SEARCH

SEARCH BY CITATION

Keywords:

  • Adherence;
  • compliance;
  • costs;
  • graft survival;
  • immunosuppressant adherence;
  • immunosuppression;
  • kidney;
  • patient survival

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

We describe factors associated with immunosuppression compliance after kidney transplantation and examine relationships between compliance with allograft outcomes and costs. Medicare claims for immunosuppression in 15 525 renal transplant recipients with at least 1 year of graft function were used to calculate compliance as medication possession ratio. Compliance was categorized by quartiles as poor, fair, good and excellent. We modeled adjusted associations of clinical factors with the likelihood of persistent compliance by multiple logistic regressions (aOR), and estimated associations of compliance with subsequent graft and patient survival with Cox proportional hazards (aHR). Adolescent recipients aged 19–24 years were more likely to be persistently noncompliant compared to patients aged 24–44 years (aOR 1.49 [1.06–2.10]). Poor (aHR 1.80 [1.52–2.13]) and fair (aHR 1.63[1.37–1.93]) compliant recipients were associated with increased risks of allograft loss compared to the excellent compliant recipients. Persistent low compliance was associated with a $12 840 increase in individual 3-year medical costs. Immunosuppression medication possession ratios indicative of less than the highest quartile of compliance predicted increased risk of graft loss and elevated costs. These findings suggest that interventions to improve medication compliance among kidney transplant recipients should emphasize the benefits of maximal compliance, rather than discourage low compliance.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Recent literature reviews combining the results from over 300 articles estimated the prevalence of poor immunosuppression compliance among kidney transplant recipients to be between 22% and 28% (1,2). Twenty percent of late acute rejections and 16% of graft losses were attributed in part to poor compliance (2). Other literature reviews have suggested that the impact of compliance on graft loss may be even greater in children (3–5).

Compliance has been measured in many ways. The measurement of immunosuppressant drug levels is performed routinely in kidney transplant recipients to assess drug exposure levels and can be used to directly measure drug compliance. The drawback is that these assays are dependent on the half-life of the metabolite and provide information only on days when the patient visits the clinic (6). Electronic monitoring has become the gold standard for measuring compliance in prospective studies and clinical trials, but monitoring the opening of pill dispensers is intrusive and not commonly done at most transplant centers (7–9). The majority of studies have used surveys to obtain a patient self-report of compliance (1,2,10). Unfortunately, this methodology may overestimate compliance rates because noncompliant patients may be hesitant to reveal their true pill-taking behavior (2,11).

Many previous studies on immunosuppression compliance have been limited to clinical trials or single-center studies of samples ranging approximately from 100 to 1500 subjects (2). A recent review outlined three approaches for examining compliance using insurance claims electronically submitted to obtain reimbursement for dispensed medications: fixed time point, gaps in prescription filling and medication possession ratio (MPR) (12). Using a fixed time point to measure compliance defines a patient as compliant if they have a medication fill with a prespecified time frame. For example, to measure compliance at a 1-year time point, a patient would be noncompliant if there was no prescription filled between days 335 and 395. Measuring compliance by looking for gaps in prescriptions defines a patient as noncompliant if they have a gap of greater than 30 days between run out of pills and the next prescription fill. MPR is defined as the number of days medication is supplied over a 1-year time interval (13). For example, if a patient receives a 30-day supply of immunosuppression, but consistently refills their medication after 35 days, they would have had medication for 320 days out of 360, MPR = 89% (320/360). For a study that has more than 1- year of follow-up, the MPR offers valuable strengths over fixed time point and gap methods. First, it allows for compliance to be measured as a continuous variable. Second, compliance can be calculated on a daily basis, which will allow patient compliance to be measured in a more fluid manner. We have recently applied this methodology in a study of MMF compliance and found that poor compliance with this drug was associated with 43–46% increased risk of graft failure (14).

