Association Between Nephrologist Ownership of Dialysis Facilities and Clinical Outcomes

Key Points

Question  Is nephrologist ownership of US dialysis facilities associated with worse clinical outcomes?

Findings  In this cross-sectional cohort study of 251 651 adults receiving dialysis, nephrologist ownership of dialysis facilities was associated with increased home dialysis use and decreased erythropoietin-stimulating agent use but not adverse outcomes associated with lower doses, such as more severe anemia or increased blood transfusions.

Meaning  This study found that nephrologist ownership was not associated with worse clinical outcomes, suggesting that such ownership may be associated with improved care quality in some domains: home dialysis and erythropoietin-stimulating agent use.

Importance  Ownership of US dialysis facilities presents a financial conflict of interest for nephrologists, who may change their clinical practice to improve facility profitability.

Objective  To investigate the association between nephrologist ownership of freestanding dialysis facilities and clinical outcomes.

Design, Setting, and Participants  This cross-sectional study was conducted using US Renal Data System data linked to a data set of freestanding nonpediatric dialysis facility owners. Participants were a sample of all adults with fee-for-service Medicare receiving dialysis for end-stage kidney disease from January 2017 to November 2017 at included facilities. Data were analyzed from April 2020 through August 2022.

Exposures  Outcomes associated with nephrologist ownership were assessed using a difference-in-differences analysis comparing the difference in outcomes between patients treated by nephrologist owners and patients treated by nonowners within facilities owned by nephrologists after accounting for differences in patient outcomes between nephrologist owners and nonowners in other facilities.

Main Outcomes and Measures  Outcomes plausibly associated with nephrologist ownership were evaluated: (1) treatment volumes (missed treatments and transplant waitlist status); (2) erythropoietin-stimulating agent (ESA) use and related outcomes (anemia, defined as hemoglobin level <10 g/dL, and blood transfusions), (3) quality metrics (mortality, hospitalizations, 30-day readmissions, hemodialysis adequacy, arteriovenous fistula use, and hemodialysis catheter use for ≥3 months), and (4) home dialysis use.

Results  A cohort of 251 651 patients (median [IQR] age, 66 [46-85] years; 112 054 [44.5%] women; 9765 Asian [3.9%], 86 837 Black [34.5%], and 148 617 White [59.1%]; 38 938 Hispanic [15.5%]) receiving dialysis for end-stage kidney disease were included. Patient treatment by nephrologist owners at their owned facilities was associated with a 2.4 percentage point (95% CI, 1.1-3.8 percentage points) higher probability of home dialysis, a 2.2 percentage point (95% CI, 3.6-0.7 percentage points) lower probability of receiving an ESA, and no significant difference in anemia or blood transfusions. Patient treatment by nephrologist owners at their owned facilities was not associated with differences in missed treatments, transplant waitlisting, mortality, hospitalizations, 30-day readmissions, hemodialysis adequacy, or fistula or long-term dialysis catheter use.

Conclusions and Relevance  This cross-sectional cohort study found that nephrologist ownership was associated with increased home dialysis use, decreased ESA use, and no change in anemia or blood transfusions.

Dialysis in the US is a multibillion-dollar industry associated with suboptimal clinical outcomes.¹ Some dialysis facilities are owned by nephrologists, either wholly or partially through joint venture partnerships with corporate owners. Although it is legal, nephrologist ownership creates financial conflicts of interest because it introduces incentives to change clinical management.² Little is known about nephrologist ownership of dialysis facilities. Recently, the Centers for Medicare & Medicaid Services (CMS) requested information on whether nephrologist ownership requires more transparency and on its association with patient care.³

Health care payers pay nephrologists a monthly professional fee for supervising dialysis, the “monthly capitated payment” in Medicare.⁴ This professional fee does not vary based on intensity or frequency of dialysis or on prescribed medications. Separately, payers reimburse facilities for providing dialysis treatments (and related medications) in a per-treatment bundle. Facilities can maximize profits by reducing empty dialysis chairs (eg, minimizing missed treatments or reducing transplant referrals) and reducing costs (eg, providing fewer medications).⁴

Because nephrologists are not typically employed by facilities, nephrologists who are not owners can make clinical decisions free of personal financial gain, although they are subject to operating constraints imposed by facility corporate owners. Conversely, nephrologist owners who share in facility profits may increase their personal compensation by altering the clinical treatment of their patients within their owned facilities.

