Projection of the health and economic impacts of Chronic kidney disease in the Chilean population

Background Chronic Kidney Disease (CKD) is a leading public health problem, with substantial burden and economic implications for healthcare systems, mainly due to renal replacement treatment (RRT) for end-stage kidney disease (ESKD). The aim of this study is to develop a multistate predictive model to estimate the future burden of CKD in Chile, given the high and rising RRT rates, population ageing, and prevalence of comorbidities contributing to CKD. Methods A dynamic stock and flow model was developed to simulate CKD progression in the Chilean population aged 40 years and older, up to the year 2041, adopting the perspective of the Chilean public healthcare system. The model included six states replicating progression of CKD, which was assumed in 1-year cycles and was categorised as slow, medium or fast progression, based on the underlying conditions. We simulated two different treatment scenarios. Only direct costs of treatment were included, and a 3% per year discount rate was applied after the first year. We calibrated the model based on international evidence; the exploration of uncertainty (95% credibility intervals) was undertaken with probabilistic sensitivity analysis. Results By the year 2041, there is an expected increase in cases of CKD stages 3a to ESKD, ceteris paribus, from 442,265 (95% UI 441,808–442,722) in 2021 to 735,513 (734,455–736,570) individuals. Direct costs of CKD stages 3a to ESKD would rise from 322.4M GBP (321.7–323.1) in 2021 to 1,038.6M GBP (1,035.5–1,041.8) in 2041. A reduction in the progression rates of the disease by the inclusion of SGLT2 inhibitors and pre-dialysis treatment would decrease the number of individuals worsening to stages 5 and ESKD, thus reducing the total costs of CKD by 214.6M GBP in 2041 to 824.0M GBP (822.7–825.3). Conclusions This model can be a useful tool for healthcare planning, with development of preventive or treatment plans to reduce and delay the progression of the disease and thus the anticipated increase in the healthcare costs of CKD.


Results
By the year 2041, there is an expected increase in cases of CKD stages 3a to ESKD, ceteris paribus, from 442,265 (95% UI 441, 722)   Introduction the long-term pathways of CKD [31,32] can provide useful information for more effective healthcare planning, resource allocation, and ultimately the delivery of the best possible healthcare service to patients [30]. In this study, we develop a population model to project CKD in Chile and estimate the future economic and health consequences of CKD from the perspective of the Chilean public healthcare system.

Model construction specifications
A dynamic stock and flow model was developed to represent the natural history of CKD for the Chilean population aged 40 years and older with reduced kidney function from stage 3a to death (Fig 1) with a time horizon of 20 years, up to 2041, to provide sufficient results of future healthcare costs and cases for decision making [25]. The model included six mutually exclusive states replicating the progression of the disease: from CKD stage 3a through to stage 5, with or without diabetes mellitus, ESKD with need of RRT, and death. The CKD stages were defined based on the Kidney Disease: Improving Global Outcomes (KDIGO) [5] 2012 classification (S1 Fig): stage 3a: eGFR 45-59 ml/min/1.73m 2 with normal or increased albuminuria; stage 3b: eGFR 30-44 ml/min/1.73m 2 with normal or increased albuminuria; stage 4: eGFR 15-29 ml/min/1.73m 2 with normal or increased albuminuria, and stage 5: eGFR <15 ml/min/1.73m 2 with normal or increased albuminuria. We could not include CKD stages 1 or 2, defined as normal eGFR but with increased albuminuria [5], as this variable was not measured in the whole sample of ENS 2016-17, so we had to limit our analysis to stages 3a to ESKD. Therefore, we refer hereafter to CKD as signifying stages 3a to ESKD excluding stages 1 and 2. The full model can be found in S2 Fig. The progression of CKD was assumed based on the development of the disease and the available evidence reporting progression across the different stages. As time moves forward, a proportion of individuals remains in the same health state, some progress to the next CKD stage, began RRT or die [30], depending on the decrease of eGFR and mortality rates. Transitions were assumed to go in only one direction (Fig 1), due to the low possibility of regression or remission of the disease, and without skipping CKD stages [25]. Transition to death was possible from any of the other health states [31], depending on the difference in mortality rates between CKD stages, age of the population and presence of comorbidities. Death was considered an 'absorbing' state [30].
The progression of CKD stages depended on the annual decrease of eGFR reported in the literature, with decrements of 1.4, 3 or 5 ml/min/1.73m 2 per year (depending on the presence of diabetes mellitus and/or increased albuminuria) [30,[32][33][34]. The annual progression of eGFR assumed that individuals with CKD but with no diabetes or increased albuminuria would progress at a rate of 1.4 ml/min/1.73m 2 per year (slow progression); those with diabetes only would progress at a rate of 3 ml/min/1.73m 2 per year (medium progression); and those with diabetes and increased albuminuria in each stage would progress at a rate of 5 ml/min/ 1.73m 2 per year (fast progression) [30], replicating the proportion of individuals with diabetes and increased albuminuria in the Chilean population with CKD stages 3a to 5 (see Table 1). The proportion with hypertension was not included when determining the progression between stages, as 86% of the adults with CKD had survey-defined hypertension [35].
The model was developed from the perspective of the Chilean public healthcare system as CKD treatment is guaranteed for everyone diagnosed with CKD stages 3a to ESKD [36], independently of whether the individual belongs to the public or private healthcare sector. The model was limited to the Chilean population 40 years or older; it accounted for populationlevel heterogeneity in age, eGFR, albuminuria and diagnoses of hypertension and diabetes mellitus based on national data [9,18]. Ethnic differences were not included due to data limitations, and so the CKD-EPI equation used to determine eGFR assumed a white population [37].

