Patients with kidney failure live longer with a better quality of life after kidney transplantation when compared with maintenance dialysis.1,2 Recipient outcomes are further improved when the transplant comes from a living rather than deceased donor.3-5 As shown in modeling studies, an increase in the rate of living donor kidney transplantation is an effective strategy to ameliorate the burden of kidney disease, and remains cost-effective even if donors are paid.6-17 Additional healthcare resources, however, are needed to evaluate, perform donor nephrectomy, and follow living kidney donors after donation.18,19
From a health system payer perspective, an accurate estimate of the costs of living kidney donation is important for several reasons. First, a better understanding of the true healthcare costs of donation would improve estimates regarding incremental costs and benefits of living donor kidney transplantation. Second, as countries across the globe seek to better address the demand for transplantable kidneys, a thorough understanding of donation-related healthcare costs will help project the anticipated expenses that could occur with initiatives aimed at increasing rates of living kidney donation. Third, detailed cost estimates would better inform the funding allocated to hospitals or clinics which provide the service. Finally, a better understanding of current costs may serve as an important baseline measure for future efforts to improve the efficiency and cost of the donor candidate evaluation.
Many prior studies only considered the surgical costs of donation and did not fully account for the additional costs of the evaluation and follow-up care of living kidney donors.9,16 Studies that did include donor evaluation costs used prespecified donor evaluation protocols of minimum required testing.6,7,15 This method of costing underestimates the cost of donation because donors may require repeat tests, additional tests due to incidental findings, and tests that are not standard to donor evaluation protocols but are needed because of the donor's personal medical history.
To contribute to the literature, we conducted this detailed costing study in a universal healthcare system where most healthcare resource use and costs are incurred by a single payer. We investigated donor costs in 3 time periods: the predonation evaluation period (beginning of the donor evaluation until donation), the perioperative period [the nephrectomy and the perioperative period (30 days postdonation)] and the follow-up period (after the perioperative period until 1 year after donation).
MATERIALS AND METHODS
Design, Population, and Setting
This was a retrospective analysis of living kidney donors who donated at 1 of Ontario's 5 transplant centers between April 1, 2004, and March 31, 2014.20,21 Ontario residents have access to universal health insurance coverage through a public payer system, which includes all aspects of predonation and postdonation care. All living donors in this study were required to be Ontario residents for at least 2 years before donation. As per donor evaluation criteria, all donors had at least 1 nephrology consult and 1 surgery consult during the evaluation period (described in detail previously21).
The cost of living kidney donation was estimated separately for 3 periods, corresponding to 3 phases of the donation process: (1) the evaluation period (from the date the donor started the evaluation until the day before donation), which captures costs associated with the living donor assessment; (2) the perioperative period (from the day of nephrectomy to 30 days postdonation), which captures costs related to the donor surgery, hospitalizations, and any possible perioperative complications (including early readmissions); and (3) the 1-year follow-up period (from day 31 postdonation until 1 year postdonation), which includes costs related to longer-term or ongoing complications and any routine plus as-needed follow-up care.
Data Collection and Costing Sources
All costs were measured from the perspective of the Canadian payer. Donors were identified from the Trillium Gift of Life Network (TGLN) database through the Institute for Clinical Evaluative Sciences (ICES).20 Several health administrative datasets at ICES were used to link the data using unique encoded identifiers. These databases included the Ontario Health Insurance Plan, which captures all primary care and specialist physician billings, Canadian Institute for Health Information Same-Day Surgery and Discharge Abstract Database (hospitalizations); National Ambulatory Care Reporting System (emergency visits); Ontario Drug Benefits (prescription drug costs for citizens 65 years of age and older or receiving social assistance); National Rehabilitation Services; Complex and Continuing Care; and Long-Term Care. The ICES-derived costing method was used to obtain all costs from the various linked databases for a specified period (inpatient and outpatient costs for the periods described above). In addition to individual billing, this costing method uses resource intensity weights multiplied by the cost per weighted case to derive the cost per case.22
We derived the frequency of each healthcare procedure received and calculated the cost of each procedure using the physician claims database codes deemed relevant to the donor evaluation.21 These costs were totalled for each donor’s evaluation period (we restricted this to the evaluation period only as we could not prespecify relevant healthcare use during the perioperative and follow-up periods). To estimate the total cost of donation, all costs (instead of prespecified procedures) from the above databases were summed over each costing period. By including all costs accrued by the living donor, there is risk that some costs may have accrued for reasons unrelated to the evaluation (ie, consulting the general physician for a nonspecific viral illness). To account for this potential overestimation of costs, a baseline non–donation-related healthcare cost was estimated using a cohort of matched healthy nondonor controls (ie, individuals with similar indicators of baseline health as the donors; described in the Matching section below). To estimate the cost of donation with the baseline cost (the cost of the controls) removed, we developed a series of regression models (described in Statistical Methods below).
