• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of jnephrolLink to Publisher's site
J Am Soc Nephrol. Dec 2011; 22(12): 2296–2302.
PMCID: PMC3250211

Both Patient and Facility Contribute to Achieving the Centers for Medicare and Medicaid Services' Pay-for-Performance Target for Dialysis Adequacy


The Centers for Medicare and Medicaid Services (CMS) designated the achieved urea reduction ratio (URR) as a pay-for-performance measure, but to what extent this measure reflects patient characteristics and adherence instead of its intent to reflect facility performance is unknown. Here, we quantified the contributions of patient case-mix and adherence to the variability in achieving URR targets across dialysis facilities. We found that 92% of 10,069 hemodialysis patients treated at 173 facilities during the last quarter of 2004 achieved the target URR ≥65%. Mixed-effect models with random intercept for dialysis facility revealed a significant facility effect: 11.5% of the variation in achievement of target URR was attributable to the facility level. Adjusting for patient case-mix reduced the proportion of variation attributable to the facility level to 6.7%. Patient gender, body surface area, dialysis access, and adherence with treatment strongly associated with achievement of the URR target. We could not identify specific facility characteristics that explained the remaining variation between facilities. These data suggest that if adherence is not a modifiable patient characteristic, providers could be unfairly penalized for caring for these patients under current CMS policy. These penalties may have unintended consequences.

Governments and health care systems increasingly employ public reporting and financial incentives to improve health care delivery.1,2 Financial incentives are often described as representing “pay for performance.” The ideal pay for performance measure, on a facility level, should tackle a disease with significant public health burden, and performance should be modifiable by implementation of good facility practices. Finally, little variation in results should be explained by patient case mix.3 Even with these conditions in place, pay-for-performance programs can reach plateaus in improvement of care, and can have unintended consequences.46

Under the Medicare Improvements for Patients and Providers Act of 2008 (MIPPA), the Centers for Medicare & Medicaid Services (CMS) implemented a case-mix adjusted bundled prospective payment system (PPS) for Medicare ESRD dialysis facilities beginning January 1, 2011.7,8 As part of the PPS, the quality incentive program (QIP) will reduce the bundled payment to a dialysis facility for failure to meet a dialysis dose performance target. The performance standard is that, at each facility, the urea reduction ratio (URR) on monthly blood testing should exceed 65% for the lesser of 96% of patients or the percentage of patients at that facility whose average URR exceeded 65% in 2007.8

Hemodialysis dose, as measured by URR, is an interesting pay-for-performance target. In theory, the facility more easily controls the URR, a function of blood and dialysate flow, membrane size and permeability, and treatment time, than it does other process indicators. In practice, patients influence the delivery of their dialysis prescription, particularly treatment duration. Facilities in which large numbers of patients routinely shorten treatment time will be penalized under the QIP, although these facilities may be providing high-quality care. The objective of this study is to examine the association between patient and facility factors and the achievement of the URR target and the extent to which patient factors explain the variation of this target.


Study Population

The median URR across the cohort was 73%, with interquartile range 70% to 77%. Ninety-two percent (9232) of patients had a median achieved URR ≥65% over the quarter. The median proportion of patients meeting the URR target ranged from 83% (80,87) in the lowest facility quartile to 97% (96,100) in the highest facility quartile.

Patient and Facility Characteristics across Quartiles of URR Target Achievement

As shown in Table 1, patients in the lowest quartile of median URR were more likely to be younger, African American, and have a higher body surface area (BSA). In addition, patients in the lowest quartile of median URR were more likely to have diabetes as a cause for kidney failure, had lower rates of cardiovascular, cerebrovascular, and peripheral vascular disease. Lower median household income and an urban location were associated with lower facility quartile of median URR. Patient staffing ratios and the proportion of patients receiving kidney transplants did not differ significantly across facility quartiles of median URR.

Table 1.
Patient and dialysis facilities characteristics across quartiles of patient median urea reduction ratio

Factors Associated with Failure of URR Target Achievement

Table 2 shows patient and facility factors associated with failure of achievement of a median URR of ≥65% in the final multivariable model. Younger age, shorter dialysis vintage, African-American race, and male sex were associated with higher odds of not meeting the URR target. Dialyzing with a catheter as opposed to an arteriovenous fistula or arteriovenous graft, and higher BSA were also strongly associated with higher odds of target failure.

Table 2.
Variables significantly associated with a patient not meeting the URR target of >65% in the final multivariable model

In addition, patient nonadherence variables, particularly treatment shortening, had a strong association with the outcome. Treatment skipping was also associated, and correlated with treatment shortening, and other non-nonadherence behavior such as interdialytic weight gains of >5.7% of dry weight and hyperphosphatemia.

