Among the most striking changes in the dialysis industry over the last two decades has been the growth of dialysis chain organizations (i.e., multi-unit facility enterprises). According to the United States Renal Data System (USRDS), the number of chain-owned units has grown over 11-fold since 1990 (1). Currently, more than 85% of facilities are affiliated with a chain organization, including roughly 85% of all for-profit and 33% of nonprofit facilities (2). The tremendous consolidation of independent dialysis facilities and small chains into large corporate for-profit dialysis chains have raised concerns for the quality of care delivered to end-stage renal disease (ESRD) patients (3–5).
Renal transplantation therapy (RTT) is the preferred and optimal treatment for ESRD because it provides a superior quality of life, improved long-term survival, and is more cost-effective than dialysis (6, 7). Despite strict equal-access regulations to ensure ethical and equitable distribution of organs, previous research has shown that not all ESRD patients are provided equivalent access to the kidney transplant waitlist (8). Both patient- and facility-level factors have been shown to influence access and allocation of cadaveric donor kidneys in the United States (9, 10). The impact of facility ownership on access to renal transplants suggests that the legal structure of organizations providing dialysis influences their organizational and clinical practices. For example, Garg et al., using 1996 data, found that for-profit ownership of dialysis facilities was associated with decreased placement on the renal transplant waiting list (8). However, the impact of other aspects under which dialysis centers operate—such as their affiliation with larger chain corporations—on access to RTT is not known, nor is the potential ways in which chain affiliation might mediate the ownership-related differences in access documented in previous research.
Researchers have speculated that a number of possible pathways might influence how chain affiliation affects patient access to RTT, including influences that might reduce or expand ownership-related differences. In this study, we assessed the impact of dialysis facility chain status—measured as affiliation, size, and chain ownership—on access to renal transplantation among the U.S. hemodialysis population.
Distribution of Patients and Facilities by Dialysis Facility Chain Status
There were 66,780 adult ESRD patients who started hemodialysis in 2006 (Fig. 1). After excluding 2,992 patients older than 70 years (and therefore less suitable for transplantation), 2,342 patients with history of kidney transplants or waitlisting, 2,156 patients with a living donor (therefore not a consequence of referrals from the dialysis center), and 2,476 patients with incomplete baseline information, the final study cohort comprised 56,714 patients receiving dialysis at 4,465 dialysis facilities. More than 60% of patients received dialysis in one of four large dialysis chains that comprised 100 or more facilities (58% in for-profit chains 1, 2, and 3; and 3% in nonprofit chain 4), 14% in mid or small chains that comprised less than 100 facilities, and 25% in nonchain facilities (10% in independent freestanding and 15% in hospital-based facilities). To ensure non-differential exclusion of patients, we examined the distribution of patients who were excluded from the study cohort among seven facility categories, and the proportions are similar to what is reported for the entire study cohort.
Patient and Facility Characteristics by Facility Chain Status
Comorbid conditions were highly prevalent among dialysis patients in all facilities. Before the start of dialysis, 57% had diabetes, 50% had cardiovascular disease, 9% had one or more traumatic conditions such as an amputation, and 11% needed assistance with daily activities (data not shown). Patient characteristics such as gender, age, race, Hispanic ethnicity, income, and insurance status were significantly different by facility chain status (P<0.0001, Table 1). Patient case mix including comorbid conditions, assistance with daily living, BMI, lifestyle choices, and serum albumin varied as well by facility chain status (P<0.0001, Table 1). Pre-ESRD care, including involvement of a nephrologist or dietitian, and type of vascular access also differed by facility type.
Facility characteristics were also significantly different by chain status (P<0.0001, Table 1). The majority of chain facilities are for-profit except chain 4, the nation’s largest nonprofit chain; 10% and 15% of mid and small chain and nonchain freestanding facilities were nonprofit, respectively. Almost all hospital-based facilities are nonprofit (97%). Facility size (number of dialysis patients), geographic location, rural status, racial composition, and staffing patterns differed by facility chain status. All large chains were more likely to be found in the Southeast U.S. region whereas hospital-based facilities were more likely to be located in Northeast. Number of registered nurses and social workers per 100 patients were highest among chain 4 facilities and lowest among chain 1 facilities.