All Ontario residents were considered possible controls if they were alive, younger than 80 years as of April 1, 2006, were not missing sex, and had no prior history of living kidney donation themselves. The eligible 17092895 control candidates were assigned a random date (a fake “donation date”) to match the distribution of donation dates observed in the donors. Controls who were >79 years or died before their assigned donation date were excluded (ineligible to donate), resulting in 16 640 699 potential controls. Since donors are a highly selected healthy subset of the population, controls with any diagnostic, procedural, or intervention codes which suggested ill health or a contraindication to donation were excluded. These included codes related to dialysis, cancer, cardiovascular disease, human immunodeficiency virus, nephrectomy, renal biopsy, pulmonary disease, liver disease, systemic lupus erythematosus, rheumatoid arthritis, genitourinary disease, or alcoholism (full list of codes in Table S1, SDC,https://links.lww.com/TP/B559).
Potential controls were excluded if they were not Ontario residents for at least 2 years before their donation date or gave birth between 2 months before and 6 months after the donation date (similar exclusions were previously applied to the donors). A total of 6 151 385 potential controls and 1214 donors were available for matching (not missing matching covariates). Matching was done by donation date (±6 months), age at donation date (±2 years), sex, rural/urban status, and neighborhood-level income quintile. Four controls were matched to each donor.
To estimate the cost of living donation, we used a series of multivariable regression models applied to the matched cohort and conducted various statistical tests to assess the fit of each model (described below). This approach is recommended for cost data because of its positive and skewed distribution.23,24 Covariates included an indicator for donor/control status, age at the donation date, sex, urban/rural status, the year of donation (2003-2007, 2008-2010, and 2011-2014), neighborhood-level income quintile, and the total evaluation time in months. The effect of a variable on costs was reported using the marginal effects postestimation procedure in STATA, which uses the method of recycled predictions to provide estimates of mean cost. We were interested in the marginal effect of the dichotomous variable donor/control status, which represents the additional cost associated with living kidney donors (compared to controls). The marginal effect of any other covariate is interpreted as the incremental cost associated with a change in 1 unit of that covariate, holding the other factors constant. To assess whether the cost for donors was different from controls across levels of a covariate, we introduced an interaction term with the donor indicator in a separate model. A significant interaction term (Pint < 0.05) means that the cost of the donor is significantly different from controls across levels of the covariate included in the interaction. In the case where the predictor is only present in donors (ie, recipient’s dialysis status,21,25 transplant center), then the analysis is automatically restricted to donors only and the marginal effect is reported (Pint is not available).
We tested and compared the fit of ordinary least squares regression on untransformed costs, log-transformed costs, and square-root-transformed costs. Nonlinear models included the exponential conditional means model and Poisson regression with maximum likelihood estimation. We fit a variety of generalized linear models (GLMs) using a combination of link functions (identity, square-root, log) and distributional families (Gaussian, gamma, inverse Gaussian, and Poisson). We also attempted to fit the generalized gamma model assuming homoskedastic and heteroskedastic versions, extended estimating equations, and 2-component finite mixture models using gamma distributions.24 If more than 1 GLM had similar indices of fit, we performed a Park test to choose the best fit model. We also explored various statistical indices of badness of fit, including the Pregibon link test, a modified Hosmer-Lemeshow test, and the Pearson correlation P value.23,24 For each of these tests, a higher P value is more desirable. We also considered statistics such as R2, root mean square error, and mean absolute percentage error, where a lower value is desired. To guard against potential overfitting, we assessed statistics after cross-validation, including root mean square error and mean absolute percentage error (lower is better), mean prediction error (zero is desirable), and the P value for the Copas test (a test for overfitting; higher is desirable).23,24 Ultimately, GLMs were selected for each analysis: log-normal for the evaluation period, square-root Poisson for the perioperative period, log-gamma for the follow-up period, and square-root-gamma for the all periods together. Robust standard errors were calculated in all models to accommodate clustering by transplant center where the donation occurred.