We also compared patients with the worst adherence with treatment (highest decile of treatment shortening) to patients with the best adherence (lowest decile of treatment shortening) (Table 3). The nonadherent patients were 15 years younger, more likely to be men and African American and to have a high body mass index and poor BP control. These patients also had higher interdialytic fluid gains and serum phosphorous levels and longer prescribed durations of treatment. Although the comorbidity burden in this group was lower than the best nonadherence group, hospitalizations were higher. These patients tended to cluster in urban settings with lower median income. No other differences in facility characteristics, including staff to patient ratio and availability of dietitians or social workers, were observed.

Table 3.
Patient characteristics by treatment adherence as defined by lowest and highest decile of treatment shortening

Facility Effect Analysis

Table 4 shows the effects of successively adding patient- and facility-level characteristics on the explained variance in URR target achievement in total (column 1), which can be further divided into the “between-patient” variance (column 2) and the “between-facility” variance (column 3). The “between-patient” variance is due to patient factors whereas the “between-facility” variance is the “center effect”. In the base model (model 1), which is adjusted for facility random intercept only, we find that facility accounts for 11.5% of the variation in URR target achievement; the remaining 88.5% of the variability in URR target achievement is due to patient factors. With adjustment for patient factors (demographics, comorbidities, laboratory data, and nonadherence) in models 1 through 5, the unexplained variance both at the patient (88.5% to 53.5%) and facility level decreases (11.5% to 6.7%). There is a 42% reduction in “center effect” with adjustment for patient factors; that is, patient factors explain a large part of the center effect. Addition of the facility-level variables did not explain the patient- or facility-level variance further.

Table 4.
Measures of facility effect on patient URR target achievement after sequentially adjusting for patient and facility characteristics

Sensitivity Analyses

In sensitivity analyses examining the effect of secular trends in dialysis adequacy (2009 versus 2004), and using the inclusion/exclusion criteria that will be used under the final rule,9 we found that significant facility-level variation in URR target achievement remained in 2009. The unexplained facility-level variation in 2009 in the URR target was 8.1%.


Most patients undergoing hemodialysis in the United States receive adequate small solute clearance, as defined by a URR ≥65%. Consistent with results of past studies that have examined for a facility effect, our study shows that dialysis facility significantly influences the likelihood that the URR target will be met.1012 Patient factors are much more influential (accounting for 88% of the variance) than facility factors (accounting for 12% of the variance) in URR target achievement. The present study also shows considerable attenuation of the facility effect with case-mix adjustment: nearly 42% of the facility effect is explained by patient case-mix and nonadherence factors.

This is in contrast to an older study by Fink et al. who also quantified the patient and center level variation in dialysis dose in the ESRD Network 5 population from the year 1997, and determined that the variation from the “center” was greater than the variation from “individual level” covariates.11 Fink's study, and others showing higher mortality at centers with lower URRs, was instrumental in adoption of URR as a quality measure for dialysis care, as they suggested a strong facility effect beyond case mix for URR target achievement. However, these studies predate the widespread secular increase in URR over the past 10 years.10,13

Our findings regarding the effect of case-mix factors and patient nonadherence on achievement of the URR target are particularly relevant in the pay-for-performance era. One of the most influential factors relating to failure to achieve a median URR of ≥65% in our models was treatment shortening. Furthermore, we observed a strong correlation between treatment shortening and other nonadherence indicators, such as treatment skipping, poor phosphorous control, and increased intradialytic weight gain. Patients who regularly shortened their treatments were more likely to be young, African American, and men with a higher body surface area and yet fewer comorbid illnesses, suggesting that the poor nonadherence behavior may be related to social/psychologic factors rather than treatment-related adverse effects. Poor adherence has been associated with hospitalizations and long-term survival in dialysis patients, and although potentially modifiable, a successful strategy for improving nonadherence has not been found.14,15

Unfortunately, we believe that pay-for-performance measures may achieve the opposite and unintended result. Recent surveys of dialysis staff including physicians, nurses, and medical directors indicated that at least three fourths of respondents felt that “cherry picking” occurred sometimes or frequently in dialysis units, and the practice of excluding chronic late arrivals/no shows was considered to have the strongest impact on dialysis outcomes.16 With a 2% cut in payment legislated by Congress for 2011 and with additional reductions for failing to meet the URR quality indicator under the QIP, we suspect that cherry picking will increase, widening the disparity in care provided to this high-risk subgroup of patients.