Association Between Chain Affiliation and Placement on the Waiting List
Mean patient follow-up time was 27 months. During the follow-up period, 19% of all patients were waitlisted, 40% died, and 41% reached the end of follow-up on December 31, 2009 without being waitlisted. Compared to the largest nonprofit chain 4, there was no difference in the likelihood of placement on the renal transplant waitlist among other facility types including the three other large for-profit chains, mid and small chains, and independent facilities (Table 2). Indeed, no significant difference was found between largest chains 1 to 4, mid and small chains, and nonchain independent facilities regardless of reference group (data not shown). Consistently, there was no difference in the likelihood of being waitlisted by chain affiliation (chain vs. nonchain: adjusted HR=1.08, 95% CI 0.98–1.19).
Effect of Chain Size and Ownership on Placement on the Waiting List
In contrast to chain affiliation, a significant relationship was found between chain size and placement on the waiting list: compared to smaller chains, large chains were 8% less likely to place patients on the waiting list (95% CI 0.84–1.00) (Table 2). Furthermore, chain ownership was associated with the likelihood of placement on the renal waitlist; compared to nonprofit chain facilities, patients from for-profit chain facilities were 13% (95% CI 0.77–0.98) less likely to be waitlisted (Table 3). However, among nonchains, ownership did not influence the likelihood of being waitlisted (HR=0.92, 95% CI 0.80–1.05. Overall, patients from for-profit facilities were 13% (95% CI 0.78–0.97) less likely to be waitlisted compared to nonprofit facilities. Other covariates adjusted in the multivariate regression model in addition to facility ownership significantly predict placement on the waiting list and are shown in Table 4. In contrast to previous studies using earlier data that found racial disparities in waitlist placement (11), we found similar rates in our study between blacks and whites, suggesting that the racial gap in waitlisting appears to have diminished in recent years. Notably, the only significant facility-level predictor was geographic region; patients residing in the East had decreased waitlisting compared to those in the West.
Secondary and Sensitivity Analyses
In the primary analysis, patients treated in hospital-based facilities are excluded because data on staffing was not available for these dialysis units. But this omits a substantial portion of patients treated under nonprofit auspices; in secondary analyses, we included hospital-based facilities in our analyses and the results for both chains versus nonchain and for-profit versus nonprofit ownership did not change (Table S1, SDC, https://links.lww.com/TP/A967). We performed other sensitivity analyses to ensure robust study results. This included exclusion of facility characteristics that might be affected by dialysis facility chain or ownership status, including patient comorbidities, serum albumin levels, disease severity, and facility variables such as staffing patterns. After removal of these variables (individually and as a group), the estimated hazard ratios were similar to those reported herein.
To address the potential collinearity between income and insurance status, and race and ethnicity, we selected one of the variables in each set, and results did not change. We also performed an analysis to test the proportional hazards (PH) assumption by examining the interaction effect between facility chain status and follow-up time (cubic splines). Results were not found to be significant, so the PH model assumption is justified. To test whether assigning the facility type to each patient at the time of dialysis initiation (rather than at another point in time) influenced results, we repeated the analyses for patients who survived at least 90 days and did not switch dialysis facilities during the first 90-day period (51,948 patients), and the results did not materially change. Finally, analyses were repeated by adjusting for the variables with missing values (i.e., nephrology care, dietitian care, and serum albumin level) in two ways: (1) removing patients with missing values and (2) generating a separate category for the missing group. Again, results were similar to those reported above.
Using the most recent data available on all incident U.S. hemodialysis patients, this study examined the impact of chain affiliation, size, and ownership on the likelihood of being waitlisted for a renal transplant. Although dialysis chain affiliation was not found to be independently associated with placement on waiting list, both chain ownership and size were shown to be significant explanatory factors in predicting access to transplants. Compared to nonprofit chain facilities, patients from for-profit chain facilities were 13% less likely to be waitlisted and large chains compared with mid and small chains were 8% less likely to place patients on the waiting list.