Here we present an unobtrusive method for assessing compliance using Medicare pharmacy claims for patient immunosuppression benefits. We aimed to identify patient characteristics and other clinical factors associated with suboptimal compliance with calcineurin inhibitor and antimetabolite use among renal allograft recipients. We examined the relationship of less than optimal compliance patterns with 3-year graft and patient outcomes. Finally, we quantified medical costs associated with persistent noncompliance.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Study population

This was a retrospective cohort study of data from the United States Renal Data System (USRDS) registry (based on data available on August 10, 2008). We included first-time recipients of solitary kidney transplants from 1995 through 2001 with Medicare as the primary insurance carrier at time of transplant. We used Medicare claims to limit the study sample to patients with indicated use of mycophenolate mofetil, azathioprine, cyclosporine, or tacrolimus as maintenance immunosuppression in the first year after transplant. To limit inaccuracies in compliance measures resulting from early graft failure, we also required that participants survived with allograft function to the first posttransplant anniversary.

Defining compliance from claims

Compliance was measured by an MPR for mycophenolate mofetil, azathioprine, cyclosporine or tacrolimus. To create the immunosuppression MPR we calculated a medication possession timeline for each drug independently. We then looked at each day to see if the patient had any of the drugs on that day (Figure 1). As Medicare covers immunosuppression fills for a maximum 30-day supply and these medications are taken chronically, we assumed that each fill covered a 30-day medication requirement, and that patients were not ordered by their physician to cease all medication. The MPR was calculated as the number of days that the patient had pills over a 360-day period divided by 360 days. The MPR calculation was censored when a patient was without immunosuppression for more than 90 days.

image

Figure 1. Creating the Immunosuppression medication possession timeline. The medication possession timeline for each immunosuppression drug (MMF, AZA, TAC and CSA). The overall timeline is a composite of all four drug timelines. The gaps in the overall timeline indicate times when the patient was going without any immunosuppression.

Download figure to PowerPoint

MPR was categorized based on the 75th percentile, median and 25th percentile as indicative of poor, fair, good and excellent compliance, respectively. The percentile cutoffs were determined at the following number of days after transplant: 360 (first-year compliance), 735 (second-year compliance) and 1096 (3-year compliance). Table 1 shows the quartile cutpoints for categorizing compliance levels at 1 year, 2 years and 3 years, as well for the full duration of the study. For the purposes of creating the cutpoints, once a person was censored their last compliance value was assumed as representative of their compliance level for that year. For example, if a patient was censored between the first and second transplant anniversary, their MPR at time of censoring was used to quantify their compliance level in the second year. Notably, such a subject censored before the second anniversary was not included in computation of sample compliance rankings in the third year after transplant.

Table 1. Medication possession ratio quartile cutpoints for categorization of compliance over the study period.
 First yearSecond yearThird yearOverall
25%0.7310.8160.8270.811
50%0.8960.9510.9620.951
75%0.9640.9971.0000.998

We examined cumulative compliance by examining only patients who survived into their third-year posttransplant. ‘Overall high compliance’ was defined by excellent compliance ratings for the first, second and third posttransplant year. ‘Overall low compliance’ was defined by fair or poor compliance rankings at all three evaluation points. Patients who demonstrated neither consistently high nor low compliance were considered to have normal compliance.

Covariate definitions

Patient characteristics recorded in the registry at transplant included recipient age, gender, race, ethnicity, cause of end-stage renal failure, duration of dialysis prior to transplant, comorbid conditions (diabetes, hypertension, cardiovascular disease), peak panel reactive antibodies (PRA), cold ischemia time, degree of human leukocyte antigen (HLA) matching; Cytomegalovirus (CMV) seropairing, and delayed graft function. The donor characteristics included age, gender, race and cause of death. Patients’ immunosuppression regimen as defined by the OPTN discharge records, induction therapy and rejection in the first year were also included. We defined new-onset diabetes mellitus after transplant (250.xx, 357.2, 362.0, 366.41 and 641.0x), gastrointestinal complications (531.xx–536.xx, 540.x, 541, 556.x, 558.x, 562.xx, 564.xx, 569.xx, 578.x, 783.xx, 787.xx and 789.xx) and infections (001.xx–137.xx) by Medicare billing claims with corresponding International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) submitted from 1995 to 2001.

Computation of posttransplant costs

Costs were computed based on payments submitted on all Medicare Part-A and Part-B claims in 1995–2001. All costs were adjusted for inflation with the medical component of the consumer price index using the year 2000 as the base year. Average accumulated cost curves were created to visually examine the difference in costs between compliance levels. We calculated individual 3-year accumulated costs for every patient who was not censored before their third anniversary. We then subtracted the cost of immunosuppression from the patient's costs to compute the percentage of cost increase due to poor compliance.

