The Last Year Before Graft Failure Negatively Impacts Economic Outcomes and is Associated With Greater Healthcare Resource Utilization Compared With Previous Years in the United Kingdom: Results of a Retrospective Observational Study.

BACKGROUND: Kidney and liver transplantation is the standard of care for end-stage renal or liver disease. However, long-term survival of kidney and liver grafts remain suboptimal. Our study aimed to understand the healthcare resources utilized and their associated costs in the years before graft failure.
METHODS: Two noninterventional, retrospective, observational studies were conducted in cohorts of kidney or liver transplant patients. Once identified, patients were followed using the UK Clinical Practice Research Datalink linked to the Hospital Episode Statistics databases from the date of transplantation to the date of the first graft failure. Total healthcare costs in the year before graft failure (primary endpoint) and during years 2-5 before graft failure (secondary endpoint) were collected.
RESULTS: A total of 269 kidney and 81 liver transplant patients were analyzed. The mean total costs were highest for all resource components in the last year before graft failure, except for mean costs of immunosuppressive therapy per patient, which decreased slightly by index date (ie, graft failure). The mean total healthcare costs in the last year before graft failure were £8115 for kidney and £9988 for liver transplant patients and were significantly (P < 0.05) higher than years 2-5 before graft failure. Mean healthcare costs for years 2, 3, 4, and 5 before graft failure were £5925, £5575, £5469, and £5468, respectively, for kidney, and £6763, £7042, £6020, and £5651, respectively, for liver transplant patients.
CONCLUSIONS: Total healthcare costs in the last year before graft failure are substantial and statistically significantly higher than years 2-5 before graft failure, in both kidney and liver transplant patients. Our findings show the economic burden placed on healthcare services in the years before graft failure.

INTRODUCTION

Renal transplantation is considered the gold standard treatment choice among the available renal replacement therapy options for patients with end-stage renal disease.[1] Compared with dialysis, renal transplantation is associated with reductions in cardiac events, improved quality of life, and increased survival.[2] For patients with end-stage liver disease, liver transplantation provides the only effective treatment option.[3,4]

Both kidney and liver transplantation are cost-effective interventions compared with dialysis and other treatments.[5-7] However, the number of kidney transplant patients with a failed allograft from 1988 to 2010 in the United States has substantially increased[8] and complications following liver transplantation are common, with approximately 22% of patients in the United States developing biliary complications post-liver transplantation.[9] In addition, 20-year graft survival rates for both kidney and liver transplant patients in Europe are poor (<22%).[10,11] Due to the increasing size of transplant waiting lists (United Kingdom patients awaiting a liver transplant have tripled from 1999 to 2009) and shortage of grafts available,[12,13] maximizing the long-term survival and function of each transplant by minimizing or controlling the foremost causes of graft loss is a clinical priority.

In many areas of medicine, costs rise substantially in the year before significant clinical events,[14-17] indicating the need for specific strategies to be determined before these events occur. Although the average costs for transplantation, specifically the surgical procedure and preoperative care, are well established in the United Kingdom,[18] knowledge of healthcare resource utilization and the associated costs during the last year before kidney and liver graft failure is limited.

In this analysis, we present the results of two similarly designed studies. The primary aim was to describe the distribution of healthcare costs during the last year before kidney or liver graft failure in the United Kingdom. Secondary aims were to describe the distribution of these healthcare costs between years 2 and 5 before graft failure and compare these costs with the last year before graft failure. We hypothesized that the total healthcare costs in the last year before graft failure would be significantly higher than the costs during years 2–5.

MATERIALS AND METHODS

Study Overview/Design

Two noninterventional, retrospective, observational studies examined cohorts of either kidney or liver transplant patients in the United Kingdom. Data were collected from the National Health Service (NHS), Clinical Practice Research Datalink (CPRD), and Hospital Episode Statistics (HES) databases, allowing for a longitudinal analysis of resource and drug utilization. For both studies, the index date was defined as the date of the first graft failure. Details of graft failure and how it was identified are described further in the methods below.

