Five‐Year Outcomes of Chronic Kidney Disease in a Longitudinal Population‐Based Cohort in Western Australia

ABSTRACT

Aim

To describe 5‐year outcomes of chronic kidney disease (CKD) in a large, population‐based cohort in Western Australia, including rates of progression, regression, kidney failure and death.

Background

CKD is a major public health challenge associated with increased morbidity and mortality, yet the natural history of early‐stage disease in population settings remains poorly defined.

Methods

We conducted a retrospective cohort study using linked pathology and hospital data for adults with incident CKD between 2006 and 2022. CKD stage was defined by two estimated glomerular filtration rate (eGFR) values ≥ 3 and < 12 months apart: mild (45–59), moderate (30–44) and severe (15–29) mL/min/1.73 m². Outcomes were assessed over 5 years using a competing risks framework.

Results

Among 153 527 individuals with incident CKD during the study period (mean age 75.6 years, 52.6% women) 118 248 had mild, 58 323 moderate and 20 322 severe CKD. At 5 years, death occurred in 16.8% (mild), 26.7% (moderate) and 30.5% (severe); kidney failure occurred in 0.6%, 2.5% and 30.0%, respectively. CKD progression was more common than regression in mild (20.1% vs. 11.8%) and moderate (18.9% vs. 14.5%) disease. Albuminuria testing was recorded in < 35% of patients.

Conclusion

People with early‐stage CKD face substantial risks of progression, kidney failure and death within 5 years. The low rate of albuminuria testing suggests missed opportunities for early identification of high‐risk individuals and timely intervention. Enhancing CKD detection and monitoring in primary care may improve outcomes.

Summary at a Glance

This large population‐based study examined 5‐year outcomes in nearly 200 000 Western Australian adults with incident chronic kidney disease (CKD), using linked pathology and hospital records. Even among individuals with early‐stage CKD, risks of progression, kidney failure and death were substantial, with fewer than half of those with mild disease remaining event‐free at 5 years. CKD regression was observed but occurred less frequently than progression. Notably, fewer than one‐third of patients had recorded albuminuria testing, limiting accurate risk stratification. These findings highlight critical gaps in CKD monitoring and reinforce the need for earlier detection and implementation of guideline‐recommended care to reduce preventable progression and mortality.

1. Introduction

Chronic kidney disease (CKD) is a global health problem affecting around 10% of the population, with incident cases of CKD increasing 33% from 7.80 million in 1990 to 18.99 million in 2019 [1, 2]. The ageing population and rising prevalence of risk factors such as diabetes are expected to exacerbate this problem [1, 3]. Particularly at its later stages, CKD is associated with high morbidity, mortality and healthcare costs, with many health systems unable to meet demand for kidney replacement therapies (KRTs) [4, 5]. In early 2025, the World Health Organization recommended Member States urgently integrate kidney disease prevention and care into national health systems to reduce the burden of disease [6].

Early stages CKD can often be asymptomatic, leading to delayed presentation, diagnosis and intervention [7]. Despite the lack of symptoms, early‐stage CKD significantly increases cardiovascular death risk, highlighting the importance of early intervention [8]. In addition to diet and lifestyle modifications, there are an increasing number of medical interventions that can slow progression and reduce morbidity of CKD [4, 9, 10]. However, there is a lack of research on early‐stage CKD, risks for progression or regression and patient trajectories following diagnosis [11].

CKD is defined by a sustained (more than 3 months) reduction in estimated glomerular filtration rate (eGFR) or presence of albuminuria, with or without radiological signs of kidney damage. Many individuals will present to different health service providers over time and may undergo testing which is stored in different health records. Without a complete medical history, it is often difficult for a clinician to make a diagnosis of CKD, and this is contributing to the underestimation of the burden of CKD.

Much research in CKD has focussed on risks for progression of disease and end‐stage kidney disease (ESKD) [12]. Diabetes and cardiovascular disease are well established comorbidities associated with disease onset, progression and outcome [12]. Costs of ESKD and delivery of KRT are also well explored [13]. What is less well established is the rate of disease progression and probability of disease regression at different stages of CKD. A recent Canadian study showed that the 5‐year probability of CKD regression was similar to that of disease progression or kidney failure for people with advanced CKD [14].