Data sources for the model parameters
Model parameters needed to populate the model were as follows: initial CKD prevalence; CKD incidence; population growth; prevalence of comorbidities; and mortality rates. These are set out in Table 1. Below we briefly discuss each in turn.
Initial prevalence of CKD. Prevalence data of CKD stages 3a-5, obtained from the nationally representative Chilean national health survey 2016-2017 (Encuesta Nacional de Salud, ENS 2016-17), was used for the initial distribution of eGFR and increased albuminuria in each stage and for the proportion of diabetes in the population with CKD. The sampling design and methods of data collection of the ENS 2016-17 have been reported elsewhere [9]. The estimated number of individuals at each discrete value of eGFR was obtained by fitting a linear regression with the natural logarithm of the number of cases as the outcome (y) and the discrete values of eGFR (range 1 to 59) as the independent variable (x). The fitted equation was ln(y) = -0.0556 + (0.077 � x); predicted values were obtained by using the exponential transformation (S3 Fig). For example, the estimated number at eGFR = 59ml/min/1.73m 2 was equal to exp(-0.0556 + (0.077 � 59)) �90. By using the fitted equation of the discrete values of eGFR in the Chilean population, we could estimate the proportion of individuals in each stage that progressed based on the different progression rates assumed in the model.
In Chile, adults in need of RRT can receive kidney transplantation, haemodialysis (HD) or peritoneal dialysis (PD), although around 95% of Chilean patients in need of RRT are currently treated with HD and 4.5% with PD [16,18,19] with few patients receiving a transplant [16,18,38]. Thus, for this model, data about the ESKD group (initial prevalence, eGFR distribution, age, presence of comorbidities, mortality experience and costs) were obtained based on the information of patients having HD and PD from the Chilean Haemodialysis registries [16,18,19]. We therefore refer to RRT as signifying HD or PD excluding renal transplant.
Incidence of CKD. As one component of inflow, the annual incidence of CKD stage 3a among the general population and among those with diabetes was taken from the literature  (Table 1).
Mortality rate. All-cause mortality rate for the population 40 years and older was calculated using the Chilean data for all-cause mortality rate by age group provided by INE [41], and then adjusted with the hazard ratios of all-cause mortality per CKD stage found in the literature [42]. For adults with ESKD, the mortality rate was compared with the rate published in the Haemodialysis registries [16,18,19].