This has in fact been shown in a study that evaluated BP control and HbA1C measurements pre- and postintroduction of a pay-for-performance (P4P) system in family medicine practices in the United Kingdom. Although improvements were noted in glycemic control across the board, African Americans and those with lower socioeconomic status had lesser improvements over Caucasians and those of higher socioeconomic status.6 This experience indicates that it would be important to examine care after P4P is introduced in the U.S. dialysis program and to particularly look for discrepancies across ethnic and socioeconomic groups. Alternatives to penalizing units that care for patients with complex medical and/or social problems are needed. For example, reimbursement should not be withheld if a low URR is followed by an increase in prescribed time or dialyzer size, creating a more tightly linked quality measure.17 Furthermore, if treatment shortening and suboptimal dialysis continued after appropriate dialysis prescription changes, the dialysis facility could obtain an informed “against medical advice” letter from the patient, thereby avoiding the penalty for a potentially nonmodifiable behavior. Applying the final rule calculations to DCI facilities in 2009 shows that 29% of facilities do not meet the URR target, and of these, almost one third (27%) miss it on the basis of one or two patients, suggesting that noncompliance, even for a small minority, could play a major role in determining the penalty from the URR P4P criteria.

The strengths of our analyses are that we used a large contemporary patient population of 10,069 prevalent hemodialysis patients across 173 facilities with geographical variation across the United States. We also performed extensive patient case-mix adjustment, beyond the existing literature using comorbidity, demographics, laboratory, and patient nonadherence. Finally, we incorporated facility-level variables in our regression models to confirm a true facility effect and to quantify the extent of facility variation in URR target achievement.

Our analysis has some limitations: All our dialysis facilities were from a single not-for-profit provider and therefore we were not able to capture the effect of dialysis ownership characteristics on outcomes. Second, an important consideration in our analysis is that we calculated achievement of the URR target measure based on the Core Performance Measures criteria (CPM), rather than the ESRD payment final rule, which was released only in August 2010 and after completion of the present analysis. Important differences between our calculation and the final rule include the use of the annual average URR in the final rule versus the quarterly median URR in the CPM. Furthermore, the final rule uses an achievement threshold of 96% of patients meeting the URR target, which is higher than the median URR in our 2004 data set (92%), but identical to our 2009 data, reflecting secular trends in improving dialysis adequacy. Despite these differences between the CPM criteria and the final rule, we performed sensitivity analyses (unadjusted) that showed significant between-facility variation in meeting URR targets still exist. We did not have the gamut of case-mix factors to determine the extent to which facility-level variation is explained by patient factors, as claims data (to code comorbid conditions) are currently unavailable, but we have no reason to believe results would have changed from those presented in Table 3.

Finally, we looked at a small number of facility-specific factors. It is possible that undesirable facility-level characteristics, such as unprofessional staff or poor communication, influence treatment shortening and skipping. We did not have information on medical director or nephrology provider contact time, and the availability of comprehensive psychologic/social supports to explore this possibility.15 We also acknowledge that the influence of patient factors may be even greater than estimated here, if other case-mix factors that were not accounted for in our study, are important to this outcome.

In conclusion, we find that a significant portion of the variability in dialysis dose target achievement is explained by patient case-mix factors and nonadherence. These findings highlight the potential risk of worsening “cherry picking” against disadvantaged patients for financial resources. Our study supports the use of a process-oriented quality measure at the facility level to improve patient nonadherence before implementation of the URR pay-for-performance measure.


Patient Population

During the last quarter (October 1 through December 31) of 2004, 12,747 patients received hemodialysis treatment at 173 Dialysis Clinic Inc. (DCI) facilities across the United States. We used the same exclusions as the Clinical Performance Measures Program7 to identify the study population. We excluded patients younger than 18 years of age as of October 1, 2004 (n = 53), patients not undergoing 3 times weekly in-center treatment (n = 261), patients who died before December 31, 2004 (n = 547), patients for whom at least one laboratory measurement for hemoglobin, albumin, or Kt/V was not reported in the fourth quarter (n = 546), and patients who started dialysis between September 1 and December 31, 2004 (n = 839). Dialysis facilities treating fewer than 20 patients (n = 400) and patients without matching CMS claims files were excluded (n = 32). This left 10,069 patients in the study population.

Sources of Predictor and Outcome Data.