Data Sources

Patients were selected from the CPRD and cross-referenced with HES databases[19,20] based on evidence (medcodes) for kidney or liver transplantation between 2004 and 2013 and subsequent graft failure between 2005 and 2014. Patient data were collected on demographic information, prescription details, clinical events (symptoms, diagnoses), preventive care provided, diagnostic/pathological tests, immunizations, specialist referrals, hospital admissions and their major outcomes, and details relating to death.

Patient Populations

Kidney and liver transplant patients were identified using transplant medcodes from CPRD and ICD-10 code Z94.4 from HES (Supplementary Tables 1 and 2, SDC, http://links.lww.com/TXD/A203). Kidney transplant patients receiving dialysis treatment were identified using dialysis medcodes from CPRD and ICD-10 codes T86.1, Z49.1, and Z99.2 from HES. Kidney transplant patient deaths were identified from the CPRD and HES and combined with the dialysis dataset to determine the index date (Figure 1A). Patients who received additional liver transplants were identified using the same codes from CPRD and HES as before. Liver transplant patient deaths were identified, and these data were combined with datasets for additional liver transplants to determine the index date (Figure 1B). Duplicate patients receiving kidney or liver transplants and those receiving additional liver transplants or dialysis were removed from both cohorts. Patients were matched to kidney and liver transplant patients who were receiving dialysis, a secondary liver transplant, or who had died using the common patient identifier. The time differences between liver transplant and a second transplant or death, and kidney transplant and dialysis or death, were calculated for each patient (Figure 1).

FIGURE 1.

Summary of search strategy and number of patients included. A, Kidney transplant cohort. B, Liver transplant cohort. CPRD, Clinical Practice Research Datalink; HES, Hospital Episode Statistics.

The studies included all kidney or liver transplant patients identified between 2004 and 2013 and those with graft failure between 2005 and 2014, for whom the time between receiving a transplant and subsequent graft failure was >365 days. To prevent the costs of initial posttransplantation resource use (not related to graft failure) affecting total healthcare costs, the studies excluded patients with <1 year between kidney or liver transplant and graft failure, and those patients who died within <1 year of receiving a transplant.

Outcomes and Endpoints

The primary endpoint for both studies was the total healthcare costs collected in the year before graft failure. A secondary endpoint was the total healthcare costs in years 2–5 before graft failure. For the primary objective, there was no comparison group or control. For the secondary objective, a repeated measures analysis was used.

Resource Utilization

Healthcare resource utilization was calculated for the following elements (each costed by their appropriate, respective 2015/2016 tariff): healthcare visits, hospital length of stay, procedures and operations performed, transplant-related resources, and drugs dispensed. The covariates examined included sex, age at kidney or liver failure (children <18 y vs adults), and type of donor (deceased vs live).

Costs

Cost data for resource components were obtained from the same sources for both kidney and liver transplant patients. The costs for services provided in the community were obtained from the Unit Costs of Health and Social Care[21] and the cost of immunosuppressive agents from the British National Formulary.[22] Diagnostic and pathology test costs were obtained from the NHS Reference Cost Schedule 2014–2015,[23] and the cost of services provided in secondary care were obtained from the NHS Reference Schedule 2015–2016.[18] A summary of all inpatient events captured by the Healthcare Resource Group in the last year before graft failure and from years 2–5 before graft failure are shown in Supplementary Tables 3, 4, 5, and 6, SDC, http://links.lww.com/TXD/A203.

The unit costs for the resources utilized leading up to graft failure are summarized in Table 1. Total healthcare costs were calculated by combining the following individual resource components: general practitioner (GP) consultations, diagnostic tests, immunosuppressive drugs, outpatient visits, day cases, and inpatient stays with associated interventions.

