Many dialysis facilities have financial relationships with nephrologists, including joint venture agreements, where the nephrologist owns a minority share of the dialysis facility. Such agreements could present a conflict of interest with respect to patient care. This study will investigate whether these joint venture agreements are associated with differences in the quality of care provided by dialysis facilities.
The investigators will use the United States Renal Data System (USRDS), a registry of all patients on dialysis in the US irrespective of payer. The dataset includes patient characteristics (biologic and sociodemographic) within 45 days of initiating dialysis for all patients with ESKD irrespective of insurance coverage, death data for all patients with ESKD irrespective of insurance, dialysis facility characteristics, which are updated annually, longitudinal treatment data submitted by the end-stage renal disease (ESRD) networks for all patients irrespective of insurance, and CROWNWeb clinical data: monthly treatment data (e.g., Kt/V, treatment time, serum hemoglobin, vascular access) submitted by dialysis facilities for all patients irrespective of insurance. The registry is linked to Fee-for-service Medicare claims for all patients with this payer. All data have already been collected (i.e., this is a retrospective study) and deidentified by a data distributor. The investigators have also obtained a cross-sectional dataset of dialysis facilities and physician owners in 2017 from the Centers for Medicare \& Medicaid Services (CMS) through a Freedom of Information Act request. In this study, the investigators will link the USRDS to this cross-sectional dataset. The investigators will also link the data to publicly available data from Dialysis Facility Compare (which contains quality performance of each dialysis facility, published by the government on a quarterly basis) and Census data (which contains geographic sociodemographic characteristics). From this data linkage, the investigators will study differences between facilities that are physician owned and those that are not physician owned. The investigators will study outcomes both at the facility level and the patient level. All models will have an alpha of 5% and will have 2-sided statistical tests. For facility level outcomes: The investigators will construct a facility level dataset and compare physician-owned facilities to non-physician owned facilities adjusting for facility characteristics and regional (zipcode level) sociodemographic characteristics. The investigators will also test the effect of incorporating patient characteristics into the model. For patient characteristics, the investigators will take the average for each facility's population (e.g., average age, % of patients male, etc.). The investigators will use ordinary least squares for continuous outcomes and logistic regression for binary outcomes. The investigators will use robust standard errors. For patient level outcomes: The investigators will construct a patient-month panel dataset and compare patients dialyzing at physician-owned facilities to those dialyzing at non-physician owned facilities. The investigators will adjust for patient, facility, and zipcode level sociodemographic characteristics. Since all outcomes are binary, the investigators will use logistic regression for all models. The investigators' primary analysis will be logistic regression, adjusting for patient, facility, zipcode characteristics, with patient-level fixed effects and non-parametric bootstrap standard errors. In order, The investigators will explore the sensitivity of results to the following: * logistic regression with all adjusters, patient-level fixed effects, and robust standard errors * ordinary least squares with all adjusters, patient-level fixed effects, cluster-robust standard errors at the facility level * ordinary least squares with all adjusters, patient-level fixed effects, robust standard errors The investigators also pre-specify the adjusters below: Patient characteristics (comorbidities will be obtained using the Chronic Conditions Warehouse software on a 12 month lookback of Medicare fee-for-service claims) * Age * Sex * Race * Ethnicity * Prior transplant * Incident patient (first 120 days of dialysis) * Years with ESRD * Dual Eligibility * Hypertension * Alzheimers * Atrial fibrilation * Prior myocardial infarction * Asthma * Breast Cancer * Cataract * Chronic obstructive pulmonary disease * Colorectal Cancer * Depression * Diabetes * Endometrial Cancer * Glaucoma * Congestive Heart Failure * Hip Fracture * Hyperlipidemia * Hypertension * Ischemic heart disease * Lung cancer * Osteoporosis * Prostate Cancer * Rheumatoid Arthritis / Osteoarthritis * Prior stroke / transient ischemic attack * Benign prostatic hyperplasia Facility characteristics * For-profit status * Chain owned * Number of patients at facility * Patient:staff ratio * ESRD Network Regional (zipcode level) sociodemographic characteristics * Median Income * % of zipcode with high school degree * % of zipcode below poverty line We pre-specify a subgroup analysis by whether the dialysis facility is owned by a large dialysis organization (i.e., Davita, Fresenius).