Despite the global burden of CKD, few studies have been able to examine incident CKD at a whole‐of‐population level with statewide pathology data. Western Australia is one of the only jurisdictions internationally with near‐complete linkage of all community and hospital serum creatinine testing, enabling accurate identification of incident CKD and long‐term follow‐up. Using this unique dataset, we applied a multi‐state competing risks framework to quantify 5‐year probabilities of progression, regression, kidney failure and death, providing contemporary population‐level insight into the natural history of early‐stage CKD.

2. Methods

2.1. Data

We utilised a WA dataset that used Privacy Preserving Record Linkage (PPRL) [15, 16, 17] to link pathology, hospital, emergency department and death records. Pathology data was provided by the four main pathology providers in WA, which have almost complete coverage of pathology testing in WA, including hospital and non‐hospital testing [18]. Each pathology provider contributed at least serum creatinine records, and three out of four contributed a range of additional variables including albumin‐to‐creatinine ratio (ACR) and demographic information. One of the pathology providers was only able to provide data from the period 2017 to 2022. Pathology data was linked to hospital admissions (Hospital Morbidity Data Collection), emergency department presentations (Emergency Department Data Collection) and death records (Death Register; Cause of Death Unit Record File), which also have state‐wide coverage. Hospital administrative variables included diagnosis codes, procedural codes, cost codes, socioeconomic index and other demographic information. Diagnosis codes and procedural codes for comorbidities were sourced from the website of the Australian Institute of Health and Welfare (AIHW) [19, 20, 21].

2.2. Study Cohort

Individuals were included in the study if they were over 18 years old and had received a serum creatinine test from one of four pathology providers in WA during the period 1 January 2006 and 22 March 2022. Incident cases of mild (eGFR, 45–59 mL/min/1.73 m²), moderate (eGFR, 30–44 mL/min/1.73 m²) and severe (eGFR, 15–29 mL/min/1.73 m²) CKD were identified by at least two eGFR records within the same stage separated by at least 3 months and within 12 months. The date of the second eGFR record was the date of qualification for that stage. Individuals could qualify as incident in more than one CKD stage over time. However, once an individual qualifies in a higher stage, they can no longer qualify in a lower stage at a later date. For example, an individual can't qualify as an incident for Stage 3a after they have already qualified for Stage 3b and, Stages 3a and 3b after qualifying for Stage 4. If multiple eGFR results within the 3–12‐month qualifying window fell into different CKD stages, the individual can only qualify when test results were within the same stage or more severe as proposed by Liu et al. [14]. For example, if the index test is Stage 3b then the initial test and in between tests have to be Stage 3b or more severe (e.g., Stage 4). Individuals who never met this requirement were not included as incident CKD cases. This approach ensures that baseline CKD staging reflects a sustained and internally consistent level of kidney function, and prevents misclassification of individuals with fluctuating or borderline values. We used eGFR and hospital diagnosis and procedural codes to exclude individuals with a previous more severe stage of CKD or who had a recorded history of KRT.

2.3. Outcome Variables

Progression was defined as a sustained change in eGFR of at least 3 months and within 12 months to a higher stage CKD, and a ≥ 25% decrease in eGFR from baseline to limit a boundary effect of small fluctuations around the cutoff value. Regression was defined as a sustained change in eGFR of at least 3 months and within 12 months to a lower stage CKD and a ≥ 25% increase in eGFR from baseline. Kidney failure was defined as a sustained change in eGFR for at least 3 months and within 12 months to below 15 mL/min/1.73 m² or when a diagnosis/procedure related to kidney failure (e.g., CKD Stage 5, dialysis) was identified from hospital records. Death was defined from mortality records. Date of progression, regression and kidney failure was defined in the same way as stage qualification, where the second of two sequential eGFR records for that stage determined the event date. We observed outcomes at 5‐years follow‐up.

Microalbuminuria was determined by a urinary ACR (UACR) of 2.5–25 mg/mmol for men and 3.5–35 mg/mmol for women, and macroalbuminuria was determined by a UACR of more than 25 mg/mmol for men and more than 35 mg/mmol for women. Albuminuria records were only included if they were reported within 365 days of the date of qualification. Comorbidities were determined by any record of a diagnosis code (Supporting Information) for that disease (defined by the AIHW) prior to the date of qualification.

