The STARRT trial: a cost comparison of optimal vs sub-optimal initiation of dialysis in Canada

Abstract

Background:

Sub-optimal transitioning of patients from chronic kidney disease (CKD) to end stage renal disease (ESRD) may result in poor clinical outcomes and increased healthcare costs. The objectives of this study were to estimate the average total cost per patient who requires initiation of renal replacement therapy (RRT) stratified by status at initiation; optimal (RRT initiation as an outpatient with an arterio-venous [AV] Fistula, Graft or Peritoneal Dialysis [PD] catheter), and sub-optimal (RRT initiation as an inpatient and/or via central venous catheter [CVC]).

Methods:

Data from the Study To Assess Renal Replacement Therapy (STARRT), a Canadian, multi-centre, 6 month retrolective study (n = 339), were used for this analysis. Unit costs for resources were obtained from participating hospitals, the literature, and/or standard costing sources. The analysis was performed from the perspective of healthcare payors and reported in 2011 Canadian Dollars (CAD). A propensity score technique was applied to control for potential confounders between the two groups.

Results:

Two hundred of the eligible patients for analysis (61.9%) were sub-optimally and 123 (38.1%) were optimally prepared. For this analysis, 106 “matched” pairs were used. The average total cost per patient was estimated to be $63,225 (with a 95% CI ranging from $58,663–$67,958) for the sub-optimally initiated patients, and $39,260 (with a 95% CI ranging from $35,683–$43,007) for the optimally initiated patients (p < 0.001).

Limitations:

Costs were calculated utilizing a conservative approach, using the cheapest available prices for medications and other resources. Assumptions had to be made for the costing of dialyses.

Conclusion:

The results of this study indicate, after adjusting for potential confounders, that optimally initiated patients for RRT have significantly lower healthcare-associated costs compared to sub-optimally initiated patients.

Key words::

• cost comparison
• dialysis
• Canadian
• STARRT

Background

The transition from chronic kidney disease (CKD) to end stage renal disease (ESRD) presents a significant challenge to patients and caregivers, as sub-optimal transition of patients to renal replacement therapy (RRT) is linked to poor clinical outcomes including higher mortality rates1,2. Of importance and growing concern is the economic impact of sub-optimal initiation of dialysis on healthcare systems, as a result of longer hospitalizations and higher resource utilization.

The treatment options available for a patient with ESRD include renal transplantation, hemodialysis (HD), or peritoneal dialysis (PD)3. While renal transplant is associated with significantly better outcomes with respect to survival, health-related quality-of-life (HRQOL), and cost savings, only a minority of patients with ESRD are candidates for transplantation4. Those that do qualify are subject to long waiting lists for an appropriate deceased donor if a live donor is not available. As a result, most patients initiate RRT on dialysis, with in-centre hemodialysis being both the most commonly selected and most expensive option3.

In Canada, an estimated $55,466 is spent every year to manage one ESRD patient5. This represents ∼2–3% of the total healthcare budget, despite ESRD affecting only 0.02% of the Canadian population6. In-centre hemodialysis is the most expensive form of renal replacement therapy (RRT), estimated to cost up to ∼$59,000 per patient in 20047. These costs are further skewed when confounding factors such as age, gender, and comorbidities like diabetes, cardiovascular disease, and duration of the kidney disease are taken into account8–12.

Improving disease management of CKD and ESRD and optimizing the transition from CKD to ESRD may improve patient care; perhaps even lower morbidity and mortality rates—and reduce costs1,3,13. Our hypothesis was that patients starting renal replacement therapy in an optimal manner would accumulate lower healthcare-related costs than those starting in a sub-optimal manner.

Objectives

The objectives of this analysis were 2-fold: (1) To estimate the average total cost per patient over the first 6 months for patients who need to initiate RRT; (2) To stratify these costs by their initiation status, which include: optimally initiated (defined as RRT initiation as an outpatient and dialysis initiation with an arterio-venous [AV] fistula [AVF], graft [AVG] or peritoneal dialysis [PD] catheter) and sub-optimally initiated (defined as RRT initiation as an inpatient and/or initiation of hemodialysis [HD] with a central venous catheter [CVC]). Costs are presented from the perspective of the healthcare payor in 2011 Canadian Dollars.