Study Type
OBSERVATIONAL
Enrollment
208,213
The "intervention" or exposure is whether the dialysis facility was owned by a physician in 2017
The "intervention" or exposure is whether the dialysis facility was owned by a large dialysis organization in 2017
Facility level outcome: Facility 5-star rating
Whether a facility was a "high-quality" facility using the Dialysis Facility Compare aggregate score in 2017. The variable will be dichotomized into: * High-quality: 4-5 stars * Low-quality: 1-3 stars
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with long-term catheter
The fraction of patient-months of a dialysis facility where the patient used a catheter for 3 months or longer in 2017
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with hemoglobin < 10 g/dL
The fraction of patient-months of a dialysis facility where the patient had a hemoglobin \< 10
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with hemoglobin > 12 g/dL
The fraction of patient-months of a dialysis facility where the patient had a hemoglobin \> 12
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with serum calcium > 10.2 mg/dL
The fraction of patient-months of a dialysis facility where the patient had a serum calcium \> 10.2
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with serum phosphate > 7 mg/dL
The fraction of patient-months of a dialysis facility where the patient had a serum phosphate \> 7
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with Kt/V >= 1.2 (for hemodialysis)
The fraction of patient-months of a dialysis facility where the patient had a Kt/V \>= 1.2 (hemodialysis only)
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient-months with Kt/V >= 1.7 (for peritoneal dialysis)
The fraction of patient-months of a dialysis facility where the patient had a Kt/V \>= 1.7 (peritoneal dialysis only)
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: standardized mortality rate
Facility's standardized mortality rate according to Dialysis Facility Compare
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: standardized hospitalization rate
Facility's standardized hospitalization rate according to Dialysis Facility Compare
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: standardized 30-day readmission rate
Facility's standardized 30-day readmission rate according to Dialysis Facility Compare
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: standardized infection ratio
Facility's standardized infection ratio according to Dialysis Facility Compare
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: standardized transfusion rate
Facility's standardized transfusion rate according to Dialysis Facility Compare
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Facility level outcome: % of patient months where fistula is being used (hemodialysis)
The fraction of patient-months of a dialysis facility where the patient was using a fistula (hemodialysis only)
Time frame: Cross-sectional from October 2017 dataset (aggregates quality performance from the calendar year 2016)
Patient level outcome: mortality
Assess the exact date of death (reported in the administrative dataset)
Time frame: For each calendar month in 2017, determine whether the patient was alive or dead (administrative data indicate the exact date of death)
Patient level outcome: home dialysis (peritoneal dialysis or home hemodialysis)
Using claims data (granular at a daily level), determine the dialysis modality used for the plurality of that month
Time frame: For each calendar month in 2017, determine whether the patient used home dialysis for the plurality of that month (claims indicate type of dialysis for each calendar day of the month)
Patient level outcome: serum hemoglobin < 10 g/dL
Whether the patient's serum hemoglobin \< 10 at the end of a given month, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient's serum hemoglobin was < 10 at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: serum hemoglobin > 12 g/dL
Whether the patient's serum hemoglobin was \> 12 at the end of a given month, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient's serum hemoglobin was > 12 at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: serum calcium > 10.2 mg/dL
Whether the patient's serum calcium was \> 10.2 at the end of a given month, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient's serum calcium > 10.2 at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)h and year
Patient level outcome: serum phosphate > 7 mg/dL
Whether the patient's serum phopshate was \> 7 at the end of a given month, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient's serum phosphate was > 7 at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: hospitalization in that month
Whether the patient had a hospitalization in a given month, using claims data (claims are granular at a daily level)
Time frame: For each calendar month in 2017, determine whether the patient experienced a hospitalization (claims data provide a daily assessment of whether patient was hospitalized)
Patient level outcome: 30-day readmission following a discharge in that month
Whether the patient had a discharge in a given month that was followed by a 30-day readmission, using claims data (claims are granular at a daily level)
Time frame: For each calendar month in 2017, determine whether the patient experienced a hospitalization that was later followed by a 30-day readmission (claims data provide a daily assessment of whether patient was hospitalized)
Patient level outcome: unplanned 30-day readmission following a discharge in that month
Whether the patient had a discharge in a given month that was followed by an unplanned 30-day readmission, using claims data (claims are granular at a daily level)
Time frame: For each calendar month in 2017, determine whether the patient experienced a hospitalization that was later followed by an unplanned 30-day readmission (claims data provide a daily assessment of whether patient was hospitalized)
Patient level outcome: using a fistula in that month (hemodialysis)
Whether a fistula was used at the end of a given month, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient was using a fistula at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: using a catheter in that month (hemodialysis)
Whether a dialysis catheter was used at the end of a given month, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient was using a dialysis catheter at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: using a catheter for 3 or more months
Whether a dialysis catheter was used at the end of a given month and at the end of the previous 2 calendar months, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient was using a fistula at the end of the calendar month, and at the end of the previous two months (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: using an erythropoietin stimulating agent
The type of erythropoietin stimulating agent in a given month and the total dose administered, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine whether the patient was using an erythropoietin stimulating agent at the end of the calendar month (dialysis facilities report to CROWNWeb at the end of each month)
Patient level outcome: receipt of a blood transfusion
Whether the patient received a blood transfusion in a given month, using claims data (granular at a daily level)
Time frame: For each calendar month in 2017, determine whether the patient received a blood transfusion in the calendar month (claims data report on a daily basis when a patient received a blood transfusion)
Patient level outcome: dose of erythropoietin stimulating agent
The type of erythropoietin stimulating agent in a given month and the total dose administered, using CROWNWeb data (reported at the end of each calendar month)
Time frame: For each calendar month in 2017, determine the type of erythropoietin stimulating agent used for that month and the total dose (dialysis facilities report to CROWNWeb at the end of each month)