It is unclear whether nephrologist ownership is associated with patient outcomes. Patients may be harmed if practice changes contradict optimal patient care. Alternatively, nephrologist ownership may be associated with improved care quality because profitability may align with better care or because nephrologist owners may promote more patient-centered care. Moreover, nephrologist owners may have different incentives and constraints at large dialysis organizations (LDOs), which are publicly traded and prioritize shareholder profits, compared with other facility types. Because nephrologist ownership has increased over time,⁵ understanding its association with patient outcomes and heterogeneity by type of facility is paramount.

Using 2017 to 2018 data from CMS on dialysis facility owners linked to a national end-stage kidney disease patient registry, we investigated whether nephrologist ownership in freestanding dialysis facilities, wholly or through joint venture arrangements, was associated with outcomes nephrologists could plausibly influence. We compared nephrologist owners with non-owners within facilities owned by nephrologists after accounting for differences between the same nephrologists in other facilities.

The University of Southern California and University of Pennsylvania institutional review boards approved this cross-sectional cohort study. These boards granted waivers of informed consent because the study was a secondary data analysis and deemed to be of minimal risk to participants. We adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Through a Freedom of Information Act request to CMS, we obtained data on individual and organizational owners of freestanding dialysis facilities registered in the CMS Medicare Provider Enrollment, Chain, and Ownership System as of May 1, 2018. Because some owners were not nephrologists, we identified the subset of owners who were nephrologists and aggregated individual nephrologists into group practices using business addresses (see technical details in the eAppendix in the Supplement).

For simplicity, we call nephrologist-owned facilities owned facilities and nonnephrologist-owned facilities nonowned facilities. We call nephrologist owners owners and nephrologists who did not own facilities nonowners.

We used the US Renal Data System (USRDS), a national registry of US patients with end-stage kidney disease. The USRDS contains Medicare Part A and Part B claims for patients enrolled in fee-for-service Medicare.¹ The USRDS includes the Consolidated Renal Operations in a Web-Enabled Network (CROWNWeb), which contains monthly reports of clinical measures, including vascular access type, laboratory values, and hemodialysis adequacy (ie, whether a patient received sufficient dialysis or a single-pool Kt/V ≥1.2).

We identified adults with fee-for-service Medicare who received dialysis in US freestanding facilities between January 1 and November 30, 2017. We excluded patients receiving dialysis at facilities with more than 15% pediatric patients because pediatric facilities have different operating constraints. We also excluded patients treated by nephrologists or at facilities with fewer than 10 patients. To ensure complete comorbidity capture, we excluded patients with less than 12 months of Medicare prior to study inclusion. We included patients who died, received a transplant, or stopped receiving dialysis at an included facility.

We assigned each patient-month to a treating nephrologist-dialysis facility pair using physician and facility claims, respectively.⁴ Unassigned months, when patients missed physician visits or outpatient dialysis due to hospitalization or death, could result in biased outcomes. Thus, we assigned an unassigned month to the previous month’s nephrologist-facility pair. For patients with consecutive unassigned months, only the first was assigned. We aggregated patient-months to the patient level and assigned each patient the nephrologist-facility pair responsible for the plurality of patient-months. Our exposure was nephrologist ownership status, which stratified patients into those treated by nonowners at nonowned facilities, nonowners at owned facilities, owners at their own facilities, and owners at other facilities.