We reported mean (standard deviation [SD]), median (25th, 75th percentile), and the mean difference (95% confidence interval [CI]) from marginal effects, where appropriate. All costs are reported in 2017 Canadian dollars.
As of April 1, 2006, the ICES-derived costing method improved to accommodate primary care physician payments under capitation following primary care reform to the costing data, in addition to dialysis facility visit costs and cancer clinic visit costs.22 We conducted sensitivity analysis in the subset of our cohort with an evaluation start date of April 1, 2006, or later using the updated costing method to accommodate these costs. Sensitivity analyses were also performed on the model chosen to estimate the donor costs since more than 1 model may fit the data for some analyses.
Software and Privacy
We used Statistical Analysis Software SAS v9.4 or SAS Enterprise Guide 6.1 (2013 SAS Institute Inc., Cary, NC) and STATA v13.0 (StataCorp LP, College Station, TX). Procedures and consultations for 5 or fewer donors are not reported to comply with privacy requirements for minimizing the chance of patient identification. The study was approved by the research ethics board at Sunnybrook Health Sciences Centre, Toronto, Canada.
We identified 1256 living kidney donors who completed the evaluation and donated in Ontario during the study period (Figure 1). Donors had a median of 28 (20-39) healthcare procedures (tests and consults) during the evaluation phase, performed during a median 16 (11-24) separate visits (which meant the procedures/tests were performed on different dates). Donors were a mean 45 (SD, 11) years of age, were mostly female (63%), white (78%), lived in urban areas (87%), and lived in higher-income neighborhoods (23% in the highest quintile vs 15% in the lowest).
Healthcare Utilization Patterns
Table S2, SDC (https://links.lww.com/TP/B559) presents the list of procedures considered relevant to living kidney donation and the frequency of use during the evaluation (1256/1256, 100%), donation (1240/1256, 99%), and follow-up (1223/1256, 97%) periods. The most costly healthcare procedure during the evaluation was consultation with a nephrologist, with a mean 1.91 consultations per donor at a mean cost of $135 per consultation, accounting for 15% of the total evaluation cost (Tables S3-S4, SDC,https://links.lww.com/TP/B559). The second most costly test was computed tomography, which were performed a mean 1.08 times per donor, representing 10% of the total evaluation at mean a cost of CAD $170 per examination. This was followed by nuclear medicine glomerular filtration rate test (9.0%), consultation with a surgeon or urologist (8.1%) and bloodwork (7.9%).
Cost of the Living Donor Transplantation Process
A total of 1099 (91%) of 1214 donors were successfully matched to 4396 controls (4 controls per donor; Figure 1). The matched donors were similar to the unmatched donors with respect to age (P = 0.47), sex (P = 0.27), urban status (P = 0.82), and neighborhood income quintile (P = 0.13) (Table 1). The mean total healthcare costs during the evaluation period for donors and their matched controls were CAD $4522 (SD, CAD $1073) and CAD $881 (SD, CAD $3061), respectively. The mean adjusted total cost attributable to the donor evaluation process was CAD $3596 (95% CI, CAD $3,350-CAD $3,842) (Table 2). Similarly, the mean cost attributable to the perioperative period was CAD $11,694 (95% CI, CAD $11 415-CAD $11 973) and the mean cost attributable to follow-up in the first year after donation was $1011 (95% CI, CAD $793-CAD $1230). The incremental cost of living donor-related care to the payer across all observation periods was CAD $16 290 (95% CI, CAD $15 814-CAD $16 767) (Table 2; Figure 2A-D). Using prespecified healthcare procedures, the cost of the living donor evaluation was a mean CAD $2108 (SD, CAD $968) (n = 1214).