Laboratory parameters, BP readings, delivered dialysis dose, and details of dialysis treatments, including the vascular access type used, were abstracted from DARWIN, a proprietary electronic medical information system used by all DCI facilities. We considered the patient to meet the CMS pay-for-performance URR target if the median achieved URR for the last quarter of 2004 was ≥65%.7 The facility percentage URR target achievement is defined as the number of patients meeting the URR target in the quarter divided by the total number of patients in the unit that quarter. Hemoglobin, albumin, Kt/V, urea, calcium, and phosphorus are measured at least monthly and are processed at a central laboratory in Nashville, Tennessee. For each patient, the last value of the month was averaged for the 3 months in the last quarter of 2004. The study was approved by the Institutional Review Board at Tufts Medical Center.


Information about patient comorbidity was based on two sources: (1) The comorbidity checklist on the Medical Evidence Form (Form 2728), which is completed for all Medicare-entitled patients at the start of dialysis, and (2) diagnostic codes accompanying CMS Institutional and Physician/Supplier Claims for the years 2002 through 2004.8 Standard methods were used to increase the specificity of using claims data to identify comorbid conditions.18,19 Diagnostic codes accompanying institutional claims for which the facility was a hospital or rehabilitation facility were included, but diagnoses accompanying laboratory tests or diagnostic imaging studies were excluded because these may have been ordered as screening tests. Diagnostic codes accompanying equipment service types were also excluded because these are not entered by physicians. For the remaining physician/supplier claim types, we required three uses of a diagnostic code, each instance separated from the previous instance by a minimum of 14 days, for the condition to be considered to be present. Dates of hospitalizations were obtained from DARWIN, which tracks missed dialysis treatments, and the reasons for the patient absences, with weekly data reconciliation. This process provides reliable information about dates of hospitalization.


We defined our primary nonadherence variables as treatment shortening or treatment skipping. Additional nonadherence variables included intradialytic weight gain >5.7% of body weight and a serum phosphorous level >7.5 mg/dl.14,20 Shortening was defined as the difference between prescribed and delivered dialysis treatment time divided by the total time prescribed for all treatments in the quarter. Treatment skipping was defined as a binary variable, with at least one skipped treatment during the quarter versus none.15 Patients in the highest decile of treatment shortening were compared with those in the lowest decile to highlight differences in patient and facility characteristics that may determine nonadherence.

Center Characteristics

Center characteristics for the dialysis facilities were derived from the CMS-Form 2744, the annual ESRD Facility Survey completed by all Medicare-approved dialysis providers. Geographical factors including urban/rural status and median income at facility zip code were obtained from U.S. Census data for the year 2000. A facility was considered rural if >50% of the population in the facility's zip code resided in areas designated as rural.

Statistical Analysis

Patient and facility characteristics were compared across quartiles of facility URR target achievement. Facility median URR was calculated by taking the median of the 3-month median URR for each individual across all patients. ANOVA and χ2 tests were used to compare continuous and categorical variables respectively with trend P values. To assess the magnitude of dialysis facility effect on URR target achievement, we used a two-level logistic regression model with patients nested within dialysis facilities.

  • πij = probablity of meeting URR target of patient i within dialysis facility j
  • Logit(πij) = αj + ∑βixij
  • αj = α + uj

uj is normally distributed with mean 0 and between dialysis facility variance σu2 and xij indicates characteristics of the ith patient in the jth dialysis facility.

To partition the variation in URR target achievement at facility and patient levels, we used an R2 measure derived from a two-level logistic regression with random intercept.21,22 This partition makes the assumption that the binary URR target is the result of an observed continuous latent variable and that the level-1 residuals for this latent variable model has a logistic distribution. The random intercept models for the binary URR target were fitted using the GLIMMIX procedure in SAS (version 9.2).

Sensitivity Analyses

We calculated the unadjusted patient- and facility-level variation in the 2009 data set derived using the exclusion criteria from the Final Rule, which was implemented in January 2011.9




The authors are grateful for grant support for N.T.: KRESCENT Post-Doctoral Fellowship (Kidney Foundation of Canada, Canadian Society of Nephrology, and the Canadian Institute of Health Research) and also for NIH 5 K23 DK066273 (D.C.M.).


Published online ahead of print. Publication date available at www.jasn.org.