Methods

Data source

The patients and resources used in this analysis were derived from the STARRT study14; a Canadian, multi-center, retrolective15 cohort study designed to assess a number of factors related to pre-dialysis care and patient status at the time of initiation of dialysis, which has recently been published. Briefly, subjects in the STARRT study were identified from charts of consecutive incident RRT patients from July 1st to December 31st, 2006. Information was collected until 6 months after the initiation or until death, renal transplant, or transfer to another CKD program. Data from the STARRT study were used to estimate and compare the average direct medical cost per patient. The study cohort for this analysis was divided into two groups: patients who were optimally initiated on RRT and patients who were sub-optimally initiated. More complete methods are available in the primary manuscript14.

STARRT study population

Data from 339 patients who started chronic RRT at 10 Canadian study centers were available. In order to be included in the STARRT study, subjects must have been at least 18 years of age or older, starting any form of chronic HD or PD, at home or in-center, or having pre-emptive transplantation. Subjects were excluded from the STARRT study if they required temporary RRT due to drug or environmental intoxication, acute renal failure treated in an intensive care unit, and those who started dialysis due to kidney transplant rejection.

Population for costing study

For the purpose of this analysis, subjects from the STARRT study were examined in two groups: Optimally initiated patients (i.e., RRT initiation as an outpatient with dialysis access using either an AVF, AVG, or PD catheter), or sub-optimally initiated patients (i.e., RRT initiation as an inpatient or HD initiation using a CVC). As 16 of the 339 patients in the STARRT study received transplants, data from 323 patients were examined in this analysis.

Propensity scores

As the data used for this analysis were not extracted from a randomized trial, a need exists to control for naturally occurring systematic differences in demographic characteristics between the optimally and sub-optimally prepared groups16. When comparing these groups, a propensity score technique was utilized as introduced by Rosenbaum and Rubin17. Propensity score methods calculate the likelihood that a person would start dialysis optimally based on their baseline demographics (known as their propensity score), then using this score in a regression model that adjusts the patient’s cost during RRT for their propensity to have started optimally. Patients with similar propensity scores have a similar likelihood of starting dialysis in an optimal fashion, regardless of whether they in fact started dialysis optimally, and so propensity scores can be used as a way of matching patients from each group.

The “nearest available matching” technique, as suggested by Rosenbaum and Rubin18, was applied for constructing the matched sample for this analysis. This method consists of randomly ordering the optimally and sub-optimally initiated subjects, then selecting the first optimally initiated subject and finding the sub-optimally initiated subject with the closest propensity score. Both subjects are then removed from consideration for matching, and the next optimally initiated subject is selected, as suggested by D’Agostino19. Before calculating the actual propensity scores, the predictors were sub-classified and checked for balance, as suggested by Rosenbaum and Rubin20.

The five a priori identified confounding covariates of interest for this analysis included: age, gender, presence of cardiovascular diseases, diabetes, and pre-dialysis care duration. All of these covariates are believed to influence the cost of CKD management.

Data from 212 subjects could be used to create 106 matched pairs to estimate the difference in average total cost per patient between the optimally and sub-optimally initiated patients requiring CKD management.

Data collection

Data collected for this analysis included patient demographics at study initiation (age and gender), prior and updated access information (by graft, fistula, or catheter), initial RRT and modality changes, RBC transfusions, hospitalizations, concomitant therapy to treat anemia, erythropoietin therapy, and all reported deaths.

Costing

Unit costs for resources were obtained from participating hospitals, the literature, and/or standard costing sources (i.e., provincial fee schedules). The analysis was performed from the perspective of a Canadian healthcare payor and reported in 2011 Canadian Dollars (CAD).

If unit costs for resources were published prior to 2011, a cost calculator21 was used to adjust the cost based on the reported prices.

For this cost comparison, only resources utilized during enrollment in the STARRT study were included. The start day of enrollment into the study was considered the date when RRT was initiated. Study termination was considered either 6 months from the start date, or earlier if the patient was terminated due to reasons such as death, transplant, transfer out of a program, or lost to follow-up.