Predictors of Costs
Over the total donation period, healthcare costs were higher for women (CAD $534 [CAD $179-CAD $890] higher than men), older persons (CAD $316 [CAD $172, CAD $460] per 10-year increase in age), and over a longer predonation evaluation period (eg, a longer window for healthcare utilization; CAD $52 [CAD $46-CAD $58) per month] (Table 3). Healthcare costs were lower in more recent years (−CAD $718 [−CAD $1217, −CAD $218] in 2011-2014 compared with 2004-2007), healthcare costs did not differ by the person's neighborhood income quintile (P = 0.66) or urban versus rural residence (P = 0.77). There was no significant difference in costs due to an interaction between the donor/control indicator and sex, age, or duration of the evaluation (Pint > 0.1 for all). There was a significant interaction with era: in 2011 to 2014 the incremental cost of donation was CAD $810 (CAD $44, CAD $1577) higher than before 2008.
In the subset of donors (n = 1099), healthcare costs were higher if the recipient started dialysis during the donor’s evaluation (CAD $886 [CAD $19, CAD $1,752) compared with preemptive transplants], but this did not affect healthcare costs during the perioperative (P = 0.82) or follow-up (P = 0.68) periods (Table 3). There was a nonsignificant trend in different costs across transplant centers for the evaluation and follow-up periods (P = 0.07 and P = 0.09, respectively) (Table 3). Costs were significantly different across transplant centers during the perioperative period, ranging from −CAD $1318 (−CAD $1971, −CAD $664) to CAD $599 (−CAD $502, CAD $1701) compared with 1 referent center (P < 0.0001).
Estimated Cost of the Evaluation Process for Candidates Who Did Not Donate
We estimated the cost of the evaluation assuming the donors were donor candidates who only completed a portion of their evaluation. The donor candidate evaluation cost was $1633 ($1452, $1813) if the candidates completed 50% of the evaluation and CAD $2699 (CAD $2463, CAD $2936) if they completed 90% of the entire evaluation (Table 4). There was a nearly linear relationship between the proportion of the evaluation completed and the cost of the donor evaluation (Figure 3).
When we restricted the analysis to donations after April 1, 2006, to accommodate the cost of capitation, the total cost of healthcare for living donors was very similar [mean, CAD $16 666 [95% CI, CAD $15 799-CAD $16 867]). The cost of the evaluation, perioperative, and follow-up periods were also similar: CAD $3565 (CAD $3319, CAD $3810), CAD $11 741 (CAD $11 409, CAD $12 073), and CAD $989 (CAD $747, CAD $1,232), respectively.
In sensitivity analyses which altered the regression model selected, models that fit the data as well or only slightly worse than the base case did not change the marginal effect estimates for perioperative and follow-up donor costs, but the evaluation costs ranged from a mean CAD $3414 (95% CI, CAD $3229-CAD $3599) with the square-root gamma model to CAD $4239 (95% CI, CAD $3852-CAD $4626) with the log-gamma model.
In this study from Ontario Canada, we found that the average cost to the healthcare system attributable to a living kidney donor was CAD $16 290. Most of these costs were incurred in the perioperative period (CAD $11 694), with costs also accrued during the evaluation period (CAD $3596). The cost of the evaluation for potential donors who completed 25% of their evaluation was CAD $865.
After adjusting for the donor/control indicator, higher healthcare consumption was observed during the evaluation period for women, older individuals, and those with a longer evaluation period. These observations were expected and consistent with the literature.26 We did not find any evidence that the cost of donation was different concerning these factors (nonsignificant interaction terms). However, we found that the evaluation costs were significantly higher for donors if their intended recipient started dialysis partway through their evaluation. This may be due to incidental donor candidate findings that require further work-up (eg, characterising ovarian or hepatic lesions identified by renal ultrasound). In turn, a prolonged evaluation caused by the recipient initiating dialysis may result in some tests being repeated, which is additional to background healthcare consumption. Although we did not observe any differences across transplant programs for the cost of the evaluation, there was significant variability in the cost of the perioperative period. Because this is the most costly period of donation, understanding the reasons for these differences and any effects on outcomes may identify opportunities for cost-savings.