2.4. Statistical Analysis

We summarised descriptive statistics at baseline (date of cohort qualification) and report these for age, sex, albuminuria and comorbidities of diabetes, cardiovascular disease and cancer. A nonparametric Aalen‐Johansen method was used to determine cumulative incidence functions of competing events (progression, regression, kidney failure and death) for each stage of CKD over 5 years [22]. Median time to event (within 5 years) was determined for each outcome for the first event. For individuals in the groups that had an outcome of progression or regression, the competing events model was repeated to determine the probability and time to the second event using the same outcomes (progression, regression, kidney failure and death).

3. Results

3.1. Cohort

A total of 153 527 individual people were eligible for the cohort analysis (Figure 1). Cohort characteristics at the time of qualification (baseline) are summarised in Table 1. At baseline there were 118 248 people (52.7% women) with mild, 58 323 people (53.1% women) with moderate and 20 322 people (50.8% women) with severe CKD, and the mean age for each stage was 74.8 (SD 11.1) years, 77.9 (SD 11.5) years and 76.6 (SD 14.0) years, respectively. In every stage the majority of people were over 65 years old. Albuminuria increased with advancing stage, but albuminuria was unmeasured for over two‐thirds of the cohort. The proportion of people with comorbid diabetes increased from mild (24.0%) to moderate (32.8%) and severe (41.6%) CKD, and cardiovascular disease increased from 61.3% in mild CKD to 76.1% in severe CKD. The proportion of people with comorbid cancer was relatively constant across all stages (17.3% mild, 18.7% moderate and 18.6% severe). Because individuals could qualify as incident at multiple CKD stages over time, the sum of stage‐specific incident cohorts exceeds the number of unique individuals.

FIGURE 1.

Cohort selection. Individuals 18 years or older were filtered to create a final cohort of 153 524 people for analysis.

TABLE 1.

Cohort statistics at baseline.

	Mild CKD a	Moderate CKD a	Severe CKD a
n	118 248	58 323	20 322
Age, years, mean (SD)	74.8 (11.1)	77.9 (11.5)	76.6 (14.0)
Age group, years, n (%)
18–64	18 697 (15.8%)	6739 (11.6%)	3693 (18.2%)
65–74	33 779 (28.6%)	11 482 (19.7%)	3344 (16.5%)
75–84	43 520 (36.8%)	22 201 (38.1%)	6525 (32.1%)
≥ 85	22 252 (18.8%)	17 901 (30.7%)	6760 (33.3%)
Sex
Women	62 327 (52.7%)	30 943 (53.1%)	10 323 (50.8%)
Men	55 921 (47.3%)	27 380 (46.9%)	9999 (49.2%)
Index eGFR, mean (SD) (mL/min/1.73 m2)	53.1 (4.3)	38.3 (4.2)	24.3 (3.9)
Albuminuria
Unmeasured	90 204 (76.3%)	42 603 (73.0%)	13 586 (66.9%)
Microalbuminuria	7752 (6.6%)	4881 (8.4%)	1943 (9.6%)
Macroalbuminuria	4154 (3.5%)	4205 (7.2%)	3423 (16.8%)
Normoalbuminuria	16 138 (13.6%)	6634 (11.4%)	1370 (6.7%)
Comorbidities
Diabetes	28 365 (24.0%)	19 159 (32.8%)	8448 (41.6%)
Cardiovascular disease	72 491 (61.3%)	41 910 (71.9%)	15 472 (76.1%)
Cancer	20 452 (17.3%)	10 902 (18.7%)	3789 (18.6%)

^a Mild (eGFR 45–59 mL/min/1.73 m²), moderate (eGFR 30–44 mL/min/1.73 m²) and severe (eGFR 15–29 mL/min/1.73 m²).