TABLE 1

Summary of unit costs for resources utilized by kidney and liver transplant patients

Statistical Analyses

Continuous and nominal variables were described using standard statistical measures. Repeated measures mixed model analysis (Stata 12.1)[24] was conducted to estimate the impact of time to graft failure on resource utilization while adjusting for covariates (sex, age, and type of donor). Covariates were added to model the outcome, starting with a random intercept model. Generalized linear models (log-link with a gamma error) were used to model healthcare costs in the last year before graft failure. Specifically, total healthcare costs for the period 5 years to >1 year before graft failure were compared with total healthcare costs in the last year before graft failure. Because they avoid the restrictions of repeated-measures analysis of variance, linear mixed models (containing both fixed and random effects) were used to ensure that total costs in the last year before graft failure were different to the total costs in years 2–5 before graft failure.[25]

RESULTS

Patients

The final number of patients included in the analysis was 269 kidney and 81 liver transplant patients. The number of patients available for analysis varied according to the time lapse between transplant and graft failure or death (Table 2). In the kidney cohort, the mean (standard deviation [SD]) age of patients was 57.4 (15.4) years (males: 56.8 [15.4] y; females: 58.4 [15.5] y). In the liver cohort, the mean (SD) age of patients was 57.8 (16.4) years (males: 60.0 [13.6] y; females: 54.9 [19.3] y).

TABLE 2

Number of patients available for analysis based on the time before graft failure

In the kidney transplant cohort, the graft failure categories at the index date were as follows: death (41%; n = 111), dependence on renal dialysis (23%; n = 61), complications of kidney transplant (18%; n = 49), extracorporeal dialysis (7%; n = 18), and other categories (11%; n = 30). In the liver transplant cohort, graft failure categories at the index date were as follows: death (89%; n = 72) and additional liver transplant (11%; n = 9).

GP Consultations

The mean number of GP consultations in the year before graft failure were 30.0 in the kidney transplant cohort and 34.1 in the liver transplant cohort with mean total durations of 7.3 and 9.3 minutes, respectively (Table 3). The mean duration and number of GP consultations for both studies, in years 2–5 before graft failure increased in the time periods closest to graft failure. Compared with years 2–5, the last year before graft failure demonstrated the highest mean duration and number of GP consultations for both studies (Table 3).

TABLE 3.

Mean healthcare resource use for kidney and liver transplants in the year before and years 2–5 before graft failure

Drug Utilization and Costs

The mean number of prescriptions per patient was 110.9 and 97.5 in the kidney and liver transplant cohorts, respectively (Table 3). The total cost of immunosuppressive therapies in the year before graft failure were £673 709 and £179 618 in the kidney and liver transplant cohorts, respectively (Table 4). Mean (SD) per patient costs were £2504 (£2298) and £2218 (£2849), respectively (Table 4). In the 5 years before graft failure, mean immunosuppressant costs per patient were lowest between years 1 and 2 for the kidney cohort and in the year before graft failure for the liver cohort. In the kidney transplant cohort, the other most frequently used drug therapies in the last year before graft failure were as follows: statins (59.9%; n = 161); calcium-channel blockers (55.4%; n = 149); proton pump inhibitors (54.3%; n = 146); antiplatelet drugs (52.4%; n = 141); and loop diuretics (52.0%; n = 140; Table 5). In the liver transplant cohort, these included proton pump inhibitors (63.0%; n = 51); broad-spectrum penicillin antibiotics (46.9%; n = 38); opioid analgesics (43.2%; n = 35); and antiplatelet drugs (40.7%; n = 33; Table 5). In years 2–5 before graft failure, other frequently used drug therapies for the kidney and liver cohorts included statins and proton pump inhibitors and broad-spectrum penicillin antibiotics, calcium-channel blockers, and antiplatelet drugs, respectively (Supplementary Tables 7 and 8, SDC, http://links.lww.com/TXD/A203).