There are two possible explanations for the pattern observed in our study. The first involves a loss of community control as organizations affiliate with a chain (12, 13). Under this scenario, for-profit ownership creates incentives to cut corners in terms of quality, but local community control buffers these incentives (potentially through stronger professional norms among clinicians or greater influence by patient advocacy groups) and reduces the extent to which they translate into clinical practice. This appears to be the case with nonchains in our study, but not with chains. The second possible explanation involves differences extending from the role of investors for large for-profit corporations. When these companies are publicly traded, the value of their stock depends upon maintaining what is often an unsustainable rate of revenue growth (14). For example, for-profit dialysis providers might be less willing to refer a patient for evaluation for RTT (especially the healthiest patients who are most likely to then receive a renal transplant) thereby removing a constant stream of revenue from their facility (15–17). Furthermore, lower levels of staffing in for-profit facilities could result in insufficient attention to patient education, which is an important component of the informed decision making regarding ESRD treatment options including transplantation. Because large nonprofit chains are not subject to the same sort of investor pressures, their behavior appears more distinct than do the practices of unaffiliated nonprofit and for-profit agencies.
The extant literature on multi-institutional healthcare systems identifies several different ways in which chain ownership and size might influence organizational behavior (18). Some of these pathways suggest that chain ownership might reduce the ownership-related differences identified in previous research, whereas others suggest, in contrast, that ownership-related differences might expand under chain affiliation (19). Because the majority of these theories predict that chains would exert isomorphic pressures, it is often presumed that expanded corporate control induces nonprofit and for-profit performance to become more similar (20, 21). In fact, however, the opposite effect has been observed among hospitals (22), insurance companies (23), and in other aspects of dialysis (22). This same pattern emerged here for access to transplantation: profit status did not influence placement on the renal waitlist among nonchains, but had a significant impact on access among chain-affiliated dialysis facilities.
As the first major payment reform in nearly 30 years, Congress has enacted the ESRD Prospective Payment System in January 2011, bundling almost all routine services used by dialysis patients. Concerns have arisen regarding potential “cherry picking” of the healthiest, least costly patients (24). Specifically, that providers most concerned with profit margins (those belonging to large for-profit dialysis chains) would perhaps be least likely to place their healthiest patients on transplant waitlists thereby losing the revenues from these patients at a more cost-conscious time. Future research is warranted to see if the findings observed in this study between 2006 and 2009 of a decreased likelihood of waitlist placement in for-profit chains have become more pronounced after implementation of ESRD PPS. Continued research to ensure that all interested and medically suitable individuals have an equal chance of obtaining a renal transplant is warranted.
The observed relationship between chain ownership and size and access to transplantations should not be interpreted as causal given the observational nature of the study design. Our study has several additional limitations. First, because data on referral by the dialysis facility for evaluation for possible transplantation was not available, we used the placement on the waiting list as a proxy measure for access to transplantation. Given placement on the waiting list is a crucial and unambiguous step toward the receipt of a cadaveric transplant, we feel its use as our study outcome is appropriate. Furthermore, because transplantation centers that receive referred patients are not assumed to play a role in the link between dialysis facility to referral, the use of this proxy is likely to have biased our results only if there are significant disparities in patient health by dialysis facility organizational status resulting in different proportions of patients placed on the waiting list by transplantation centers. However, previous studies have suggested that compared to patients treated in for-profit dialysis facilities, patients treated in nonprofit facilities had either similar (9) or worse health status (25, 26). Therefore, if such a bias exits, our findings would be conservatively biased toward the null by using waitlist placement (vs. an antecedent end point such as referral for evaluation for transplantation).
Second, receiving a transplant involves a series steps related to medical suitability, interest in transplantations, pretransplantation workup, and movement up a waiting list to eventual transplantation. Given use of observational Medicare data, we were not able to capture and adjust for potential barriers for each of these steps such as health literacy of the patients, misinformation, poor provider communication, and misperceptions or decision-making capacity (27, 28).