Statistical analyses

We compared the distributions of compliance categories according to baseline clinical characteristics by chi squared tests. We modeled covariate-adjusted associations of baseline clinical factors with overall high and low compliance patterns (adjusted odds ratio, aOR) by separate multivariate logistic regressions.

We estimated the incidence of allograft failure and patient death after the first transplant anniversary according to 1-year compliance patterns by the Kaplan–Meier method. Multivariate Cox regression models were used to determine the associations of compliance category with graft and patient survival after the first transplant anniversary. Compliance was modeled as a time-varying covariate, such that each patient was categorized according to compliance quartile ranking over time. Compliance quartile cutoffs were defined as the mean of the daily quartile cutoffs over the follow-up period. Patient survival was censored at loss of Medicare coverage, 3 years posttransplant, or end of study (December 31, 2001). Loss of Medicare coverage was defined as the date of last medical claim, loss of eligibility as defined by the USRDS, or 30 consecutive days without any immunosuppression claims. Because the compliance measure required survival with graft function to the first transplant anniversary, the origin time for time-to-event analyses was 365 days after transplant. Models were adjusted for recipient, donor and transplant factors defined as covariates above.

Three-year medical costs were examined in patients who had a compliance measured at the end of the third year using multiple linear regression. Cost curves were generated to show patient medical cost differences between patients in the persistently high and persistently low compliance groups. This method did not adjust for other factors. Multiple linear regressions were also performed to adjust the cost contributions of compliance for other observed recipient, donor and transplant factors. All statistical analyses were calculated using SAS software for Windows, version 9.1 (SAS Institute, Inc., Cary, NC). Differences were considered statistically significant at a 2-sided p < 0.05. The study was approved by the Saint Louis University Institutional Review Board.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

There were 83 283 patients who received primary kidney transplants from 1995 through 2001, of which 15 525 recipients met selection serial criteria. Criteria for inclusion in the sample included Medicare coverage (46 593 patients excluded), use of Medicare immunosuppression benefits (14 346 patients excluded), survival with graft function and Medicare coverage to the first anniversary (6401 patients excluded), and evidence of immunosuppression in UNOS (422 patients excluded). The sample population included slightly more African-American recipients, as they were on dialysis longer and were less likely to receive a living donor transplant (results not shown). Table 2 describes the demographic characteristics of the full sample and distributions of first-year compliance patterns, based on quartiles, according to baseline traits.

Table 2.  Demographic traits of the study cohort, and distributions of compliance patterns according to baseline characteristics
 Overall compliance NPoor compliance N (%)Fair compliance N (%)Good compliance N (%)Excellent compliance N (%)p-Value1
  1. 1As measured by a chi-squared test to examine if there is a difference between groups.