TABLE 4.

Mean annual costs of immunosuppressive therapies

TABLE 5.

Use of other drug therapies for kidney and liver transplants in the last year before graft failure

Diagnostic Tests

In both cohorts, the mean number of diagnostic tests in the year before graft failure (61.7 and 65.4 per patient in the kidney and liver transplant cohorts, respectively) was higher compared with years 2–5 (Table 3). The most frequently used tests in the year before graft failure included serum creatinine (kidney: 5.1%; n = 839; liver: 3.9%; n = 208), potassium (kidney: 3.8%; n = 636; liver: 3.6%; n = 189), and sodium (kidney: 3.4%; n = 563; liver: 3.5%; n = 185; Table 6).

TABLE 6.

Diagnostic tests used for kidney and liver transplants in the last year before graft failure

Referrals from GPs

Compared with years 2–5, the mean number of referrals from GPs to other providers differed were higher in the year before graft failure for both cohorts (0.9 and 1.1 per patient for kidney and liver transplant cohorts, respectively; Table 3).

Inpatient Stays

The mean number of inpatient procedures in both cohorts was higher in the last year before graft failure (1.4 and 2.2 per patient for kidney and liver transplant cohorts, respectively) compared with years 2–5 (Table 3).

Day Cases

The mean number of day cases in the last year before graft failure was 1.1 and 4.4 per patient in the kidney and liver cohorts, respectively (Table 3). Day care procedures were highest in the last year before graft failure for the kidney transplant cohort and from years 3 to 4 (4.9 per patient) for the liver transplant cohort.

Outpatient Visits

In both cohorts, the mean number of outpatient visits in the last year before graft failure (7.5 and 2.9 per patient in kidney and liver transplant cohorts, respectively) was higher compared with years 2–5 (Table 3).

Total Costs by Year

The mean totals per patient healthcare costs in the last year before graft failure were £8115 (SD: £4539; 95% confidence interval [CI], £7570–£8659) for the kidney transplant cohort and £9988 (SD: £6703; 95% CI, £8506–£11 470) for the liver transplant cohort (Figure 2). For years 2, 3, 4, and 5, mean (SD) total healthcare costs for kidney transplant patients were £5925 (£3155), £5575 (£3253), £5469 (£2976), and £5468 (£3242) and for liver transplant patients were £6763 (£4940), £7042 (£5812), £6020 (£5518), and £5651 (£3074), respectively (Figure 2). Results of mixed-level modeling demonstrated total healthcare costs as a function of time to graft failure (last year compared with years 2–5) were statistically significant (P < 0.05; Figure 3).

FIGURE 2.

Mean total healthcare costs by year and component. A, Kidney. B, Liver. GP, general practitioner; IP, inpatient visits; OP, outpatient visits.

FIGURE 3.

Total costs as a function of time to graft failure and sex. A, Kidney transplant cohort. B, Liver transplant cohort.

For the kidney transplant cohort, the main cost components were immunosuppressive drugs, inpatient stays (both displayed similar costs), and outpatient visits. For the liver transplant cohort, the main cost components were inpatient stays (approximately twice the cost of immunosuppressive drugs), immunosuppressive drugs, and day cases. Total healthcare costs were higher in the last year before graft failure in all components compared with previous years, apart from immunosuppression (Figure 2). The mean (SD) inpatient cost per patient in the year before graft failure was £2521 (£3001) for kidney and £4494 (£4761) for liver transplant cohorts and was higher in the last year before graft failure compared with previous years (Figure 4). Median (interquartile range) inpatient costs per patient in the year before graft failure were £1510 (£3674) and £3221 (£5573) for kidney and liver transplant cohorts, respectively.

FIGURE 4.

Mean inpatient costs per patient in the year before graft failure and years 2–5 before graft failure. A, Kidney transplant cohort. B, Liver transplant cohort.