Third, misclassification in the assignment of facility type might arise if patients switched facilities or facility type during the period between study baseline and waitlisting. We addressed the former in a sensitivity analysis and the latter (switching facility types) is rare according to our analysis. Fourth, we can not entirely rule out residual confounding despite the adjustment for numerous patient and facility characteristics. It is documented that for-profit facilities tend to locate in certain economically advantageous areas, instead of locating randomly (29). Adjusting for income differences and geographic region based on ESRD networks might not fully address geographic difference. Other facility variables, not captured here, such as local market competition, might also underlie the disparities observed in this study. Future prospective studies with more complete confounding control should be conducted to confirm our study findings. Despite these limitations, our study represents the first to evaluate the role of dialysis chain status—currently the dominant organizational paradigm in the dialysis industry—in obtaining a renal transplant.
MATERIALS AND METHODS
Study Design and Sample Selection
This study is a retrospective cohort design of incident hemodialysis patients and their providers. Peritoneal patients were excluded because of differences in case mix, higher rates of transplantation, and overrepresentation in nonprofit facilities compared to hemodialysis patients (30–33). Similar to the strategy of Garg et al. (8), because the steps needed to be placed on the renal waitlist are usually taken early in the course of ESRD (34), each patient was assigned to the specific facility where she was receiving care at the initiation of hemodialysis. Follow-up started at the date of hemodialysis initiation (study baseline) and ended on December 31, 2009, or death, or placement on the United Network for Organ Sharing (UNOS) waiting list for a cadaveric transplant, whichever occurred earlier. Given the unlikelihood of obtaining a transplant after age 70, only adults between 18 and 70 years were deemed eligible for study inclusion. Patients with a previous transplant or waitlisting history, a living donor transplant at any time during the study, or incomplete baseline information were also excluded.
The primary data source was the USRDS standard analytical files (SAFs). The variables included in the USRDS SAFs, as well as the data source, collection methods, and validation studies, are described at the USRDS website, http://www.usrds.org. Specifically, we used USRDS files to collect patient demographics, clinical history, and patient dialysis provider on all ESRD patients initiating hemodialysis in 2006, including patients who used Medicare as secondary payor. As a special data request, USRDS provided the provider number for each patient at dialysis initiation, thereby allowing us to include all patients in our analysis. Kidney transplant waitlist information was determined from UNOS data and linked to USRDS. Facility organizational status was obtained from the quarterly updated CMS Dialysis Facility Report. The 2000 U.S. Census data were used to obtain information on several ecological characteristics including median household income, percent white population, and urbanicity status. All files were merged based on unique patient and facility identifiers.
Patient and Facility Variables
The study endpoint is defined as first placement on the deceased (cadaveric) donor waitlist during the study follow-up period. The main exposure of interest is facility chain status identified by seven facility types including four large chains (chains 1–4 with >100 dialysis units), mid and small chains (≤100 units), independent, and hospital-based (nonchain freestanding facilities), representing the full spectrum of facilities providing outpatient dialysis to U.S. patients. Facilities were also disaggregated by chain size, affiliation, and ownership. Several other facility-level characteristics, including size (number of beds), geographic region, staffing patterns, urbanicity, and area racial composition, were included in the model. Numerous patient-level characteristics, listed in Table 1, known to be potential confounders (35–37) were used to adjust for differences in case mix among facilities.
Crude waitlisting rates per 100 person-years of dialysis were calculated. To disentangle the facility-level factors from patient determinants that might influence access to the renal waitlist, three-level pooled regression modeling (random-effects regression or hierarchical regression modeling) (38, 39) with facilities at the third level, patients at the second level, and patient months at the first level was used. Specifically, observations over 36 monthly intervals were pooled into a single sample and a random intercept model that treats the dialysis facility effects as random effects was fit with patient months during follow-up as the unit of analysis. Pooled logistic regression, also called discrete-time hazard modeling (40), was conducted using SAS GLIMMIX procedure to estimate hazard ratios (HRs) comparing the incidence of waitlist placement in each facility group (41, 42). This procedure accounts for the within-facility and within-chain correlations or clustering. Because patients treated in hospital-based facilities might be more complex compared to patients treated in freestanding facilities (43), hospital-based facilities were excluded from the primary analyses and analyzed separately. All analyses were performed using SAS, version 9.1.3 for Linux. Statistical significance is defined as a P value less than 0.05.