Overall   15 5253848 (24.8%)3833 (24.7%)4118 (26.5%)3726 (24.0%) 
Recipient age
 0–18246 (1.6%)89 (36.2%)63 (25.6%)54 (22.0%)40 (16.3%)<0.0001
 19–24705 (4.5%)229 (32.5%)152 (21.6%)182 (25.8%)142 (20.1%) 
 24–446125 (39.5%)1506 (24.6%)1470 (24.0%)1658 (27.1%)1491 (24.3%) 
 45–605384 (34.7%)1300 (24.2%)1364 (25.3%)1409 (26.2%)1311 (24.4%) 
 Over 613065 (19.7%)724 (23.6%)784 (25.6%)815 (26.6%)742 (24.2%) 
Gender
 Female6257 (40.3%)1612 (25.8%)1490 (23.8%)1646 (26.3%)1509 (24.1%)0.0564
 Male9268 (59.7%)2236 (24.1%)2343 (25.3%)2472 (26.7%)2217 (23.9%) 
Physical limitations1162 (7.5%)336 (28.9%)288 (24.8%)292 (25.1%)246 (21.2%)0.0033
Recipient race
 Caucasian7894 (50.9%)1971 (25.0%)1925 (24.4%)2096 (26.6%)1902 (24.1%)0.0122
 African American4555 (29.3%)1185 (26.0%)1133 (24.9%)1200 (26.3%)1037 (22.8%) 
 Other3076 (19.8%)692 (22.5%)775 (25.2%)822 (26.7%)787 (25.6%) 
Donor age
 0–181977 (12.7%)473 (23.9%)486 (24.6%)529 (26.8%)489 (24.7%)0.0169
 19–5510 023 (64.6%)2492 (24.9%)2454 (24.5%)2624 (26.2%)2453 (24.5%) 
 Over 561808 (11.7%)423 (23.4%)489 (27.1%)512 (28.3%)384 (21.2%) 
 Missing1717 (11.1%)460 (26.8%)404 (23.5%)453 (26.4%)400 (23.3%) 
Dialysis duration
 0–24 months5185 (33.4%)1303 (25.1%)1219 (23.5%)1362 (26.3%)1301 (25.1%)0.0592
 25–60 months7305 (47.1%)1808 (24.8%)1816 (24.9%)1942 (26.6%)1739 (23.8%) 
 Over 60 months3035 (19.6%)737 (24.3%)798 (26.3%)814 (26.8%)686 (22.6%) 
Cold ischemia time
 Living donor3369 (21.7%)803 (23.8%)822 (24.4%)845 (25.1%)899 (26.7%)0.0058
 1–24 h10 421 (67.1%)2629 (25.2%)2555 (24.5%)2811 (27.0%)2426 (23.3%) 
 25–36 h1457 (9.4%)351 (24.1%)387 (26.6%)378 (25.9%)341 (23.4%) 
 Over 36 h278 (1.8%)65 (23.4%)69 (24.8%)84 (30.2%)60 (21.6%) 
Recipient comorbidities
 HTN12 347 (79.5%)3039 (24.6%)3043 (24.7%)3254 (26.4%)3011 (24.4%)0.1587
 CVD1407 (9.1%)361 (25.7%)333 (23.7%)362 (25.7%)351 (25.0%)0.5567
 Diabetes5073 (32.7%)1315 (25.9%)1300 (25.6%)1352 (26.7%)1106 (21.8%)<0.0001
PRA
 ≤ 80%14 769 (95.1%)3650 (24.7%)3640 (24.7%)3912 (26.5%)3567 (24.2%)0.2687
 >80756 (4.9%)198 (26.2%)193 (25.5%)206 (27.3%)159 (21.0%) 
IS regimen
 CSA and MMF6336 (40.8%)1383 (21.8%)1505 (23.8%)1765 (27.9%)1683 (26.6%)<0.0001
 TAC and MMF3617 (23.3%)863 (23.9%)882 (24.4%)946 (26.2%)926 (25.6%) 
 CNI and AZA2845 (18.3%)738 (25.9%)731 (25.7%)764 (26.9%)612 (21.5%) 
 CNI and Rapa717 (4.6%)276 (38.5%)195 (27.2%)158 (22.0%)88 (12.3%) 
 CNI only1470 (9.5%)425 (28.9%)397 (27.0%)337 (22.9%)311 (21.2%) 
 ADJ only540 (3.5%)163 (30.2%)123 (22.8%)148 (27.4%)106 (19.6%) 
 Induction6702 (43.2%)1737 (25.9%)1686 (25.2%)1813 (27.1%)1466 (21.9%)<0.0001
 Steroids15 052 (97.0%)3719 (24.7%)3719 (24.7%)4004 (26.6%)3610 (24.0%)0.4896
 DGF4227 (27.2%)1106 (26.2%)1103 (26.1%)1238 (29.3%)780 (18.5%)<0.0001
Adverse Outcome in the first year
 Rejection3586 (23.1%)931 (26.0%)941 (26.2%)979 (27.3%)735 (20.5%)<0.0001
 Diabetes6622 (42.7%)1691 (25.5%)1716 (25.9%)1748 (26.4%)1467 (22.2%)<0.0001
 GI8093 (52.1%)2112 (26.1%)2119 (26.2%)2155 (26.6%)1707 (21.1%)<0.0001
 Infection2981 (19.2%)855 (28.7%)787 (26.4%)765 (25.7%)574 (19.3%)<0.0001

There were 11 199 recipients who had a compliance measure into the third year of our study. Of these, 2589 (23.1%) were identified as having overall low compliance and 705 (6.3%) were identified has having overall high compliance. Table 3 shows the general demographic traits associated persistent low and high compliance.