Total Costs by Sex, Age, and Type of Donor

For patients receiving a kidney transplant, the mean total costs during the last year before graft failure for males and females were £8413 (SD: £4726; 95% CI, £7685–£9142) per patient and £7648 (SD: £4210; 95% CI, £6833–£8462) per patient, respectively (Table 7). For patients receiving liver transplant, mean annual costs for males and females during the last year were £8421 (SD: £5651; 95% CI, £6743–£10 099) and £12 048 (SD: £7468; 95% CI, £9483–£14 613), respectively. Total costs by sex were statistically significantly different (P < 0.05) for liver transplant patients only, with higher total costs for females compared with males (Figure 5).

TABLE 7.

Mean total healthcare costs by year

FIGURE 5.

Mean total healthcare costs by age and sex. A, Kidney transplant cohort. B, Liver transplant cohort.

Kidney grafts from live donors were less expensive than those from deceased donors (mean costs: £5511 vs £9054), although not reaching statistical significance (due to small numbers of clearly identifiable grafts from live or deceased donors). For liver transplant patients, there were no sufficient data to analyze costs by donor type.

DISCUSSION

To our knowledge, this is the first study to accurately describe the distribution of healthcare costs and resource utilization in the years leading up to kidney and liver graft failure. The results of this study, based on real-world data, confirm the underlying hypothesis of the study that total healthcare costs in the last year before graft failure are significantly higher (P < 0.05) than years 2–5. Therefore, these studies show the later stages of a graft’s lifetime, specifically the last year before graft failure, to be associated with greater consumption of healthcare resources, with inpatient stays being the main cost driver.

Traditionally, economic evaluations in the field of transplantation (eg, a German study by Jüurgensen et al[26]) have taken into consideration potential changes in posttransplantation costs. However, few studies adequately reflect the actual patient pathway of graft failure. Kidney transplant patients typically present with slow functional decreases over time, eventually culminating in graft failure.[27] Comparably, liver transplant patients also follow a similar course with 10-year graft failure estimated at approximately 35%.[28] Moreover, studies have generally not accounted for cost variations in transplant patients using graft failure as the index date. Therefore, our analysis provides a novel insight into the additional costs incurred in the later years of a kidney and liver transplant patient’s clinical course.

In the present study, the mean total costs (median; interquartile range) during the last year before graft failure were £8115 (£7450.26; £5840.11) and £9988 (£8237.24; £8187.15) for kidney and liver transplants, respectively. Total costs substantially increased during the last year before graft failure, compared with the relatively stable costs reported during years 2–5. For liver transplant patients, costs were statistically significantly (P < 0.05) higher for female patients compared with male patients. Although the exact explanation for the higher reported cost in female liver transplant patients is unclear, it is possible that this may reflect, in part, the disparities in posttransplantation outcomes between male and female patients. For example, compared with male transplant patients, females have a slightly greater incidence of retransplantation.[29] This is possibly related to the recurrence of diseases, such as primary biliary cirrhosis and autoimmune hepatitis, which were responsible for the primary liver transplant.[29] Likewise, the recurrence of hepatitis C is also possibly related.[30] However, the development of effective antiviral treatments[31] questions whether sex differences will exist in future hepatitis C populations.

Inpatient visits were the biggest cost driver in the year before graft failure, confirming findings from other studies which analyzed the costs associated with kidney and liver transplant patients.[32,33] In the last year before graft failure, the reasons for hospitalization in our study were widespread and either related or unrelated to transplantation. Therefore, if strategies to avoid or defer inpatient interventions, such as those recommended by the Consensus on Managing Modifiable Risk in Transplantation group,[34] can be implemented, this will most likely impact the total healthcare costs in the last year before graft failure. In addition, the assessment of hospitalizations that could potentially be managed in a less costly outpatient setting may also help to reduce overall costs.