To the best of our knowledge, only 1 study attempted to describe the costs of the evaluation, donation, and follow-up periods separately for a small sample of living donor kidney transplants (n = 130).8 The cost of donor candidates (those who did not donate) were included in 3 studies.8,10,27 However, the cost of a partial donor assessment may vary from center-to-center depending on their procedures for living donor work-up: transplant programs that perform multiple tests on the same day28 or those that evaluate multiple candidates simultaneously may incur higher evaluation costs since donor candidates who did not donate will have a greater number of tests performed. We reported the cost of partial evaluations, assuming nondonors would be scheduled to receive the same evaluation process as donors. Individual programs can interpret these costs in a manner that most closely fits their current or prospective operations.
Previous studies have estimated the total cost of living donation-related care as $23937 in Alberta Canada, CAD $15 462 in France, and CAD $15 850 in Spain (cf. CAD $16 290 in this study, all in 2017 $CAD).7,8,11 The donor evaluation period accounted for a significant proportion of the total cost of living donation: 11% in the Alberta study, 12% in the Spanish study, and 22% in this study (the periods differed in the French study). The estimated healthcare costs in the current study are lower than the Alberta study,8 particularly those related to the perioperative period ($11 644 vs. $18 482). Our study captured the healthcare utilization for donors more comprehensively, so it is unclear why the donation costs are much higher in Alberta. Although we did not include the cost of partial evaluations, the cost of the predonation phase was similar.
Having a nondonor control group and adjusting for covariates is a novel approach to estimate the incremental costs associated with living donation. This methodology (where all costs are included) guards against both underestimation (does not omit relevant costs since all costs are captured) and overestimation (does not include irrelevant costs since on average these are removed by the controls), and also provides a measure of precision. However, there are some limitations that should be acknowledged. First, the reimbursement costs for out-of-pocket expenses borne by the donors were not available.29 These costs are often remunerated by government-funded organizations and should be included if a governmental perspective is desired.30 Second, in this study we only considered persons who became donors. The healthcare utilization of donor candidates who did not donate should be described in future work. Several candidates may be evaluated to realize 1 kidney transplant, and some evaluations may not result in the identification of a suitable donor. Summating all these candidate evaluation costs may be important for some purposes. Third, the cost of running a living donor program was not measured. This includes the cost of personnel (eg, living donor nurse coordinator, social work support, administrative assistant), equipment, and overhead. Fourth, we only looked at 1-year follow-up costs. Some costs related to donation may take decades to manifest (ie, possible donation-related kidney disease). Fortunately, the 15-year increase in the absolute risk of kidney failure attributable to donation appears to be small.31 Finally, these estimates pertain to a universal healthcare system, where 78% of donors were of white race, and may not generalize well to other countries or healthcare systems.
We found that the cost of living kidney donation in Ontario, Canada is on average CAD $16 290 per donor. The perioperative period is the largest component of the costs (CAD $11 694 per donor) followed by the evaluation (CAD $3596 per donor) and follow-up periods (CAD $1011). Although substantial costs of living donor care are related to the nephrectomy procedure, comprehensive assessment of costs must include evaluation and follow-up care. These estimates are informative for planning future work to support and expand living donation and transplantation, and directing efforts to improve the cost efficiency of living donor care.