Inclusion of sodium-glucose cotransporter-2 (SGLT2) inhibitors and predialysis treatment
We simulated the effect of including the treatment with SGLT2 inhibitors for individuals with CKD stages 3a and 3b with diabetes mellitus [43,44]. Based on the available evidence, we assumed that this treatment decreased the progression of CKD (HR 0.71, 95% CI 0.57 to 0.89) [44,45]. The costs of treatment with SGLT2 inhibitors was provided by the Chilean Ministry of Health (calculated as £179.94 per patient annually).
Furthermore, we simulated the effect of including pre-dialysis treatment for individuals with CKD stages 4 and 5. Based on the evidence, we assumed that the inclusion of pre-dialysis treatment would reduce the progression of the disease (HR 0.85, 95% CI 0.74 to 0.98) [46]. The costs for the pre-dialysis treatment was estimated based on experts' opinion due to lack of Chilean data. The total annual cost per patient (in addition to the current treatment) was estimated at £360.97 and £1637.46 for stages 4 and 5, respectively.

Costs
Only direct costs related to the treatment of the disease were included in the model. We incorporated these on an annual basis to represent the evolution of the expected total cost from the CKD stages (3a-ESKD) indicated by eGFR <60 mL/min/1.73 m 2 (determined by CKD-EPI equation) with normal or increased albuminuria in accordance with KDIGO guidelines [5]. a Incidence of CKD stage 3a, given in per million population (pmp). b Fast, medium and slow progression considered as annual decline of eGFR of 5 mL/min/1.73m 2 , 3 mL/min/1.73m 2  perspective of the Chilean public healthcare system. The model was developed replicating the standard of CKD care in Chile. Currently, for stages 3a to 5, the Chilean Guidelines on CKD recommend multidisciplinary healthcare for patients [47]. All patients with stages 3a or 3b are usually diagnosed, managed and controlled in primary care by a multidisciplinary healthcare team and are referred to secondary care as the condition progresses to stage 4 and 5 [47]. For adults with ESKD, either HD or PD are provided, following the Chilean guidelines for this group of individuals [18,[48][49][50][51]. The annual cost per stage was estimated, and subsequently multiplied by the number of expected cases per year, thus obtaining the total cost per year for the total CKD population. All data about resource consumption and their costs were obtained from the Individual Expected Cost Verification Study (EVC) [52] and the National Health Fund (FONASA) [53]. Constant rates were applied throughout the 20-year simulation period, therefore the variation of costs per year was given by the variation in the number of individuals estimated for each stage. All costs were adjusted by the variation of the consumer price index, with values as at June 2019: the adjustment was based on June 2019 values due to the economic instabilities that occurred during the end of 2019 and throughout 2020 in Chile due to the social crisis and the Covid-19 outbreak. This model was intended as a baseline to project the future cases and costs of CKD in Chile, and according to Chilean guidelines, a 3% annual discount rate was applied to the costs. Costs are presented in GBP using the UK HM Revenue & Customs May 2021 exchange rate [54]. The total annual costs for each stage and associated reference are outlined in Table 2. The full details of the treatments, frequency of use and costs considered for the model can be found in the S1-S5 Tables.

Sensitivity analysis
Exploration of uncertainty in the modelled estimates was conducted with probabilistic sensitivity analyses (PSA). The PSA was undertaken whereby different probability distributions were associated with each parameter of the model depending on the data, using the mean value and the 95% confidence interval of the estimate [25]. For the costs data, we assumed a baseline cost ± 20% variation in the mean value to calculate the 95% confidence interval used in the PSA. The choice of the types of distributions was according to standard practice and to  [57]. We calibrated the model using international evidence and comparing the rates of increase in ESKD cases with the Chilean Haemodialysis registry [19]. All descriptive analyses to estimate the model parameters were adjusted for the complex survey design of the ENS and were performed using Stata V15.1 (StataCorp, College Station, Texas, USA). Stella Professional V 2.1 and Microsoft Excel Office 365 V2001 were used to construct the model and the different scenarios using the Visual Basic for Applications (VBA) macro fully parameterized to conduct the PSA. Figures were designed using Tableau Desktop Professional 2020.

Results
The results of the model show that the number of Chilean adults with CKD, ceteris paribus, are projected to increase continually in all stages to the year 2041 ( Table 3). The results from the PSA supported the continued growth of the CKD population in all stages as shown in Fig 2A.