3.2. Incidence and 5‐Year Outcome of First Event

Death was among the most common events across all stages of CKD, occurring in 16.8% of mild, 26.7% of moderate and 30.5% of severe CKD. Progression to kidney failure was rare in mild (0.6%) and moderate (2.5%) CKD, but was almost as common as death in severe CKD (30.0%). Progression of CKD was the second most likely event to death in mild and moderate CKD, occurring in 20.1% and 18.9% of the populations respectively. Regression of CKD remained relatively constant across the disease spectrum, occurring in 11.8% of mild and 14.5% of moderate and severe CKD. At the time of censoring, 50.7% of people with mild CKD remained event‐free and this decreased to 37.3% of people with moderate, and 24.8% of people with severe CKD (Figure 2).

FIGURE 2.

Five‐year outcomes of CKD by stage. Cumulative proportion of individuals with mild (eGFR 45–59 mL/min/1.73 m²), moderate (eGFR 30–44 mL/min/1.73 m²) and severe (eGFR 15–29 mL/min/1.73 m²) CKD who experienced death, progression, regression, kidney failure or remained event‐free over 5 years. Outcomes were mutually exclusive and determined using Aalen Johansen competing events model.

3.3. Time to Outcome

Of those who advanced to kidney failure or death, the median time to death or kidney failure decreased by CKD stage from mild (1.5 [IQR 0.6–2.8] and 1.5 [IQR 0.6–2.7] years, respectively), to moderate (1.5 [IQR 0.6–2.7] and 1.3 [IQR 0.5–2.5] years, respectively) and severe (1.2 [IQR 0.4–2.3] years to death) CKD, with the exception of kidney failure which increased slightly from moderate to 1.5 [IQR 0.7–2.6] years in severe CKD. For individuals who progressed in CKD stage, the median time to progression also decreased from 2.1 [IQR 1.2–3.3] years in mild to 1.9 [IQR 1.1–3.1] years in moderate CKD. Regression was observed at all stages of CKD, and for those who regressed, the median time decreased from 1.7 [IQR 1.0–2.9] years in mild, to 1.5 years [IQR 0.9–2.6] in moderate and 1.2 [IQR 0.7–2.1] years in severe CKD (Figure 3).

FIGURE 3.

Median time to outcome event. Median time to death, kidney failure, progression or regression by baseline CKD stage. Time to event calculated from index date of CKD qualification to the first observed outcome within 5 years. CKD is defined as mild (eGFR 45–59 mL/min/1.73 m²), moderate (eGFR 30–44 mL/min/1.73 m²) and severe (eGFR 15–29 mL/min/1.73 m²).

3.4. 5‐Year Outcome of Second Event

For individuals that had either a progression or regression in the initial competing events model (Figure 4a), we repeated the analysis for the same outcomes over the subsequent 5 years (Figure 4b). People who had a regression from moderate or severe CKD had a low probability of further regression, and a relatively high probability of progression or death. People who regressed from severe CKD had a lower probability of kidney failure than people with moderate CKD in their first event. People who progressed from moderate to severe CKD had a similar 5‐year outcome profile as those who had an index event of severe CKD. People who moved from mild to moderate CKD had a similar pattern as those with an index event of moderate CKD; however, the probability of remaining stable in the stage of their second event was lower, with probabilities for both progression, regression and kidney failure increasing slightly.

FIGURE 4.

(a) Cumulative incidence of first CKD outcome over time. Cumulative incidence functions showing time to first event (progression, regression, kidney failure or death) from baseline CKD stage. Events modelled using Aalen–Johansen estimator with competing risks framework. CKD is defined as mild (eGFR 45–59 mL/min/1.73 m²), moderate (eGFR 30–44 mL/min/1.73 m²) and severe (eGFR 15–29 mL/min/1.73 m²). (b) Cumulative incidence of second CKD outcome after regression or progression. Cumulative incidence functions for second event among individuals who experienced regression (top) or progression (bottom) as a first event. Probabilities estimated for death, progression, regression, or kidney failure over the subsequent 5 years. CKD is defined as mild (eGFR 45–59 mL/min/1.73 m²), moderate (eGFR 30–44 mL/min/1.73 m²) and severe (eGFR 15–29 mL/min/1.73 m²).