1. Wyld M, Morton RL, Hayen A, et al. A systematic review and meta-analysis of utility-based quality of life in chronic kidney disease treatments. PLoS Med
2. Schold JD, Buccini LD, Goldfarb DA, et al. Association between kidney transplant center performance and the survival benefit of transplantation versus dialysis. Clin J Am Soc Nephrol
3. Guimarães J, Araújo AM, Santos F, et al. Living-donor and deceased-donor renal transplantation: differences in early outcome—a single-center experience. Transplant Proc
4. Nemati E, Einollahi B, Lesan Pezeshki M, et al. Does kidney transplantation with deceased or living donor affect graft survival? Nephrourol Mon
5. Gozdowska J, Zatorski M, Torchalla P, et al. Living-donor versus deceased-donor kidney transplantation: comparison of psychosocial consequences for recipients. Transplant Proc
6. Silva SB, Caulliraux HM, Araújo CA, et al. Cost comparison of kidney transplant versus dialysis in Brazil. Cad Saude Publica
7. Sánchez-Escuredo A, Alsina A, Diekmann F, et al. Economic analysis of the treatment of end-stage renal disease treatment: living-donor kidney transplantation versus hemodialysis. Transplant Proc
8. Barnieh L, Manns BJ, Klarenbach S, et al. A description of the costs of living and standard criteria deceased donor kidney transplantation. Am J Transplant
9. Mullins CD, Thomas SK, Pradel FG, et al. The economic impact of laparoscopic living-donor nephrectomy on kidney transplantation. Transplantation
10. Bongiovanni I, Couillerot-Peyrondet A-L, Sambuc C, et al. Cost-effectiveness analysis of various strategies of end-stage renal disease patients' care in France. Nephrol Ther
11. Sainsaulieu Y, Sambuc C, Logerot H, et al. Cost of a renal transplant: medico-economic analysis of the amount reimbursed by the French national health program to finance renal transplantation. Nephrol Ther
12. Held PJ, McCormick F, Ojo A, et al. A cost-benefit analysis of government compensation of kidney donors. Am J Transplant
13. Barnieh L, Gill JS, Klarenbach S, et al. The cost-effectiveness of using payment to increase living donor kidneys for transplantation. Clin J Am Soc Nephrol
14. Shimizu U, Saito S, Lings Y, et al. Cost-effectiveness achieved through changing the composition of renal replacement therapy in Japan. J Med Econ
15. Haller M, Gutjahr G, Kramar R, et al. Cost-effectiveness analysis of renal replacement therapy in Austria. Nephrol Dial Transplant
16. Howard K, Salkeld G, White S, et al. The cost-effectiveness of increasing kidney transplantation and home-based dialysis. Nephrology (Carlton)
17. Matas AJ, Schnitzler M. Payment for living donor (vendor) kidneys: a cost-effectiveness analysis. Am J Transplant
18. Lam NN, Lentine KL, Levey AS. Long-term medical risks to the living kidney donor. Nat. Rev. Nephrol
19. Lentine KL, Kasiske BL, Levey AS, et al. KDIGO clinical practice guideline on the evaluation and care of living kidney donors. Transplantation
. 2017;101(8S Suppl 1):S7–S105.
20. Trillium Gift of Life Network. 2004–2014. Toronto, Ontario
21. Habbous S, McArthur E, Begen MA, et al. Initiating maintenance dialysis prior to living kidney donor transplantation when a donor candidate evaluation is well underway. Transplantation
22. Wodchis WP, Bushmeneva K, Nikinovic M, et al. Guidelines on Person-Level Costing Using Administrative Databases in Ontario. HSPRN
. Working Paper Series Vol 1. 2013.
23. Glick HA, Doshi JA, Sonnad SS, et al. Economic Evaluation in Clinical Trials
. Oxford: Oxford University Press; 2007.
24. Jones A, Rice N, D’Uva TB, et al. Applied Health Economics
. Second ed. London and New York: Routledge; 2013.
25. Habbous S, Arnold J, Begen MA, et al. Duration of living kidney transplant donor evaluations: findings from 2 multi-center cohort studies. [published online March 24, 2018]. Am J Kidney Dis
. doi: 10.1053/j.ajkd.2018.01.036.
26. Alemayehu B, Warner KE. The lifetime distribution of health care costs. Health Serv Res
27. Asgeirsdóttir TL, Asmundsdóttir G, Heimisdóttir M, et al. Cost-effectiveness analysis of treatment for end-stage renal disease. Laeknabladid
28. Graham JM, Courtney AE. The adoption of a one-day donor assessment model in a living kidney donor transplant program: a quality improvement project. Am J Kidney Dis
29. Klarenbach S, Gill JS, Knoll G, et al. Economic consequences incurred by living kidney donors: a Canadian multi-center prospective study. Am J Transplant
30. Sickand M, Cuerden MS, Klarenbach SW, et al. Reimbursing live organ donors for incurred non-medical expenses: a global perspective on policies and programs. Am J Transplant
31. Lam NN, Lentine KL, Garg AX. End-stage renal disease risk in live kidney donors: what have we learned from two recent studies? Curr Opin Nephrol Hypertens