The Consensus on Managing Modifiable Risk in Transplantation group reports the importance of using comprehensive methods to identify and manage potentially reversible risk factors for graft failure in kidney and liver transplant recipients.[34] These modifiable risk factors over the longer term include issues related to immunosuppression, such as nonadherence and side effects. It is possible that closer clinical management of these risk factors in transplanted patients could plateau the costs in all the years before failure rather than significantly increasing costs in the last year.

Strategies that might accurately diagnose early graft failure would be beneficial. A study being conducted by Dorling et al[35] is evaluating the use of a combined antibody/treatment program in patients receiving a kidney transplant. The study aims to enhance graft function and delay graft failure through screening patients for antibodies against human leukocyte antigens to ensure that these patients, who are at a high risk of premature graft failure, are identified and treated accordingly. If a biomarker-led care regimen proved clinically beneficial and delayed the onset of graft failure, this may also reduce treatment costs per year, as graft failure has been shown to be a cost-driving event.[36] Measurement of serum creatinine is a common approach for the assessment of graft function or risk for graft loss, and ≥3.0 mg/dL appears to be associated with the lowest projected kidney graft half-life.[37] In our study, 839 creatinine tests were performed in 269 kidney patients in the year before graft failure.

Compared with randomized controlled trials, this study was able to measure healthcare resource utilization and costs over a longer time period through the use of real-world data.[38-40] As such, our results were based on data taken from a representative transplant patient population in the CPRD and associated HES databases in the United Kingdom. While unit costs would vary, it is likely that our key findings (significantly higher costs in the last year before graft failure, compared with years 2–5) would also pertain to other countries.

Our analysis also provides an accurate source of data to estimate healthcare resource use and costs associated with graft failure in the time leading up to the event. Higher resource use and costs in the last year before graft failure compared with years 2–5 is probably not unexpected. However, this study represents the first time that resources and costs have been estimated in the year before graft failure in renal and liver transplantation. These data could be used as inputs in future health economic assessments (eg, cost-effectiveness analyses of immunosuppressant therapies) and support payers in their decision-making. In addition, there may be justification to update cost-effectiveness analyses to account for intermediate states of disease progression, where the costs substantially increase before failure. Implications from our study may suggest that the cost of treating transplanted patients with immunosuppressants has been previously underestimated.

Important limitations of our studies should be noted. For instance, these studies have the established limitations of any retrospective analysis. However, given the lengthy period of this analysis, such studies would be difficult to undertake prospectively. The possibility of misclassification bias, and the fact that temporal relationships are often difficult to assess, are also limitations. Critically, this study is a conservative analysis and potentially underestimates the actual costs incurred leading up to kidney or liver graft failure. For example, as kidney graft failure is not an acute event and happens over time, dialysis is often implemented as a part of a care package before allograft failure; however, our analysis does not account for these costs. Additional renal costs, such as management of episodes of antibody-mediated rejection and re-establishing vascular access for patients whose grafts fail and need to be returned to dialysis, have not been fully captured in the total healthcare costs. Likewise, other secondary care costs, such as those associated with radiology (inpatient/outpatient) and bed stays (by type of bed: general ward, high dependency unit, intensive care unit) are not included in our analysis.

Nevertheless, a longitudinal analysis has been possible due to the size of the original database. To this end, our novel findings highlight the substantial burden placed on healthcare services in the years leading up to graft failure. On the basis of our results, future studies are recommended to compare healthcare resource utilization and costs in patients with and without graft failure. For example, it may be of value to evaluate resource use in patients whose creatinine rises above a certain threshold (eg, 3.0 mg/dL).

CONCLUSION

In conclusion, total healthcare costs in the year before graft failure in both kidney and liver transplant patients are substantial and significantly greater (P < 0.05) than the earlier years posttransplantation.

ACKNOWLEDGMENTS

The research presented in this manuscript was funded by Astellas Pharma Inc. Medical writing support was provided by Joseph Norvill of Bioscript Medical and funded by Astellas Pharma Inc.