4. Discussion

This study provides one of the largest population‐level examinations to date of CKD outcomes in a contemporary Australian cohort. Using linked longitudinal data, we tracked 5‐year trajectories of disease progression, regression, kidney failure and mortality, revealing critical insights into the natural history of CKD in routine clinical care. Most notably, we show that even early‐stage CKD is associated with substantial risk of adverse outcomes within a relatively short timeframe.

Nearly half of individuals with mild CKD remained event‐free over 5 years, suggesting a window of opportunity for intervention. However, the risk of progression or death still exceeded that of regression, underscoring the clinical importance of early identification and management. Compared with the Canadian population‐based study by Liu and colleagues [14], our cohort demonstrated slightly lower rates of CKD regression across all stages and higher rates of progression in mild and moderate disease. These differences may reflect variations in clinical practice, comorbidity burden, or access to care and warrant further investigation.

In severe CKD, the probabilities of death (30.5%) and kidney failure (30.0%) were nearly equal, aligning with international findings that suggest late‐stage CKD carries a bifurcated risk of either death or progression to KRT [7, 8, 10, 23, 24]. This differs from the Canadian study, where death was more common than kidney failure but aligns with other international reports suggesting that the balance of these outcomes may vary depending on healthcare system performance and referral practices [8, 24, 25].

One particularly concerning finding was the low frequency of albuminuria testing, with fewer than one‐third of patients across all CKD stages receiving urinary ACR measurement. While other jurisdictions globally show significantly higher rates of ACR testing, this is not outside of what is reported in other developed countries in Europe and elsewhere [26, 27, 28]. Given that albuminuria is a cornerstone of CKD risk stratification and eligibility for disease‐modifying therapies such as SGLT2 inhibitors and GLP‐1 receptor agonists [29, 30], underuse of this test likely represents a missed opportunity for early intervention and optimal management. Prior Australian work has similarly noted poor adherence to Kidney Health Australia (KHA) guidelines regarding albuminuria testing [31].

The 5‐year trajectories following CKD regression revealed that a return to a lower stage is not always durable. Most individuals who initially regressed subsequently progressed or died, with only a small proportion experiencing sustained improvement. This finding confirms the hypothesis that regression of CKD, even for a prolonged period, does not necessarily indicate disease resolution and highlights the need for close follow‐up. The instability of eGFR trajectories ‐ especially in the context of comorbidities and ageing ‐ reinforces the utility of longitudinal monitoring frameworks that go beyond snapshot eGFR measures.

Our results also confirm prior observations that small numerical fluctuations in eGFR may not reliably reflect meaningful clinical change [12, 14]. To mitigate the risk of misclassification due to regression to the mean or lab variability, we adopted a conservative definition of change requiring both a stage transition and a 25% eGFR difference. Nonetheless, some outcome misclassification remains possible, especially in those near stage boundaries.

Our findings must also be interpreted in the context of growing concern regarding the use of fixed eGFR thresholds to define CKD in elderly populations. As recently highlighted in Nephrology, the application of a universal eGFR < 60 mL/min/1.73 m² cutoff across all age groups risks overdiagnosis among older adults for whom mild reductions in GFR may reflect physiological ageing rather than pathological decline [32]. In our cohort, the average age at diagnosis exceeded 75 years, raising the possibility that a substantial proportion of individuals (particularly those with mild CKD and no albuminuria) may have been misclassified as having disease despite low risk of progression or death. Studies have shown that eGFR values between 45 and 59 mL/min/1.73 m² in the absence of albuminuria or structural kidney abnormalities are not strongly associated with adverse outcomes in older adults, and may not warrant a CKD diagnosis [14]. Given that albuminuria testing was infrequent in our dataset, we were unable to fully differentiate between low‐risk and high‐risk individuals, potentially leading to an overestimation of disease burden and associated outcomes in this group. This underscores the need for age‐adapted diagnostic frameworks and risk models that incorporate both kidney function and damage markers to improve prognostic accuracy and avoid unnecessary medicalisation of ageing.

While our findings are broadly consistent with the epidemiology of CKD reported internationally [1, 5], they also reflect distinctive patterns within the Western Australian healthcare context. The relatively high rates of moderate to severe CKD among older adults, and the strong comorbidity burden (notably cardiovascular disease and diabetes), align with national mortality trends in which CKD is frequently listed as a contributing cause of death [33]. The short median time to death or kidney failure across all CKD stages ‐ often under 2 years ‐ further emphasises the urgency of earlier detection and risk stratification.

