD 15.2) years. Table 1, 54.4% of patients were nonanaemic while 26.7%, 17.8%, and 1% had mild, moderate, and severe anaemia, respectively. Table 2 summarises the distribution of anaemia by gender, age category, CKD stages, and primary renal diagnosis. Te prevalence and severity of anaemia were greater in males than females, and more pronounced at higher ages, especially among those aged >65 years. Anaemia was also more prevalent and severe with more advanced CKD stage. In CKD stages 1-3a, most patients were nonanaemic, while the prevalence and severity of anaemia increased signifcantly from stages 3b-5. Anaemia was more common and severe among patients with RVD and DN, while the lowest rates were among patients with GRD.

Anaemia and Related
Outcomes. Te associations of anaemia with hospital admission numbers, length of stay, and costs among the CKD patients are shown in Figure 1.

Hospital Admissions.
A total of 1,972 (85.6%) study participants had at least one hospital admission, with 14,634 total admissions: the median number of hospital admissions was 117.6 (IQR 41.7-280.9) per 100 person years, ranging from 75.1 (25.9-177.9) in the no-anaemia group, and 160.3 (61-337.6), 247.1 (108.6-510.8), and 474.3 (222.4-645.3) in the mild, moderate, and severe categories anaemia groups, respectively. Te leading principal cause of admissions was diseases of the circulatory system (n � 1,877, 12.8%) (Supplementary data, Table S1). (1,721,454−9,049,241) in those with nonanaemia, and mild, moderate, and severe degree of anaemia, respectively. Table 3, 6.9% of the cohort had admissions for GI bleeding during follow-up: rates were higher with more severe anaemia at baseline, but they still constituted the minority of all anaemic CKD patients. More than quarter of the cohort (27.5%) had admissions for malignancies, but they were not correlated signifcantly with severity of anaemia at baseline. Table 4 summarises the patient outcomes of CVE, commencement of KRT, and death without RRT by anaemia severity at baseline. among participants who were nonanaemic, mildly anaemic, and moderately/severely anaemic in the overall cohort, respectively. Te severity of anaemia was signifcantly associated (p < 0.001) with CVE events after adjusting for the potential confounders, with hazard ratios (95% CI) of 1.4, 1.7 for mild anaemia and moderate/severe anaemia, respectively, compared to those with no anaemia (Table 5).

Commencement of KRT.
Among the 2,303 participants, 10.7% (n � 247) commenced KRT, with distribution by categories of anaemia summarised in Table 4. Te incidence rate of starting KRT per 100 person years increased from 1.2 (0.9-1.5) to 3.9 (3.2-4.9) and then to 7.2 (5.9-8.9) among participants who were nonanaemic, mildly anaemic, and moderately/severely anaemic in the overall cohort. Te severity of anaemia was signifcantly associated (p < 0.001) distribution with starting KRT after adjusting for the potential confounders of age, sex, BMI, CKD stage, proteinuria, site of recruitment, and primary diagnosis, with hazard ratios (95% CI) of those with mild anaemia and moderate/ severe anaemia compared to those with no anaemia of 1  (Table 7).
3.9. Anaemia-Related Treatment. Table 8 summaries use of treatments of erythropoietin stimulating agents (ESA), iron infusions, and blood transfusions among the study cohort.
With worsening degrees of anaemia, the proportion of people who required all three treatment types increased.