Table 1 presents the healthcare resources included for this analysis as well as the assigned unit costs and the sources for assigning the unit costs per resource.

Table 1.  Unit costs of healthcare resources.

Healthcare resource	Unit cost	Cost information source
Lab cost/week (average cost for all tests and all patients)	$29.50	McFarlane and Redelmeier22
RBC Transfusion (Red Blood Cells Only) per one unit	$731.00	Dranitsaris23
Daily hospital cost per patient	$906.00	Median cost as calculated from the six sites who provided their site-specific daily average cost per patient
Medications
Iron-iv
Venofer 100 mg/5 mL Single Dose Vial	$57.63	Panthermeds24
Ferrlecit (12.5 mg/ml)	$23.44	RAMQ25
Dextran (50 mg/ml)	$27.50	RAMQ25
Iron-oral
Iron (Ferrous) Gluconate (300 mg/tablet)	$0.02	OPDP Drug Benefit Formulary26
Iron Ferrous Fumarate (300 mg/tablet)	$0.12	OPDP Drug Benefit Formulary26
Iron Sulfate (300 mg/tablet)	$0.05	OPDP Drug Benefit Formulary26
Vit-B12 (Cyanocobalamin, 1000 mcg/ml)	$3.10	OPDP Drug Benefit Formulary26
Folate-Folic Acid (5 mg/tablet)	$0.02	RAMQ25
REPLAVITE (1 mg/tablet)	$0.10	Canadian Pharmacy27
INFUFER Liquid (50 mg/ml in 2 ml)	$27.50	RAMQ25
Eprex (erythropoietin)
20,000 IU/ml	$267.90	OPDP Drug Benefit Formulary26
40,000 IU/ml	$401.85	OPDP Drug Benefit Formulary26
Aranesp (darbepoetin alfa)
150 mcg/0.3ml	$402.00	OPDP Drug Benefit Formulary26
200 mcg/0.4ml	$536.00	OPDP Drug Benefit Formulary26
Access
Graft	$6812.13	calculated from Lee et al.3
Cuffed CVC	$7864.32	calculated from Lee et al.3
CVC without cuff	$7864.32	assumed same as Cuffed CVC
Fistula	$2872.51	calculated from Lee et al.3
PD catheter	$2527.51	calculated from Lee et al.3
CVC Temp Line	$7864.32	assumed same as Cuffed CVC
Femoral Access	$7864.32	assumed same as Cuffed CVC
PICC LINE	$7864.32	assumed same as Cuffed CVC
Fistula RE-DO	$2872.51	assumed same as a first time Fistula
Modality
Hemodialysis
In centre (in-center hemodialysis)/week	$1523.00	McFarlane and Redelmeier22
Short daily/week	$1008.00	McFarlane7
Satellite	$1201.38	calculated from Lee et al.3
Peritoneal dialysis
CAPD	$1160.00	Goeree28
CCPD	$1260.00	Coyte29

All costs are reported as 2011 CAD.

RAMQ, Régie de l’assurance maladie du Québec; OPDP, Ontario Public Drug Programs; CAPD, Continuous Ambulatory Peritoneal Dialysis; CCPD, Continuous Cycling Peritoneal Dialysis.

Treatment cost (concomitant medication and erythropoietin therapy)

The total cost of medications per patient was calculated by multiplying the unit cost per dose by the total number of doses utilized during the study period (see Table 1 for cost details). Treatment costs also included costs of concomitant medications to treat anemia (such as iron, vitamin B-12, or folic acid) and erythropoietin therapy (Eprex® or Aranesp®). Other medications as reported in the patients chart were included.

Access cost

Access included the cost of Graft, Fistula, Cuffed CVC and CVC without cuff, PD Catheter, CVC temporary line, femoral access, and peripherally inserted central catheter line. The total cost for access was calculated as the sum of the individual costs for all access change procedures performed during the study period (prior access before RRT initiation was excluded). See Table 1 for cost details.