Inclusion of SGLT2 inhibitors + pre-dialysis treatment
The results of including treatment with SGLT2 inhibitors for individuals with CKD stages 3a and 3b with diabetes and the introduction of pre-dialysis for individuals with CKD stages 4 and 5 are shown in Tables 3 (number of cases) and 4 (costs). With the inclusion of these treatments, the difference in the total number of cases of CKD by 2041 would be an increase of 13,447 (from 735,513 (734,455-736,570)) in the baseline scenario to 748,960 (748,023-749,896)), but with a marked change in the distribution of cases between CKD stages. Due to a higher percentage of people progressing at a slower rate, the number of cases in earlier stages (3a and 3b) at the end of 2041 would be higher than the baseline scenario and the total cases of stages 5 and ESKD would be fewer (Fig 3). This would reduce the total costs of CKD by 214.6M GBP projected for 2041 (from the 1,038.6M GBP in the baseline scenario to 824.0M GBP in the scenario with the inclusion of treatment with SGLT2 inhibitors and pre-dialysis) (Fig 4).

Discussion
To the best of our knowledge, this is the first population model to project CKD in the Chilean population and thus simulate the future CKD economic burden for the Chilean healthcare system. Using a dynamic stock and flow model, we represented the natural evolution of CKD in the Chilean population aged 40 years or older between 2021 and 2041. According to our  results, the projected number of Chilean adults with CKD stages 3a to ESKD would almost double in number in the next 20 years. This substantial increase in the population with CKD will increase healthcare resource utilization and overall costs accordingly [32]. Even though we included only direct costs of the treatment of CKD, our projection shows that they would triple by 2041. There are other important indirect costs such as hospitalizations and costs related to adverse events due to CKD that were not considered and these may have an important impact on the overall costs of the disease. Studies projecting the disease in other countries have found similar results, with an increase in the prevalence of CKD and ESKD in the general adult population [2,28,32]. The study by Hoerger et al. [32] projected an increase in the prevalence of CKD stages 3a to ESKD in the United States from 13.2% in 2010 to 16.7% in 2030. Chile has a younger population structure compared with the United States; therefore, we would expect this difference in CKD prevalence, but the increase in percentage points has a similar trend between both countries. Wong et al. concluded that by 2050, nearly 25% of the population in Singapore would have CKD [2]. This significant difference with our projections is probably due to the inclusion of CKD stages 1 and 2 in their model (i.e. normal eGFR but with increased albuminuria). ENS 2016-17 measured albuminuria only in individuals with diabetes and hypertension; thus we could not classify adults at CKD stages 1 and 2 for the model. Additionally, we did not have sufficient information about the trajectory of these earlier stages in terms of the natural course and progression of the disease to be confident about including them in the model, therefore we limited our projection only to the population with reduced eGFR (CKD stages 3a-ESKD) with normal or increased albuminuria and considered the proportion of individuals with increased albuminuria to determine different progression rates. Nevertheless, we emphasize the importance of considering all CKD stages to project the disease when national cohorts with long follow-up become available, to provide more reliable characterization of the relationship and progression of CKD and thus a more precise projection of it.
In our model we focus on CKD in adults aged 40 years or older because eGFR is relatively constant before that age [32]; eGFR begins declining around the age of 30-35 years so we estimated that relatively few cases were missed and this reduced the misclassification of cases in younger adults [6]. The risk of developing CKD and ESKD increases strongly with age [28,58], as it is related in part with the natural decline of kidney function, so the clinical significance of early stages of the disease (stages 3a-3b) has been debated for the more elderly population [6,59]. Nonetheless, we included this population in the model as the further decline in kidney function in this group resulting in progression to more advanced stages would lead to the need for more specialized medical treatment to prevent progression, or RRT in case they progress to ESKD, thus increasing healthcare resources utilization [42]. The overall Chilean adult population is projected to increase [41], leading to an expected increase in the burden of CKD in the following years [28].