1. United States Renal Data System. USRDS
2011 Annual Data Report. Bethesda, MD: National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Disease. Available at: http://www.usrds.org/atlas11.aspx
. Accessed May 2013.
2. United States Renal Data System. USRDS
2005 Annual Data Report. Bethesda, MD: National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Disease. Available at: http://www.usrds.org/adr.htm
. Accessed May 2013.
3. Zhang Y, Cotter DJ, Thamer M. The effect of dialysis chains on mortality among patients receiving hemodialysis. Health Serv Res 2011; 46: 747.
4. Medicare Payment Advisory Commission (MedPAC). Chapter 6. Outpatient dialysis services. In: Report to the Congress: Medicare Payment Policy. Washington, D.C.: 2012: 141.
5. Hirth RA, Turenne MN, Wheeler JR, et al. Do resource utilization and clinical measures still vary across dialysis chains after controlling for the local practices of facilities and physicians? Med Care 2010; 48: 726.
6. Laupacis A, Keown P, Pus N, et al. A study of the quality of life and cost-utility of renal transplantation. Kidney Int 1996; 50: 235.
7. Wolfe RA, Ashby VB, Milford EL, et al. Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 1999; 341: 1725.
8. Gaston RS, Danovitch GM, Adams PL, et al. The report of a national conference on the wait list for kidney transplantation. Am J Transplant 2003; 3: 775.
9. Garg PP, Frick KD, Diener-West M, et al. Effect of the ownership of dialysis facilities on patients’ survival and referral for transplantation. N Engl J Med 1999; 341: 1653.
10. Alexander GC, Sehgal AR Transplant Task Force of the Renal Network, Inc. Variation in access to kidney transplantation across dialysis facilities: using process of care measures for quality improvement. Am J Kidney Dis 2002; 40: 824.
11. Epstein AM, Ayanian JZ, Keogh JH, et al. Racial disparities in access to renal transplantation—clinically appropriate or due to underuse or overuse? N Engl J Med 2000; 343: 1537.
12. Alexander JA, Weiner BJ, Succi M. Community accountability among hospitals affiliated with health care systems. Milbank Q 2000; 78: 157.
13. Proenca EJ, Rosko MD, Zinn SJ. Community orientation in hospitals: an institutional and resource dependence perspective. Health Serv Res 2000; 35: 1011.
14. Gray B, Schlesinger M. Health. In: Salamon L, eds. The State of the Nonprofit Sector. Washington, DC: Brookings Institution Press; 2002: 65.
15. Bovbjerg RR, Held PH, Diamond LH. Provider-patient relations and treatment choice in the era of fiscal incentives: the case of the end-stage renal disease program. Milbank Q 1987; 65: 177.
16. Kucirka LM, Grams ME, Balhara KS, et al. Disparities in provision of transplant information affect access to kidney transplantation. Am J Transplant 2012; 12: 351.
17. Gardner KD Jr. Profit and the end-stage renal disease program. N Engl J Med 1981; 305: 461.
18. DiMaggio PJ, Anheier HK. The sociology of nonprofit organizations and sectors. Annu Rev Sociol 1990; 16: 137.
19. Pozniak AS, Hirth RA, Banaszak-Holl J, et al. Predictors of chain acquisition among independent dialysis facilities. Health Serv Res 2010; 45: 476.
20. Alexander JA, Weiner BJ. The adoption of the corporate governance model by nonprofit organizations. N Engl J Med 1998; 8: 223.
21. Schlesinger M. Mismeasuring the consequences of ownership: external influences and the comparative performance of public, for-profit and nonprofit organizations. In: Powell WW, Clemens ES, eds. Private Action and the Public Good. New Haven, Conn: Yale University Press; 1998: 85.
22. Schlesinger M, Gray B. Nonprofit organizations and health care: the paradox of persistent attention. In: Powell WW, Steinberg R, eds . The Nonprofit Sector: A Research Handbook. 2nd ed. New Haven, CT: Yale University Press; 2004.