Table 3.  Demographics for the study population that survived into year 3, and the percentage of low and high compliance stratified by various covariates
 Overall NLow compliance N (%)Normal compliance N (%)High compliance N (%)p-Value1
  1. 1As measured by a chi-squared test to examine if there is a difference between groups.

Overall11 1992589 (23.1%)7905 (70.6%)705 (6.3%) 
Recipient age
 0–18172 (1.5%)51 (29.7%)114 (66.3%)7 (4.1%)<0.0001
 19–24522 (4.7%)163 (31.2%)332 (63.6%)27 (5.2%) 
 25–444471 (39.9%)1067 (23.9%)3137 (70.2%)267 (6.0%) 
 45–603896 (34.8%)862 (22.1%)2786 (71.5%)248 (6.4%) 
 Over 612138 (19.1%)446 (20.9%)1536 (71.8%)156 (7.3%) 
Gender
 Female4514 (40.3%)1040 (23.0%)3177 (70.4%)297 (6.6%)0.5955
 Male6685 (59.7%)1549 (23.2%)4728 (70.7%)408 (6.1%) 
Physical limitations805 (7.2%)181 (22.5%)579 (71.9%)45 (5.6%)0.5898
Recipient race
 Caucasian5837 (52.1%)1237 (21.2%)4178 (71.6%)422 (7.2%)<0.0001
 African American3117 (27.8%)815 (26.2%)2140 (68.7%)162 (5.2%) 
 Other2245 (20.1%)537 (23.9%)1587 (70.7%)121 (5.4%) 
Donor age
 0–181537 (13.7%)358 (23.3%)1086 (70.7%)93 (6.1%)0.0448
 19–557431 (66.4%)1740 (23.4%)5236 (70.5%)455 (6.1%) 
 Over 561285 (11.5%)280 (21.8%)931 (72.5%)74 (5.8%) 
 Missing946 (8.5%)211 (22.3%)652 (68.9%)83 (8.8%) 
Dialysis duration
 0–24 months3878 (34.6%)863 (22.3%)2776 (71.6%)239 (6.2%)0.3228
 25–60 months5313 (47.4%)1232 (23.2%)3738 (70.4%)343 (6.5%) 
 Over 60 months2008 (17.9%)494 (24.6%)1391 (69.3%)123 (6.1%) 
Cold ischemia time
 Living donor2375 (21.2%)508 (21.4%)1707 (71.9%)160 (6.7%)0.33
 1–24 h7527 (67.2%)1787 (23.7%)5278 (70.1%)462 (6.1%) 
 25–36 h1084 (9.7%)245 (22.6%)772 (71.2%)67 (6.2%) 
 Over 36 h213 (1.9%)49 (23.0%)148 (69.5%)16 (7.5%) 
Recipient comorbidities
 HTN8697 (77.7%)1997 (23.0%)6124 (70.4%)576 (6.6%)0.0271
 CVD968 (8.6%)216 (22.3%)684 (70.7%)68 (7.0%)0.5507
 Diabetes3600 (32.2%)883 (24.5%)2522 (70.1%)195 (5.4%)0.0036
PRA
 ≤ 80%10 672 (95.3%)2449 (23.0%)7549 (70.7%)674 (6.3%)0.1561
 >80527 (4.7%)140 (26.6%)356 (67.6%)31 (5.9%) 
IS regimen
 CSA and MMF4792 (42.8%)988 (20.6%)3447 (71.9%)357 (7.5%)<0.0001
 TAC and MMF2190 (19.6%)524 (23.9%)1514 (69.1%)152 (6.9%) 
 CNI and AZA2378 (21.2%)537 (22.6%)1734 (72.9%)107 (4.5%) 
 CNI and Rapa362 (3.2%)136 (37.6%)222 (61.3%)4 (1.1%) 
 CNI only1111 (9.9%)317 (28.5%)729 (65.6%)65 (5.9%) 
 ADJ only366 (3.3%)87 (23.8%)259 (70.8%)20 (5.5%) 
 Induction4534 (40.5%)1124 (24.8%)3158 (69.7%)252 (5.6%)0.0002
 Steroids10 887 (97.2%)2497 (22.9%)7706 (70.8%)684 (6.3%)0.0199
 DGF2746 (24.5%)664 (24.2%)1934 (70.4%)148 (5.4%)0.0382
Adverse outcome in the first year
 Rejection2754 (24.6%)629 (22.8%)1983 (72.0%)142 (5.2%)0.0098
 Diabetes4678 (41.8%)1138 (24.3%)3277 (70.1%)263 (5.6%)0.0037
 GI5721 (51.1%)1425 (24.9%)3974 (69.5%)322 (5.6%)<0.0001
 Infection2138 (19.1%)563 (26.3%)1463 (68.4%)112 (5.2%)0.0001