Simple comparisons between owned and nonowned facilities (or owners and nonowners) are confounded by unobserved differences across nephrologist and facility populations. Nephrologists may preferentially own facilities in wealthier areas, for example. We addressed this bias using a difference-in-differences approach. Conceptually, we first estimated differences in outcomes for owners at their owned facilities vs at facilities where they had no ownership stake (ie, nonowned facilities). Observable differences were attributable to prescription differences among owners between owned and nonowned facilities, selection bias from systematic differences between owned and nonowned facilities, or both. To eliminate selection bias due to systematic differences between owned and nonowned facilities, we estimated the same difference among nonowners and subtracted from the first difference. Equivalently, this difference in differences is the difference in outcomes between owners and nonowners within owned facilities after subtracting the difference between the same nephrologists in nonowned facilities.

We identified outcomes nephrologists can reasonably influence: (1) treatment volume (number of missed treatments and transplant waitlist status); (2) receipt of an erythropoietin-stimulating agent (ESA), which is an important cost for dialysis facilities, and outcomes related to ESAs, including a hemoglobin level of less than 10 g/dL (to convert to grams per liter, multiply by 10.0) and number of units of blood received from blood transfusions⁴; (3) quality metrics⁶ (mortality, hospitalizations, 30-day readmissions,^7,8 hemodialysis adequacy as measured by Kt/V, arteriovenous fistula use, and hemodialysis catheter use for ≥3 months); and (4) home dialysis use.^9-11 Analyses were at the patient level, with outcomes modeled as the mean event rate per month with treatment by the assigned nephrologist-facility pair. We winsorized blood transfusions by recoding outliers to a maximum of the ninety-ninth percentile.

Individually, these outcomes are not necessarily associated with better care or increased profitability. Such an evaluation requires assessing the entirety of outcomes in context. For instance, decreased ESA use may be good for clinical outcomes (if it is associated with lower rates of inappropriately high hemoglobin levels) or bad (if it is associated with increased blood transfusions).

We estimated a patient-level linear regression model of each outcome as a function of facility indicators (fixed effects), nephrologist fixed effects, and an indicator for whether the treating nephrologist owned the treating facility (ie, an interaction between whether the treating nephrologist was an owner and whether the treating facility was owned). The coefficient on the last indicator is the difference in differences. Facility and nephrologist fixed effects controlled for unobserved differences in fixed characteristics of nephrologists and facilities in the absence of nephrologist ownership of the treating facility.¹² As covariates, we included observable patient characteristics (age, sex, self-reported race, self-reported ethnicity, number of years on dialysis, and comorbid conditions using diagnosis codes in Medicare claims during the 12 months before the first observed treatment¹³) and zip code characteristics from the American Community Survey¹⁴ (population size, median household income, median rent, and percent of residents with a high school degree) at the first included patient-month (full list in the Table; eTable 1 in the Supplement). Self-reported race and ethnicity are in the USRDS and were included because they are important confounders. The USRDS categories for race were Asian, American Indian or Alaska Native, Black or African American, Native Hawaiian or other Pacific Islander, White, and other. We combined American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, and other into an other category because they made up a small number of patients and to adhere to data use agreement requirements that did not allow us to report cell sizes less than 11. This combined other category was used as a single category for regressions. Ethnicity options were Hispanic or Latino or not Hispanic or Latino. We clustered standard errors at the facility level.¹⁵

To assess for confounding by patient populations of nephrologists and facilities (eg, from potential selective referral of patients by owners to owned facilities), we treated each covariate as the dependent variable of the previously described difference-in-differences model, omitting patient covariates as independent variables. These models estimated difference in differences in patient characteristics between owners and nonowners within owned vs nonowned facilities. Small differences in differences in patient covariates would suggest balance on observed characteristics, thereby supporting the assumption of no selection bias.

We conducted several exploratory analyses. First, to conduct subgroup analyses by LDO ownership and by facility profit status, we included an interaction term between the difference-in-differences variable and subgroup of interest. Second, we assessed the association between nephrologist ownership and estimated glomerular filtration rate (eGFR) at dialysis start. Third, we evaluated whether changes in ESA use associated with nephrologist ownership differed by home dialysis use. Finally, we assessed the use of hospice. We considered these analyses exploratory because of smaller sample sizes (for non-LDO and nonprofit subgroups), simultaneity (for eGFR at the start of dialysis), and endogeneity (for dialysis modality subgroups) and because hospice is an intermediate outcome. We conducted sensitivity analyses in which we assigned up to 3 unassigned months to a nephrologist-facility pair, excluded patients with treatment from multiple nephrologist-facility pairs during the study period, and included months with treatment by a nonassigned nephrologist-facility pair when assessing outcomes.