Discussion
Anaemia was common among NDD-CKD patients in two public renal specialty practice sites. Participants with anaemia were signifcantly older than those without anaemi,a and proportions of patients with anaemia and higher severity of anaemia were signifcantly higher above the age of 65 years. Tese fndings are comparable to other similar international and national studies [1,[22][23][24][25][26][27]. Anaemia was more prevalent among males who contrast with the general Australian population: a study in 2011-12 showed that women were more likely to have anaemia than men (6.4% compared with 2.5%) [1]. A higher prevalence of anaemia among males was similarly seen in a US study of people with CKD [28]. Te prevalence of anaemia was higher among patients with lower kidney function, which is consistent with the published literature [22,23], and the highest proportion of patients with anaemia was seen in stages 3b and 4. Underrepresentation of stage 5 CKD patients probably represents the small sample size and earlier survival disadvantage of this group. Te CKD cohorts with the highest proportions of anaemia were those with diabetic nephropathy (27.4%) and renovascular disease (29.2%), whereas the lowest prevalence was among patients with genetic renal disease (3.3%). Other studies have reported up to 10 times higher prevalence rates of anaemia among diabetic patients with CKD compared to those without diabetes [7,22,29,30]. While hospital admissions for GI bleeding were more frequent among people with more severe anaemia, only the minority of anaemic persons had such events, and we could fnd no association of anaemia with neoplasms (Table 3).
However, it is worth noting that nearly 27.5% of patients in the study cohort had neoplasm as a comorbidity which highlights the high risk, multimorbid status and more advanced age of many CKD patients, and highlights the complexity and multisystem nature of clinical and holistic care nephrologists and their patients' face.
Evaluation of anaemia and its impact on hospitalisationrelated parameters among a CKD cohort not on KRT is a unique aspect in this study and the frst of its kind in Australia. Participants with anaemia had higher total number of hospital admissions, longer hospital stay, and higher hospital costs during follow-up than the nonanaemic cohort, and these parameters increased exponentially with worsening degree of anaemia. Nissenson et al. found similar outcomes in a US-insured population with anaemia, with outpatient visits, emergency visits, impatient admissions, LOS, and hospital costs signifcantly higher than in nonanaemic patients [27]. Smith [6] reported that people with CKD who were anaemic had twice the average annualised costs of CKD without anaemia, while Lefebvre et al. found that nondialysis patients with CKD had an 40% higher overall health care cost than nonanaemic patients, with the largest driver of cost being due to hospitalisations [31]. Further US studies have supported this fnding [7].
In our fully adjusted Cox regression models, anaemia was a signifcant and independent risk factor for CVE, starting KRT, and death. Furthermore, the incident rates and hazards of developing these outcomes signifcantly increased with worsening anaemia, in keeping the known literature. It is unclear whether anaemia is directly the causative factor or merely associated with adverse outcomes. Many International Journal of Nephrology mechanisms such as infammation, tissue hypoxia, and oxidative stress have been proposed as intermediary mechanisms. Te greater use of ESAs, iron infusions, and blood transfusions in those with more severe anaemia is intuitive; it was not possible to analyse individual responses to these treatments in this study. On a diferent topic, while the apparently higher rates of CVE outcomes by anaemia categories in site B versus site A participants (hazard ratios 1.4 (p < 0.001) are compatible with their higher rates of obesity, diabetes, macroproteinuria, and lower IRSD scores, further evaluation of this phenomenon is needed.
Our fndings enrich the knowledge on profles of anaemia and its infuence on the prognosis in NDD-CKD population in Australia. We acknowledge limitations in the study. With the study limited to CKD.QLD Registry patients from two public kidney services in Queensland, the data may not be generalizable to all CKD populations    International Journal of Nephrology 7    In conclusion, anaemia was highly prevalent in these two study CKD cohorts with signifcant adverse efects on hospitalisations, increased healthcare costs, and poor health outcomes. Mitigating anaemia has the potential to improve clinical and economic outcomes in people with CKD [4]. However, more efective management for anaemia in CKD is needed [32]. Tis study could provide a baseline for evaluation of better management strategies for NDD-CKD patients [16].

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon reasonable request.

Additional Points
Summary at a Glance. Tis large CKD.QLD Registry-based study identifed patterns and rates of anaemia among patients with established chronic kidney disease not on dialysis, and higher rates of progression to kidney replacement therapy, of cardiovascular events and of death as well as greater hospital utilisation and costs in patients with more severe degrees of anaemia.

Disclosure
Part of this study was presented as an abstract for poster presentation at the ANZSN/ASM 2020.

Conflicts of Interest
Te authors declare that they have no conficts of interest.