Modality cost

Modality included the cost of conventional HD either performed in-hospital or in satellite centers, short daily HD, and PD of two types, Continuous Ambulatory Peritoneal dialysis (CAPD) and Continuous Cycler-Assisted Peritoneal Dialysis (CCPD). The total cost for modality was calculated by multiplying the average cost per week of the specific modality by the number of weeks the modality was utilized during the study period. See Table 1 for cost details.

RBC transfusions cost

The total cost for RBC transfusions per patient was calculated by multiplying the RBC cost per unit by the total number of units utilized during the study period. See Table 1 for cost details.

Hospitalization cost

Six of the 10 sites who participated in the STARRT study provided site-specific average costs for treating one patient for 1 day at their site. The median cost was calculated and used as the average per patient daily hospital cost for this analysis. The total cost for hospitalization per patient was calculated by multiplying the average daily hospital cost by the length of stay (days) in hospital during the study period (including initial hospitalization for sub-optimally initiated patients). No hospital costs were assigned when the admission and discharge to the hospital occurred on the same day.

Laboratory cost

The total cost for laboratory per patient was calculated by multiplying the average laboratory cost per week (as reported by McFarlane and Redelmeier22) by the number of weeks of enrollment in the study. No individual laboratory data were collected.

Secondary analysis

In order to apply the identical criteria for sub-optimal initiation of dialysis set out in the original STARRT study14, a secondary analysis was conducted, excluding the costs associated with the initial hospitalizations for sub-optimally initiated patients.

Statistical analysis

The analysis was firstly based on univariate statistical procedures that include: frequency counts; capturing basic statistics; cross-tabulations for categorical variables; t-tests, and correlations (Pearson for continuous variables and Spearman’s Rho for categorical variables). Where necessary, transformations of variables with a non-normal distribution were applied.

Thereafter, multivariate statistical procedures were employed: logistic regression in the case of a dichotomous criterion (i.e., optimally prepared group vs sub-optimally prepared group for the purpose of creating propensity score) and multiple linear regression including general linear models in the case of a continuous criterion (i.e., cost). Analyses were performed using SPSS, version 19.1.

Results

Data from 323 patients were used to estimate the average total cost per patient from the initiation of RRT. As previously reported, 200 (61.9%) patients were considered sub-optimally initiated, while the remaining 123 (38.1%) were identified as optimally initiated. Approximately two-thirds of the subjects (202/323 or 62.5%) were males. The mean age of the subjects was 64.4 years (SD ± 15.0), with a range of 19–92. For details see the top half of Table 2.

Table 2.  Group comparisons before and after matching for potential confounders.

Confounders/predictors	Group comparisons
	Sub-optimal		Optimal		Comparisons
	Mean	SD	Mean	SD	t (p)	Standardized difference in %^a
Prior to “matching”
n	200		123
Gender (Female = 0, Male = 1)	0.60	0.491	0.67	0.473	−1.201	−14.5
Cardiovascular Disease (No = 0, Yes = 1)	0.62	0.487	0.44	0.498	3.216***	36.5
Diabetes (No = 0, Yes = 1)	0.42	0.495	0.41	0.495	0.095	2.0
Pre-dialysis Duration ( ≤ 1 year = 0; > 1 year = 1)	0.55	0.498	0.76	0.431	−3.709***	−45.2
Age	65.89	14.956	61.98	14.892	2.287*	26.2
Logit of the Propensity Score	0.35	0.147	0.45	0.141	−5.845***	−68.1
After “matching”
n	106		106
Gender (Female = 0, Male = 1)	0.61	0.489	0.66	0.476	−0.712	−10.4
Cardiovascular Disease (No = 0, Yes = 1)	0.48	0.502	0.51	0.502	−0.410	−6.0
Diabetes (No = 0, Yes = 1)	0.41	0.493	0.42	0.495	−0.139	−2.0
Pre-dialysis Duration ( ≤ 1 year = 0; >1 year = 1)	0.74	0.443	0.72	0.453	0.307	4.5
Age	63.42	16.098	63.19	14.744	0.111	1.5
Logit of the Propensity Score	0.50	0.030	0.50	0.031	−0.883	−13.1

^aThe mean difference as a percentage of the average standard deviation.