We used 2-tailed tests with an α of .05, and we adjusted P values using an adaptive correction proposed by Benjamini et al,¹⁶ which preserves the false discovery rate with correlated outcomes. We corrected P values only for primary analyses. To construct our analytical data set, we used SAS statistical software version 9.4 (SAS Institute), and for statistical analyses, we used Stata statistical software version 14.0, MP edition (StataCorp). Data were analyzed from March 2020 through August 2022.

Of 251 651 adults (median [IQR] age, 66 [46-85] years; 112 054 [44.5%] women; 9765 Asian [3.9%], 86 837 Black [34.5%], and 148 617 White [59.1%]; 38 938 Hispanic [15.5%]) in our sample (Figure 1), 24 063 patients (9.6%) were treated by owners at their own facilities, 66 304 patients (26.4%) were treated by owners at other facilities, and 161 284 patients (64.1%) were treated by nonowners (Table; eTable 1 in the Supplement). Patients at physician-owned facilities were less likely to be Black; more likely to be Hispanic, be dual eligible, and live in a zip code with higher median income but a lower high school graduation rate; and had a lower mean number of years on dialysis. There was 1 characteristic (the proportion of the population in the patient zip code below the poverty line) that was significantly different when assessing covariate balance, but the magnitude of the coefficient was low (0.4%) compared with the base (18.0%) (Table, eTable 1 in the Supplement). There were 2 covariates with significant differences when assessing balance with an alternate model: Hispanic ethnicity and dual eligibility for Medicare and Medicaid (eTable 2 in the Supplement).

Regression estimates are presented as absolute, not relative differences. Unadjusted and adjusted analyses were similar (Figure 2).¹⁶ Patients treated by facility owners had a 2.4 percentage point higher probability of home dialysis use than patients of nonowners in owned facilities after controlling for differences in home dialysis use between patients of the same nephrologists in nonowned facilities (adjusted difference in difference, 2.4 percentage points; 95% CI, 1.1 to 3.8 percentage points; adjusted P = .006). Treatment by an owner in an owned facility was associated with a 2.2 percentage point lower probability of ESA use (95% CI, −3.6 to −0.7 percentage points; adjusted P = .02) but not with a difference in anemia or blood transfusions. We did not observe significant difference in differences for missed treatments, transplant wait-listing, mortality, hospitalization or 30-day readmission rates, hemodialysis adequacy, or fistula or long-term catheter rates by ownership (Figure 2B).

Outcomes were generally similar by LDO (Figure 3) and profit status (eFigure in the Supplement), although we observed some heterogeneity. Nephrologist ownership of non-LDO facilities was associated with a decreased proportion of patients with long-term catheters among those treated by owners (−1.6 percentage points; 95% CI, −3.1 to −0.1 percentage points; P for interaction =.007). Additionally, ownership was associated with increased home dialysis use at LDOs (3.6 percentage points; 95% CI, 1.6 to 5.7 percentage points; P for interaction = .05) but not at non-LDOs (Figure 3).

There was no association between ownership and eGFR at the start of dialysis (eTable 3 in the Supplement). We observed similar decreases in ESA use between in-center hemodialysis (−1.6 percentage points; 95% CI, −3.1 to −0.1 percentage points) and home dialysis (−2.8 percentage points; 95% CI, −5.5% to −0.1 percentage points) (P for interaction = .38). In sensitivity analyses, findings were not materially different (eTable 4 in the Supplement).

In this cross-sectional cohort study, nephrologist ownership of dialysis facilities was associated with increased use of home dialysis. Although ownership was associated with decreased ESA use, we found no significant difference in anemia (hemoglobin level <10 g/dL) or blood transfusions. In exploratory analyses, there was no significant heterogeneity in ESA use by type of facility (LDO vs non-LDO or profit status). Ownership among non-LDOs was associated with a decreased proportion of patients with long-term catheters. The increase in home dialysis rates was higher among LDOs. In an exploratory analysis, ownership was not associated with increased eGFR at dialysis start.