Other risk factors behind the rise in CKD prevalence are probably the increase in the prevalence of diabetes, hypertension, and obesity in the general population. In Chile, according to the most recent health survey (2016-17), 12.3% of adults had survey-defined diabetes, 34.6% were obese and 27.7% had survey-defined hypertension [35]. These show a significant increase in diabetes and obesity compared with the 2009-10 survey, with prevalences of 9.4% and 25.7% respectively. These data support our projection of an expected increase in the number of individuals with CKD and emphasize the need to develop new interventions to slow the onset and progression of CKD at earlier stages and therefore reduce the incidence of ESKD [58]. As the results of the model simulations show, a reduction in the progression of the disease given by the estimated effect of the treatments: (1) SGLT2 inhibitors for those with diabetes mellitus in CKD stages 3a and 3b, and (2) pre-dialysis for those in CKD stages 4 and 5, would have a significant impact on the number of individuals advancing to later stages of the disease and therefore potential savings of around 214.6M GBP on direct costs for the Chilean healthcare system. These findings show the importance of targeting the progression of CKD as one of the key variables when establishing the treatment's goals. The evidence shows several options for effective treatments for the management of early stages of CKD to reduce progression of the disease such as the treatment with SGTL2 inhibitors for earlier stages of CKD [60][61][62][63] and the multidisciplinary pre-dialysis treatment for stages 4 and 5 [25,46], or conservative management for certain individuals with ESKD as an option instead of RRT [28,64]. These strategies need to be studied in future models with real world data to assess their implementation, cost-effectiveness and the impact on CKD in Chile.
Our results have several limitations. First, due to unavailability of Chilean cohort studies of CKD, the data sources used for the model may have introduced bias to our results. The crosssectional design of the ENS 2016-17 and the estimate of eGFR based on single-point-in-time measurements of serum creatinine, are possible sources of bias. Moreover, due to the limitations of the cross-sectional design of the Chilean data we did not include important risk factors for development and progression of CKD in Chile such as obesity [28,65], different ethnicities [66], socio-economic level [67], other clinical risk factors [68], proteinuria [2], or acute kidney injury [69,70]. Likewise, adverse events of CKD such as myocardial infarction, stroke or other cardiovascular diseases [71], although important comorbidities for CKD [39,71], were not included in this first population model for CKD stages 3a to 5. These complications will have an important impact in the number of cases and the costs of the disease; therefore, they should also be considered in a future modelling study when longitudinal data becomes available. Also, due to unavailability of Chilean renal transplantation data we could not include this treatment in the model and limited our simulation and analysis of RRT only to dialysis. Secondly, our model assumed two simplifications regarding the progression and trajectory of eGFR. We permitted the progression of the disease only by decrease in eGFR in time, without allowing possible regression of eGFR. Nevertheless, we considered that the influence would be negligible as the evidence shows that only in a minority of patients does eGFR improve and often they revert to CKD [25]. The trajectory of eGFR was assumed to decrease linearly at three constant rates with difference between stages depending on the presence of diabetes mellitus and/or increased albuminuria [30]. These simplifications may have introduced some uncertainty to our results, that we tried to overcome with the PSA, but robust longitudinal data sets with longer follow ups are needed to have a deeper knowledge of both the progression and trajectory of eGFR between stages.
Thirdly, our estimates are based on current risk factor prevalence, CKD incidence, and allmortality rates. If these rates change over time, with the inclusion of different interventions or with acute events such as the recent pandemic outbreak of Coronavirus (Covid-19), our estimates may either over-or under-estimate the CKD projections [32]. Therefore, these estimates must be treated with caution and must be validated when longitudinal data become available [30].
Our model provides essential information needed by decision-makers for future public healthcare planning; preparing resources needed; and taking effective actions to combat the problem of CKD in Chile. These results, although limited to the healthcare system perspective, indicate that the number of cases and costs of CKD would continue to increase in the future if no actions were taken. It is important to consider these results as part of a broader societal perspective, where CKD imposes large health and economic burden to individuals with the disease and their families, the national healthcare system and society (including the productivity loss of due to sick leave or early retirement). CKD and its risks factors such as diabetes and hypertension can be prevented or delayed, and therefore implementing effective prevention strategies to slow the increasing burden of the disease is an urgent public health priority.