1. Doran T, Fullwood C, Gravelle H, Reeves D, Kontopantelis E, Hiroeh U, Roland M.: Pay-for-performance programs in family practices in the United Kingdom. N Engl J Med 355: 375–384, 2006. [PubMed]
2. Rosenthal MB, Landon BE, Normand SL, Frank RG, Epstein AM.: Pay for performance in commercial HMOs. N Engl J Med 355: 1895–1902, 2006. [PubMed]
3. Landon BE, Normand SL, Blumenthal D, Daley J.: Physician clinical performance assessment: Prospects and barriers. JAMA 290: 1183–1189, 2003. [PubMed]
4. Millett C, Gray J, Bottle A, Majeed A.: Ethnic disparities in blood pressure management in patients with hypertension after the introduction of pay for performance. Ann Fam Med 6: 490–496, 2008. [PMC free article] [PubMed]
5. Millett C, Gray J, Saxena S, Netuveli G, Khunti K, Majeed A.: Ethnic disparities in diabetes management and pay-for-performance in the UK: The Wandsworth Prospective Diabetes Study. PLoS Med 4: e191, 2007. [PMC free article] [PubMed]
6. Millett C, Netuveli G, Saxena S, Majeed A.: Impact of pay for performance on ethnic disparities in intermediate outcomes for diabetes: A longitudinal study. Diabetes Care 32: 404–409, 2009. [PMC free article] [PubMed]
7. Centers for Medicare & Medicaid Services: 2007 Annual Report, end stage renal disease clinical performance measures project. Department of Health and Human Services, Centers for Medicare & Medicaid Services, Office of Clinical Standards & Quality, Baltimore, Maryland, December 2007. http://www.cms.hhs.gov/CPMProject/Downloads/ESRDCPMYear2007Report.pdf
8. U.S. Department of Health & Human Services: Centers for Medicare & Medicaid Services, 2008, https://www.cms.gov/
9. U.S. Department of Health & Human Services, Centers for Medicare & Medicaid Services: Overview End Stage Renal Disease Payment, 2011, http://www.cms.gov/ESRDPayment/
10. Charra B.: Improving adequacy improves haemodialysis outcome. EDTNA ERCA J 26: 6–10, 19, 2000. [PubMed]
11. Fink JC, Blahut SA, Briglia AE, Gardner JF, Light PD.: Effect of center- versus patient-specific factors on variations in dialysis adequacy. J Am Soc Nephrol 12: 164–169, 2001. [PubMed]
12. Fink JC, Gardner JF, Armistead NC, Turner MS, Light PD.: Within-center correlation in dialysis adequacy. J Clin Epidemiol 53: 79–85, 2000. [PubMed]
13. Charra B, Calemard E, Ruffet M, Chazot C, Terrat JC, Vanel T, Laurent G.: Survival as an index of adequacy of dialysis. Kidney Int 41: 1286–1291, 1992. [PubMed]
14. Kimmel PL, Peterson RA, Weihs KL, Simmens SJ, Alleyne S, Cruz I, Veis JH.: Psychosocial factors, behavioral compliance and survival in urban hemodialysis patients. Kidney Int 54: 245–254, 1998. [PubMed]
15. Saran R, Bragg-Gresham JL, Rayner HC, Goodkin DA, Keen ML, Van Dijk PC, Kurokawa K, Piera L, Saito A, Fukuhara S, Young EW, Held PJ, Port FK.: Nonadherence in hemodialysis: Associations with mortality, hospitalization, and practice patterns in the DOPPS. Kidney Int 64: 254–262, 2003. [PubMed]
16. Desai AA, Bolus R, Nissenson A, Chertow GM, Bolus S, Solomon MD, Khawar OS, Talley J, Spiegel BM.: Is there “cherry picking” in the ESRD Program? Perceptions from a Dialysis Provider Survey. Clin J Am Soc Nephrol 4: 772–777, 2009. [PMC free article] [PubMed]
17. Kerr EA, Smith DM, Hogan MM, Hofer TP, Krein SL, Bermann M, Hayward RA.: Building a better quality measure: Are some patients with ‘poor quality’ actually getting good care? Med Care 41: 1173–1182, 2003. [PubMed]
18. Klabunde CN, Potosky AL, Legler JM, Warren JL.: Development of a comorbidity index using physician claims data. J Clin Epidemiol 53: 1258–1267, 2000. [PubMed]
19. Klabunde CN, Warren JL, Legler JM.: Assessing comorbidity using claims data: An overview. Med Care 40: IV-26–IV-35, 2002 [PubMed]
20. Kimmel PL, Peterson RA, Weihs KL, Simmens SJ, Boyle DH, Umana WO, Kovac JA, Alleyne S, Cruz I, Veis JH.: Psychologic functioning, quality of life, and behavioral compliance in patients beginning hemodialysis. J Am Soc Nephrol 7: 2152–2159, 1996. [PubMed]
21. Larsen K, Merlo J.: Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol 161: 81–88, 2005. [PubMed]
22. Snijders TAB BR.: Multilevel analysis: an introduction to basic and advanced multilevel modelling. London, Sage, 1999

Articles from Journal of the American Society of Nephrology : JASN are provided here courtesy of American Society of Nephrology
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Compound
    PubChem Compound links
  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles
  • Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...