Historically, ESA use has received heavy scrutiny from federal regulators.¹⁷ Before CMS implemented the End-Stage Renal Disease Prospective Payment System in 2011,¹⁸ administering ESAs was highly profitable, and their use increased dramatically.¹⁹ After dialysis-related medications were bundled into dialysis reimbursements, ESAs became a significant cost center for dialysis facilities.²⁰ Simultaneously, CMS established the Quality Incentive Program, which penalized facilities for high hemoglobin levels (from ESA overuse) and currently penalizes for blood transfusions.^6,21 Additionally, 2 clinical trials^22,23 demonstrated uncertain benefit and increased cardiovascular complications with ESA overuse.^22,23 Consequently, nephrologists became more circumspect with prescribing ESAs. Our study found that nephrologist owners were less likely to prescribe ESAs. While we could not identify their motivations, owners may want to lower facility costs, improve Quality Incentive Program performance, reduce complications from ESA overuse, or achieve a combination of those outcomes. That ownership was not associated with decreased hemoglobin levels or increased transfusions suggests that lower ESA use was not associated with detrimental outcomes.¹⁹ Although increased home dialysis use could partially explain decreased ESA use, it cannot be the entire explanation because we observed decreased ESA use for in-center and home dialysis populations.

Nephrologist ownership was also associated with increased home dialysis use, which aligns with policy goals to promote home dialysis.²⁴ One explanation may be that nephrologist owners may be better able than corporate owners to build a successful home dialysis program, which requires robust education and an engaged nephrologist. Studies^25,26 have found that more patients prefer to use home dialysis than the number who use it. Moreover, nephrologists particularly motivated to provide more patient-centered care may pursue dialysis facility ownership if it facilitates aligning practice with patient preferences.

A profit motive could also be associated with these outcomes. In 2017, nephrologists’ physician fees were higher for managing in-center hemodialysis than home dialysis.²⁷ However, the End-Stage Renal Disease Prospective Payment System made home dialysis more profitable to dialysis facilities.^9,28 Thus, nephrologist owners sharing facility profits may have an overall higher income when prescribing home dialysis despite lower physician fees for doing so. Finally, nephrologist owners may be less interested in short-term financial gains than corporate owners without nephrologist partners; home dialysis programs tend to have worse short-term but better long-term profit margins than in-center hemodialysis programs that can share capital costs among patients already receiving in-center dialysis.¹¹

This study has potential implications for nephrologists’ ability to act on behalf of their patients on dialysis. Typically, facilities and nephrologists are financially separate, and without nephrologist oversight, corporate owners may have different motives. Unfortunately, we could not identify whether a profit motive, a desire to improve patient-centered care, or both were associated with decreased ESA use and increased home dialysis use. The former would suggest that policy makers should consider introducing incentives for nephrologists rather than solely for the facility (eg, including nephrologists in quality measures that have historically focused on facilities), although physician-level incentives tied to performance measures may be associated with unintended outcomes.^29-32 If the latter, it may be that corporate ownership interferes with nephrologists’ independence and desire to act in patients’ best interests. That ownership was associated with increased home dialysis use primarily in LDO facilities was consistent with this latter explanation because corporate owners (and public stakeholders) may eschew patient-centered care in favor of profits. If so, policy makers may be able to implement additional policy safeguards at the facility level.

Our study is highly relevant to recent questions CMS has raised about physician ownership of dialysis facilities.³ The paucity of research on physician ownership in dialysis may be attributed to the opaque nature of these financial relationships, notwithstanding calls for transparency.^2,33 Data for this analysis were very difficult to obtain, requiring a CMS Freedom of Information Act request and considerable manual cleaning and data-linkage efforts. These data are not provided by dialysis companies or available to departments of health or End-Stage Renal Disease Prospective Payment System networks. Additionally, nephrologist ownership is not routinely disclosed to patients. CMS, commercial payers, patients themselves, and regulatory bodies, such as state departments of health, have a legitimate interest in knowing more about these ownership relationships, especially given that information on other financial conflicts of interest is publicly available. Future research with more comprehensive data may address limitations of our study.