The adjusted likelihood of persistent low compliance by multivariate logistic regression was higher in adolescent recipients aged 19–24 years compared to patients aged 24–44 years (aOR 1.56, p < 0.0001) and in patients who experienced a gastrointestinal complication in the first year after transplant (aOR 1.20, p = 0.0001) compared to those without gastrointestinal complications (Table 4). Low compliance was more likely, and conversely high compliance was less likely, among non-Caucasian patients compared to Caucasian patients. Rejection (aOR 0.81, p = 0.04) and gastrointestinal complications (aOR 0.70, p < 0.0001) in the first year after transplant were also associated with a decreased likelihood of persistent excellent compliance. Some immunosuppression regimens were also associated with compliance; tacrolimus with MMF (aOR 1.16, p = 0.02), CNI with azathioprine (aOR 1.14, p = 0.03), CNI with rapamycin (aOR 2.35, p < 0.0001) and CNI monotherapy (aOR 1.54, p < 0.0001) were associated with poor compliance when compared to patients taking cyclosporine with MMF.

Table 4.  Logistic regression results for persistent compliance over 3 years
 Persistent compliance
Low compliance OR (95% CI)High compliance OR (95% CI)
Recipient age
 0–181.49 (1.06–2.10)0.54 (0.24–1.21)
 19–241.56 (1.27–1.91)0.59 (0.38–0.92)
 24–44ReferenceReference
 45–600.91 (0.82–1.01)1.14 (0.93–1.40)
 Over 610.87 (0.77–0.99)1.27 (1.00–1.62)
Female0.96 (0.88–1.06)1.12 (0.94–1.34)
Recipient race
 CaucasianReferenceReference
 African
American1.27 (1.14–1.41)0.60 (0.48–0.74)
Other1.15 (1.02–1.30)0.66 (0.52–0.83)
IS regimen
 CSA and MMFReferenceReference
 TAC and MMF1.16 (1.02–1.31)0.84 (0.67–1.05)
 CNI and AZA1.14 (1.01–1.29)0.55 (0.43–0.71)
 CNI and Rapa2.35 (1.87–2.95)0.08 (0.03–0.21)
 CNI only1.54 (1.32–1.80)0.56 (0.41–0.76)
 Adj only1.19 (0.92–1.53)0.65 (0.39–1.09)
DGF1.01 (0.91–1.13)0.82 (0.66–1.02)
Adverse outcome
 Rejection0.99 (0.90–1.10)0.81 (0.66–0.99)
 Diabetes1.05 (0.93–1.19)0.88 (0.69–1.12)
 GI1.20 (1.09–1.31)0.70 (0.59–0.84)
 Infection1.21 (1.08–1.35)0.75 (0.59–0.94)

Figure 2 shows the Kaplan–Meier curves for both graft and patient survival for the 2 years after the first transplant anniversary. Among survivors to the first anniversary, the incidence of subsequent graft failure varied significantly by first-year compliance categorization, with failure over the next 2 years of 11.5% in the poor compliance group compared to 7.4% in the excellent compliance group (graft loss, p < 0.0001). The incidence of patient death varied significantly by first-year compliance categorization, with patient death over the next 2 years of 5.4% in the fair compliance group compared to 3.0% in the excellent compliance group (death, p < 0.0001).

image

Figure 2. Kaplan–Meier plot of (A) graft failure and (B) death by compliance quartile at the end of the first year.

Download figure to PowerPoint

Results from the time-varying Cox proportional hazards model are shown in Table 5. Compared to compliance in the upper quartile for the sample, when a recipient's compliance was fair, their risk of dying increased 50% (2nd quartile HR = 1.54). Risk of graft failure also increased when patients fell below the median (poor: HR = 1.80, fair: HR = 1.63). While the fair compliance group had a higher risk for death in terms of point estimate, there was a large amount of overlap between the group's 95% confidence intervals.