*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

A total of 178 subjects (55.1% or 178/323) had a history of cardiovascular disease and 135 (41.8%) suffered from diabetes. Approximately one-third of the subjects (103/323 or 31.9%) had no history of cardiovascular disease or diabetes, whereas 93 subjects (28.8%) presented concomitantly with both comorbid conditions. With respect to pre-dialysis care, 201 subjects (63.2%) received nephrology follow-up for more than 1-year. Twenty-three subjects (7.1%) died during the study period.

The average number of days subjects were enrolled in the STARRT study was 162.5 days (SD ± 46.3) with a range of 1–203 days.

Results based on propensity scores

The top part of Table 2 illustrates the descriptive statistics of the five covariates and the logit (function/formula used to estimate propensity score, as described by D’Agostino19) for the optimally and the sub-optimally initiated groups before matching (n = 323). As illustrated, four of the five a priori identified confounding covariates (age, existence of cardiovascular diseases, diabetes, and pre-dialysis duration) were significantly different between the two groups. The bottom half of Table 2 illustrates that after “matching” the groups in 106 pairs, none of the identified confounders remained statistically significantly different.

The sample size for this analysis was reduced to 212 subjects (106 pairs), as 94 sub-optimally initiated subjects and 17 subjects in the optimally initiated group could not be matched.

The propensity analysis resulted in matched samples which had similar means for each of the five identified confounding covariates.

Patients demographics

Amongst the sub-set of the 212 included subjects, 76 subjects (35.8%) did not suffer from either cardiovascular disease or diabetes, whereas 56 subjects (26.4%) had a diagnosis of both. The pre-dialysis nephrology follow-up for 154 subjects (72.6%) consisted of more than 12 months of care. Thirteen subjects (6.2%) died during enrollment in the STARRT study, the majority (11 or 84.6%) were considered to be sub-optimally initiated on RRT. The mean age of the matched sample was 63.3 years (SD = 15.4) with a range of 20–92.

Resources and costs

Table 3 suggests that while patients optimally initiated on RRT accumulated more treatment days compared with patients sub-optimally initiated as a result of lower mortality (172.6 > 158.6 days), the expenses (square-root) associated with their overall cost per patient was lower (206.0 [with a 95% CI ranging from 199.2–216.7], compared to 241.6 [with a 95% CI ranging from 229.5–254.7], p < 0.001) in the sub-optimally initiated group. Table 3 also illustrates that the number of comorbidities, the number of renal diseases, and age at RRT initiation were all similar between the optimally and sub-optimally prepared groups (for details see Table 3). The captured CVD co-morbidities included hypertension, heart disease, peripheral artery disease, and stroke. All those were combined into the “category” of CVD and used for matching the groups, as well as diabetes. The main renal disease of study participants was glomerulo-nephritis.

Table 3.  Difference between the optimally and the sub-optimally initiated groups in the analysis of 106 matched pairs.*

Predictors	Group	n	Mean	SD	t
Number of comorbidities	Sub-optimal	106	2.55	1.38	0.184
	Optimal	106	2.51	1.59
Number of renal diseases	Sub-optimal	106	1.09	0.43	−0.149
	Optimal	106	1.10	0.50
Number of days in study	Sub-optimal	106	158.59	50.00	−2.426^b
	Optimal	106	172.61	31.85
Age (years) at RRT initiation	Sub-optimal	106	63.42	16.10	0.111
	Optimal	106	63.19	14.74
SQRT of overall cost	Sub-optimal	106	241.55	62.22	4.612^a
	Optimal	106	206.00	49.27

*Results in this table are based on univariate analysis.

^ap ≤ 0.001; ^bp ≤ 0.05.

The average cost per patient for the original sample (n = 323) was determined to be $54,020. After matching (n = 212), the average cost per patient was calculated as $50,532. In both cases, the average cost per patient was not normally distributed (i.e., Skewness > 2.00 and Kurtosis > 5.00), the square-root transformation was used for further analyses.

Table 4 lists the statistically significant predictors of the square-root transformation of the overall cost per patient, demonstrating that the most important predictor for the average cost is the length of time in RRT. The longer a subject was enrolled in the trial, the resources he/she utilized resulted in higher costs. The second most important contributor was determined to be the optimal or sub-optimal status at initiation. The costs associated with optimal RRT initiation were significantly less than those incurred by the sub-optimally initiated group.