Our study has several limitations. Because the physician ownership database was cross-sectional, we could not draw causal conclusions and were limited to a 1-year study period. Additionally, we could not identify the start date of each ownership agreement or changes during the study period, and our ownership data came from 6 months after our study period. However, we hypothesized that ownership agreements were unlikely to change substantially in 1 year, which may mitigate these concerns. We may have also underidentified nephrologist owners if the principals were not easily identifiable in state business registries. These measurement errors would bias our estimates to the null. Although we mitigated unobserved selection bias by conducting a difference-in-differences analysis and incorporating nephrologist-level and facility-level fixed effects, our results could still be confounded by unobserved factors associated with a patient’s treating nephrologist within a specific facility (eg, owners may self-refer specific patients to their owned facilities or preferentially treat patients using home dialysis). We provide some evidence suggesting that this difference-in-differences analysis was associated with successful mitigation of selection bias. Observable covariates were balanced and internally consistent with outcomes, such as mortality. Moreover, results were not materially different in unadjusted analysis. Additionally, this prevalent patient analysis could not rigorously study earlier dialysis starts or referral patterns at dialysis start; future studies should examine referrals of incident patients commencing dialysis. We required that patients have Medicare, which may limit generalizability, although patients with Medicare constitute more than half of the prevalent dialysis population.¹ Additionally, Medicare claims may not contain complete data on clinical outcomes, which again would bias our estimates to the null.³⁴ Despite these limitations, major strengths of our study include methods to mitigate unobserved selection bias and the consistency of results across multiple robustness checks.

The results of this cross-sectional cohort study suggest that conflicts of interest associated with nephrologist ownership of dialysis facilities were not associated with compromised patient care and were consistent with improved outcomes in 2 domains, home dialysis and ESA use. Still, given the increasing prevalence of nephrologist ownership of dialysis facilities,⁵ policy makers should mandate disclosure of physician ownership as is generally done with other financial conflicts of interest and make such information publicly available for researchers and policy makers.

Accepted for Publication: September 16, 2022.

Published Online: November 7, 2022. doi:10.1001/jamainternmed.2022.5002

Corresponding Author: Eugene Lin, MD, MS, Department of Medicine, Division of Nephrology, Keck School of Medicine of the University of Southern California, 1333 San Pablo St, MMR 622, Los Angeles, CA 90033 (eugeneli@usc.edu).

Author Contributions: Dr Lin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Co-Senior Authors: Drs Berns and Kanter.

Concept and design: Lin, McCoy, Berns, Kanter.

Acquisition, analysis, or interpretation of data: Lin, McCoy, Liu, Lung, Rapista, Kanter.

Drafting of the manuscript: Lin, Liu, Lung, Kanter.

Critical revision of the manuscript for important intellectual content: Lin, McCoy, Rapista, Berns, Kanter.

Statistical analysis: Lin, Liu, Lung, Kanter.

Obtained funding: Lin, McCoy.

Administrative, technical, or material support: Lin, McCoy, Rapista, Berns.

Supervision: Lin, Berns, Kanter.

Conflict of Interest Disclosures: Dr Lin reported receiving personal fees from Acumen outside the submitted work and serving on boards or committees for the American Society of Nephrology and National Kidney Foundation. Dr Berns reported that his division received medical director fees for other faculty members who serve as hemodialysis unit medical directors; Dr Berns does not serve as a medical director and received no compensation. No other disclosures were reported.

Funding/Support: Dr Lin was supported by grant K08 DK118213 from the National Institute of Diabetes and Digestive and Kidney Diseases and by the University Kidney Research Organization. This study was also supported by a pilot grant from the Leonard Davis Institute of Health Economics at the University of Pennsylvania.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Some of the data reported here were supplied by the US Renal Data System (USRDS). The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the US government or USRDS.