This study benefits from near‐complete state‐wide capture of pathology and hospital data over a 16‐year period, leveraging privacy‐preserving linkage to integrate fragmented health records. However, several limitations should be noted. Missing data on albuminuria, HbA1c and medication use limit our ability to fully characterise CKD risk and management. Also, while our cohort is broadly representative of the Western Australian adult population, differences in health system design, testing practices and clinical pathways may limit generalisability to other jurisdictions.

Our analyses were conducted using stage‐specific incident cohorts, whereby individuals could contribute incident observations at more than one CKD stage over time if they met incident criteria for each stage. This approach reflects the progressive nature of CKD and allows estimation of stage‐specific outcomes using independent competing risks models. However, because stage entry depends on the timing and frequency of eGFR testing, some individuals may first qualify at a more advanced stage if earlier disease was unmonitored. As a result, later‐stage cohorts may include a mix of individuals with truly rapid progression and those with unobserved earlier stages. This limitation should be considered when interpreting outcomes in the moderate and severe CKD groups, and highlights the dependence of CKD staging on testing intensity in routine care.

Our definitions of progression and regression relied on transitions between discrete CKD stages, which introduces the possibility of misclassification for individuals whose eGFR values lie close to stage boundaries. We attempted to mitigate this by requiring a ≥ 25% change in eGFR in addition to a change in stage, reducing the influence of biological and analytical variability. Nonetheless, categorisation inevitably loses information. Alternative approaches ‐ such as modelling eGFR as a continuous, time‐varying variable or estimating individual eGFR slopes ‐ may provide a more sensitive assessment of kidney function change and avoid threshold effects. These methods were not feasible within our current dataset but could offer complementary insights in future analyses, particularly for identifying subtle early declines in function.

These findings have clear implications for health service planning and clinical practice. Despite being frequently described as a slowly progressive disease, CKD ‐ even in its early stages ‐ carries a significant risk of deterioration or death within a few years. The current underuse of albuminuria testing may delay identification of high‐risk patients and hinder access to treatments known to reduce progression and cardiovascular events. Improved implementation of guideline‐recommended testing and monitoring, coupled with population‐level strategies to manage diabetes, hypertension and cardiovascular disease, may meaningfully improve outcomes. The role of primary care, in particular, is pivotal for early detection and management.

Author Contributions

E.T. led the conceptualisation, formal analysis, investigation, methodology, project administration and drafted the original manuscript. K.E.K.C. contributed to data curation, formal analysis, methodology, visualisation and manuscript review. C.M.Y.L. assisted with formal analysis and methodology and reviewed the manuscript. S.R., M.E., A.I., G.D. and N.S.R.L. contributed to manuscript review and editing. J.H.B., D.H. and S.Ro. provided conceptual input, supervision and secured funding. A.C. contributed to conceptualisation, supervision, formal analysis, investigation, methodology and funding acquisition. All authors reviewed and approved the final manuscript.

Funding

This study was supported by the Digital Health Cooperative Research Centre (DHCRC), funded under the Australian Government's Cooperative Research Centres (CRC) Program (grant number DHCRC‐0073).

Ethics Statement

A waiver of consent was provided by the WA Department of Health Human Research Ethics Committee (RGS0000001183) due to the size of the population and time elapsed since care.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Annual counts of incident chronic kidney disease (CKD) cases identified during the study period (2006–2022).

Supporting Information: Methods: detail additional methodological information, including diagnostic and procedural codes used to exclude individuals with prior kidney replacement therapy or more severe CKD before cohort entry, as well as definitions of comorbid conditions derived from linked hospital and emergency department records. These supplementary materials provide further detail to support transparency and reproducibility of the cohort construction and analytic approach.

Thomas E., Chai K. E. K., Lee C. M. Y., et al., “Five‐Year Outcomes of Chronic Kidney Disease in a Longitudinal Population‐Based Cohort in Western Australia,” Nephrology 31, no. 2 (2026): e70172, 10.1111/nep.70172.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.