Table 4.  Exploring the effects of various predictors on (square root of) total cost using regression analysis (n = 212).*

Predictors	Unstandardized coefficients		t (p)	95% CI for B
	B	SE		Lower	Upper
Group (0 = Sub-optimal, 1 = Optimal)	53.31	7.16	7.445^a	39.19	67.42
Died during RRT (0 = Yes, 1 = No)	−60.59	15.68	−3.864^a	−91.51	−29.68
Number of days on RRT	0.98	0.11	9.160^a	0.77	1.19
Group × Diabetes	29.80	8.72	3.418^a	12.61	46.99
Pre-dialysis Duration × Number of days in RRT	−0.17	0.08	−2.089^c	−0.34	−0.10
Pre-dialysis Duration × Age	0.58	0.20	2.854^b	0.18	0.97
Intercept (Constant)	81.42	17.36	4.690^a	47.19	115.65

^ap ≤ 0.001; ^bp ≤ 0.01; ^cp ≤ 0.05.

×, Interaction.

*Results in this table are based on General Linear Modelling

Other findings from the analyses were that: (a) patients with a pre-dialysis care duration of more than 1 year stayed longer in RRT and utilized more resources resulting in higher cost; (b) patients with a pre-dialysis duration of 1 year or less were younger and cost less; and (c) the average length of stay in hospital for the optimally prepared group was 6.5 (±29.5) days compared to 19.5 (±31.9) days for the sub-optimally prepared group (p = 0.01).

Finally, before controlling for the covariates, the average cost for the matched sample (n = 212) was $50,532 (with a 95% CI ranging from $47,109–$54,075). The average cost for a sub-optimally initiated subject was $58,383 (with a 95% CI ranging from $52,684–$64,375), and $43,248 ($39,689–$46,959) for an optimally initiated subject. After controlling for the identified covariates, the difference between the groups increased; $63,225 (with a 95% CI ranging from $58,663–$67,958) for the sub-optimally initiated group compared with $39,260 (with a 95% CI ranging from $35,683–$43,007) for the optimally initiated group (p < 0.001). Please see Table 5 for a summary.

Table 5.  Cost comparison of sub-optimally and optimally prepared subjects—before and after matching (n = 212) for main analysis.

	Sub-optimally prepared $ (95% CI)	Optimally prepared $ (95% CI)
Average cost before matching	$58,383 ($52, 684–$64, 375)	$43,248 ($39, 689–$46, 959)
Average cost after matching	$63,225 ($58, 663–$67, 958)	$39,260 ($35, 683–$43, 007)

Secondary analysis

Excluding the cost of the initial hospitalizations for the sub-optimally initiated patients reduced the average cost for the matched sample (n = 212) to $48,110 (with a 95% CI ranging from $45,172–$51,048). After controlling for the identified covariates the average cost for the sub-optimally initiated group was reduced to $55,655 (with a 95% CI ranging from $51,990–$59,319), lower than in the main analysis ($63,225), but still significantly (p < 0.001) higher than the optimally initiated patients ($39,260).

Discussion

The results of this analysis demonstrate that the total cost over the first 6 months for subjects who initiate RRT sub-optimally (RRT initiation as an inpatient and/or via CVC) are significantly higher when compared to those patients who were considered to be optimally initiated (RRT initiation as an outpatient and with an AV Fistula, Graft, or PD catheter).

By using the propensity score technique, an effort was made to eliminate possible confounding factors, hoping that the reported difference in cost between the two groups ($63,225 and $39,260 for sub-optimally and optimally initiated patients, respectively) can be attributed to RRT initiation status alone.

A study published by McFarlane7 in 2004 estimated the cost (in 2003 $US) for conventional home hemodialysis at ∼$35,000, for nocturnal hemodialysis at ∼$49,000, and for in-center hemodialysis at ∼$59,476, which are in range with the costs estimated in the current study.