Table 5.  Cox proportional hazards of graft loss and patient death associated with level of compliance
ComplianceGraft failureDeath
HR (95% CI)p-ValueHR (95% CI)p-Value
ExcellentReference Reference 
Good1.12 (0.92–1.36)  0.25141.26 (0.96–1.65)0.0901
Fair1.63 (1.37–1.93)<0.00011.54 (1.19–2.00)0.0011
Poor1.80 (1.52–2.13)<0.00011.24 (0.95–1.64)0.1166

Figure 3 shows a curve for transplant costs centered on transplant date. The unadjusted difference in the average accumulated cost difference between the persistently high and the persistently low compliance groups was $12 840 per individual patient. In the multivariate regression, persistent noncompliance accounted for an estimated $7253 increase in adjusted medical costs over 3 years.

image

Figure 3. All medical costs over time by persistent high and low compliance.

Download figure to PowerPoint

Figure 4 shows the cost curve for medical costs, excluding the cost of immunosuppression. The unadjusted difference in the average accumulated non-immunosuppression medical care costs between the persistently high and the persistently low compliance groups was $33 346 per individual patient. In multivariate regression, persistent noncompliance accounted for a $21 600 increase in adjusted medical costs over 3 years.

image

Figure 4. Medical costs excluding cost of immunosuppression over time by persistent high and low compliance.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Measuring patient compliance using an MPR allowed us to examine compliance with more resolution than other methods that produce a binomial measure. The majority of previous studies treated compliance as a dichotomous variable. In this study, we modeled compliance in a time-varying manner with four compliance categories. Using this methodology, we recently found that poor compliance with MMF was associated with 43–46% increased risk of graft failure (14). In this study, we expanded this technique to examine associations of compliance patterns for calcineurin inhibitors and antimetabolites with graft outcomes, patient survival and costs, allowing us to investigate a potential gradient effect that the other studies could not assess.

A major finding of this study was detection of significant increases in the risk of graft failure in the lowest two quartiles of compliance. Kidney transplant recipients in the lowest two quartiles of compliance by MPR had more than a 60% increased risk of graft failure when compared to the highest compliance quartile. Risk of both graft failure and patient death were not significantly higher for patients who were in the 3rd compliance quartile compared to the 4th compliance quartile. Physicians treating patients whom they suspect to be noncompliant will now be able to show their patients estimated risks associated with skipping their medications. These results suggest that compliance should be stressed not only for patients thought to be at risk for poor compliance, but also for those with less than optimal compliance. Counseling on medication adherence should shift from warnings against the dangers of poor compliance to an emphasis on maximal compliance as a strategy for best outcomes. Physicians should also note that simply handing a patient educational material about the dangers of noncompliance has only limited effectiveness for improving compliance. The World Health Organization states that the most effective method for improving patient compliance is education combined with a patient-tailored intervention to promote medication-taking behavior (15). While the ultimate responsibility for medication taking lies with patients, transplant care providers can help support maximal compliance by disseminating this information during patient counseling.

This study also found that some immunosuppression regimens are associated with patient noncompliance. Patients who were on a CNI with either azathioprine or rapamycin were more likely to be noncompliant. The differences in compliance rates could be attributed to several different factors. First, patients taking azathioprine or rapamycin might be more likely to be prescribed an immunosuppression dose reduction between physician visits. Second, due to the low number of patients who are taking azathioprine or rapamycin in our study, it is possible that there are factors that are related to both patients who are on azathioprine or rapamycin and being noncompliant that we were not able to control for in the current model. Finally, it could be that treatment regimens for azathioprine or rapamycin are difficult for the patient to follow, either due to patient confusion or patient side effects.

Another major finding of this analysis relates to the economic implications of noncompliance for healthcare costs. Patients who were persistently noncompliant experienced approximately $33 000 higher medical costs after 3 years compared to patients with excellent compliance. Persistence of a behavior may be a barrier to change, and is important to note that approximately 25% of the patients were persistently noncompliant over the 3 years of study. Given the trends of noncompliance over time, it is likely that this difference in medical costs will continue to expand over time after transplant. These data provide theoretical support for programmatic investments in compliance encouragement and support as a means for cost savings in the long term after transplant.