23. Schlesinger M, Quon N, Wynia M, et al. Profit-seeking, corporate control and the trustworthiness of health care organizations: assessments of health plan performance by their affiliated physicians. Health Serv Res 2005; 40: 605.
24. Desai AA, Bolus R, Nissenson A, et al. Is there “cherry picking” in the ESRD program? Perceptions from a dialysis provider survey. Clin J Am Soc Nephrol 2009; 4: 772.
25. Dalrymple LS, Johansen KL, Romano PS, et al. Comparison of hospitalization rates among for-profit and nonprofit dialysis facilities. Clin J Am Soc Nephrol 2014; 9: 73.
26. Foley RN, Fan Q, Liu J, et al. Comparative mortality of hemodialysis patients at for-profit and not-for-profit dialysis facilities in the United States, 1998 to 2003: a retrospective analysis. BMC Nephrol 2008; 9: 6.
27. Kazley AS, Simpson KN, Chavin KD, et al. Barriers facing patients referred for kidney transplant cause loss to follow-up. Kidney Int 2012; 82: 1018.
28. Grubbs V, Gregorich SE, Perez-Stable EJ, et al. Health literacy and access to kidney transplantation. Clin J Am Soc Nephrol 2009; 4: 195.
29. Hirth RA. Competition between for-profit and nonprofit health care providers: can it help achieve social goals? Med Care Res Rev 1997; 54: 414.
30. Thamer M, Zhang Y, Marayati F, et al. Factors influencing route of administration for epoetin treatment among hemodialysis patients in the United States. Am J Kidney Dis 2006; 48: 77.
31. Stack A. Determinants of modality selection among incident US dialysis patients: results from a national study. J Am Soc Nephrol 2002; 13: 1279.
32. Weinhandle ED, Foley RN, Gilbertson DT, et al. Propensity-matched mortality comparison of incident hemodialysis and peritoneal dialysis patients. J Am Soc Nephrol 2010; 21: 499.
33. Fuquay R, Teitelbaum I. Transplant outcomes and dialysis modality. Contrib Nephrol 2012; 178: 251.
34. Alexander GC, Sehgal ARTransplant Task Force of the Renal Network, Inc. Variation in access to kidney transplantation across dialysis facilities: using process of care measures for quality improvement. Am J Kidney Dis 2002; 40: 824.
35. Garg PP, Diener-West M, Powe NR. Income-based disparities in outcomes for patients with chronic kidney disease. Semin Nephrol 2001; 21: 377.
36. Ozminkowski R, White A, Hassol A, et al. What if socioeconomics made no difference: access to cadaver kidney transplant as an example. Med Care 1998; 36: 1398.
37. Rodriguez RA, Sen S, Mehta K, et al. Geography matters: relationships among urban residential segregation, dialysis facilities, and patient outcomes. Ann Intern Med 2007; 146: 493.
38. Austi P. Comparing hierarchical modeling with traditional logistic regression analysis among patients hospitalized with acute myocardial infarction: should we be analyzing cardiovascular outcomes data differently? Am Heart J 2003; 145: 27.
39. Rice N, Leyland A. Multilevel models: applications to health data. J Health Serv Res Policy 1996; 1: 154.
40. D’Agostino RB, Lee ML, Belanger AJ, et al. Relation of pooled logistic regression to time dependent Cox regression analysis: the Framingham Heart Study. Stat Med 1990; 12: 1501.
41. Bailie GR, Mason NA, Bragg-Gresham JL, et al. Analgesic prescription patterns among hemodialysis patients in the DOPPS: potential for underprescription. Kidney Int 2004; 65: 2419.
42. U.S. Renal Data System. USRDS
2002 Annual Data Report. Bethesda, MD: National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Disease. Available at: http://www.usrds.org/adr.htm
. Accessed May 2013.
43. U.S. Renal Data System. USRDS
2007 Annual Data Report. Bethesda, MD: National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Disease. Available at: http://www.usrds.org/adr.htm
. Accessed May 2013.