Sixteen subjects from the STARRT study (n = 339) who received renal transplantations (both pre-emptive and non-pre-emptive) were excluded from this economic analysis. The primary reason for exclusion was that the transplantations occurred outside enrolment in the STARRT study and that the transplant patients appear to be different from the rest of the study cohort with regard to age and co-morbidities (they were much younger with less co-morbidities), both of which, as reported, impact the cost of CKD management. Furthermore, it is well known that transplant is the best and most economical modality of RRT4.

A total of 323 subjects (23 died, 21 of them in the sub-optimally initiated group) were included for the matching. After matching, data from two patients who died in the optimally initiated group and 11 in the sub-optimally initiated group were included for analysis. Should subjects who died consume less resources, as they did not complete the 6-month period of the study (which might not be the case as patients can be resource intensive in the weeks/months prior to death), the introduced bias would be conservative, as more patients who died were part of the sub-optimally initiated group.

Our results also demonstrated that optimally initiated subjects were, on average, hospitalized for a significantly shorter period of time compared with the sub-optimally initiated subjects, and utilized far fewer healthcare resources, resulting in a lower overall cost for CKD management.

About 60% of our study cohort was sub-optimally initiated. The estimated difference in cost between an optimally and sub-optimally initiated patient was ∼$24,000. There were 5431 ESRD patients who initiated RRT in 2008 in Canada30. If all patients and not only ∼40% of them would have been optimally initiated for RRT, the Canadian healthcare system could have potentially saved ∼$78 million for the year.

Assumptions and limitations

A conservative approach was used for calculating the cost for medications, as always the cheapest available price was used. Prices for resources were taken from Standard Costing Sources such as ODB or RAMQ where possible. For the calculation of the access cost, an assumption had to be made. For example, CVC without Cuff was assumed to be the same as cuffed CVC and the cost of femoral access was assumed to be the same as cuffed CVC. The cost of conventional HD performed at a satellite center was derived from the published literature3. Because the cost per session of HD was not found, the average cost of HD per week as published by McFarlane and Redelmeier22 was used, not accounting for the specific number of sessions per week or the number of hours per session a patient in the STARRT study utilized. Because the cost of PD based on volume of dialysate was not found, the average cost of PD per week as published in the literature was used28.

Treatments for which start or end dates were missing were assigned a value of $0, which again is a conservative approach. Of the 212 subjects in the matched sample, 17 (10 in the optimal and seven in the sub-optimal prepared group) had assigned treatment costs of $0.

Only hospitalizations that occurred during enrollment in the STARRT study were included. Six of the 10 sites which participated in the STARRT study provided site-specific average per patient daily hospitalization costs. However, different sites included different resources and/or services in their cost; therefore, the median value of the site-specific average daily per patient costs was used as the daily hospitalization cost for all the hospitals in this study (including the four sites that did not provide site-specific costs).

A published weekly average cost was used to calculate the total laboratory cost for each patient22. No individual laboratory test data were collected, due to the low costs compared to the other resources.

The secondary analysis was conducted as a sensitivity analysis, to reflect the approach used in the original STARRT study, which excluded the initial hospitalization to report clinical outcomes14. From an economic perspective, the initial hospitalization still resulted in resource utilization and associated costs, and was therefore included in the primary analysis. However, even after excluding the initial hospitalizations from the costing analysis, the difference between the optimally and sub-optimally initiated groups remained robust and statistically significantly different.

Conclusions

The results of this study indicate, after adjusting for potential confounders, that optimally initiated patients for RRT have significantly lower healthcare-associated costs compared to sub-optimally initiated patients.

Transparency

Declaration of funding

This research was funded by Janssen Inc. The sponsor provided the data for this analysis.

Declaration of financial relationship

PIVINA Consulting Inc. was hired by Janssen Inc. to conduct this research. Dr Fernando Camacho analyzed the original STARRT study data as a consultant to Janssen Inc. Dr Lou Marra and Farah Jivraj (at the time of conducting the study) are/were working for Janssen Inc. Drs Phil McFarlane and David Mendelssohn were principal investigators of the original STARRT study and consultants for Janssen Inc.

Acknowledgements

No non-author assistance in the preparation of this article is to be declared.