Billing claims data have been shown to provide accurate measure of the use of calcineurin inhibitors and adjunctive agents. We previously demonstrated high concordance of Medicare claims for these agents with data reported to the OPTN as well as with electronic medical records of medication use (16,17). Administrative claims data are proving to be a useful supplement to survey data collected by the OPTN. Other data available from administrative claims include costs associated with medical care (18) and posttransplant complications not collected on the OPTN survey forms (19–22). In this study we expanded the applications of claims to demonstrate that compliance rates can be derived from these data and used to predict posttransplant outcomes and costs.

As the technical approach to calculating compliance, we used the methodology employed by Takemoto et al. to examine adherence to MMF (14). This method has several limitations. First, dose reductions made by a physician between fills may be misclassified. This reduction typically involves physicians directing their patients to take fewer pills per day, and dose reductions could be associated with poorer graft survival. The MPR cannot distinguish between noncompliance and this type of dose reduction. To reduce the possible effect of a physician-prescribed dose reduction, we included all immunosuppressants in our compliance measure. The physician would be required to reduce all of the patient's prescribed medications to have the dose reduction be considered noncompliance. We also attempted to improve upon the Takemoto et al. method by examining all immunosuppression simultaneously. This allowed us to examine a greater proportion of the target population, as well as reduce the risk that we were measuring physician-directed dose reduction as noncompliance. Unfortunately, patients who were being treated by physician/centers that utilize programs such as a tacrolimus monotherapy regimen with spaced weaning or a low dose immunosuppression regimen could be at an increased risk for graft failure and be more susceptible to being misclassified as noncompliant.

Second, taking the last known value for compliance until the end of the year for MPR computation introduces potentially skewed cut points for the quartiles. The decision to keep the last known value was an effort to increase the number of recipients who would be eligible for the cost analysis. If we were to limit the population to only those people who survived to 3 years, we would have lost 38% of our population. Third, we could not distinguish the difference between a person who was noncompliant and a person who had been ordered to stop taking immunosuppression. Finally, the MPR method treated all missed medications equally. A patient who missed a single dose 20 different times in the year would be classified the same as a patient who missed 20 doses in a row. We were unable to see if there was a clinical difference between missing doses over time or missing all doses at one time.

Our study was limited to patients with Medicare as primary insurer and our findings may not generalize to beneficiaries of private payer systems. Although our methods were designed to select patients with primary Medicare benefits, it is possible that administrative errors lead to inclusion of privately insured or patients using Medicare as a secondary payer in the study sample. If this is the case, we would have underestimated their MPR and their costs. To control for Medicare as a secondary payer, we censored patients when more than 90 days elapsed between fills. By censoring at 90 days, we allowed for enough time in between fills to allow for a patient to be reasonably noncompliant or catch a physician-related dose reduction. However, it is unlikely that a patient would restart using Medicare as their primary coverage after only 90 days. Because Medicare coverage expires at 3 years after transplant in the absence of disability or advanced age, we were restricted to only 3 years of follow-up in Medicare data. The effect of poor compliance on graft outcomes may increase in the later years (23), but later effects could not be measured with this design. Finally, based on the retrospective nature of our data we can describe associations but not prove a direct relationship. Similarly, we can illustrate the association between medical costs and compliance, but we cannot prove that poor compliance causes an increase in cost. Events that result from noncompliance as outcome mediators, such as acute rejection, may be true cost drivers but these pathways would not invalidate poor compliance as a maker of high posttransplant costs. However, we also cannot exclude that other unmeasured patient characteristics or behaviors associated with poor compliance are the true determinants of costs.

In summary, we applied the MPR as a method for assessing immunosuppression compliance after kidney transplant using Medicare pharmacy claims that quantifies compliance over time. Use of pharmacy claims is a non-intrusive approach suitable for the study of large samples. We observed significant increases in the risk of graft failure with all compliance levels less than excellent, not only with poor compliance. We also found that persistent noncompliance is associated with higher posttransplant costs. Prospective studies of the clinical and cost-effectiveness of multidisciplinary approaches for enhancing compliance over time are warranted.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Funding Sources: Portions of this work were supported by a grant from Astellas Pharma Inc., Tokyo, Japan. Dr. Lentine received support from a grant from the NIDDK, K08-0730306. Dr. Salvalaggio received support from a grant from the American Society of Transplantation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References