Summary
Background and objectives There is a trend in the United States to maintenance dialysis initiation at higher levels of estimated GFR. This study aimed to determine whether provider characteristics and pre-ESRD nephrology care and vascular access are independently associated with higher estimated GFR at initiation.
Design, setting, participants, & measurements This retrospective cohort study used US Renal Data System data for patients who initiated dialysis in 2006 (n=83,621) and American Medical Association Physician Master File data for provider characteristics. Patient characteristics and estimated GFR were defined, and providers at dialysis initiation were identified. Earlier dialysis initiation was defined as initiation at estimated GFR>10 ml/min per 1.73 m2. Nephrologist density per 100 ESRD patients was calculated by Health Service Area in 2006. Associations between provider characteristics and estimated GFR were determined using logistic regression and linear regression models, accounting for provider clustering.
Results Of the cohort, 47.8% of patients initiated dialysis at estimated GFR>10 ml/min per 1.73 m2, and 16.2% of patients initiated dialysis at estimated GFR≥15 ml/min per 1.73 m2. Predialysis nephrologist care for 0–12 months was associated with greater odds of earlier initiation compared with no care. Patients initiating with an arteriovenous fistula or graft were more likely to initiate earlier than patients initiating with a catheter. Provider sex was not associated with timing of dialysis initiation as measured by estimated GFR. Care by providers who graduated from nondomestic medical schools was associated with greater odds of earlier initiation. Greater provider experience was associated with lower likelihood of earlier initiation.
Conclusion This study supports the hypothesis that provider factors are associated with timing of dialysis initiation in the United States.
Introduction
The number of patients who undergo renal replacement therapy in the United States continues to increase (1). Patients with ESRD constituted 1.3% of the Medicare population in 2012 but accounted for 7.5% of Medicare spending (1). There is a trend in the United States to initiation of maintenance dialysis at higher levels of estimated GFR (eGFR) (2). In observational studies, higher eGFR at dialysis initiation has been associated with higher mortality risk after initiation, independent of nutritional status (3,4). Recently, the Initiating Dialysis Early and Late trial reported that, compared with a strategy of delayed dialysis initiation, earlier initiation in patients with advanced CKD was not associated with improved survival (5) or quality of life, but was associated with increased costs (6). Despite the evidence of limited benefit of earlier dialysis initiation, the percentage of patients who initiate renal replacement therapy with eGFR above 15 ml/min per 1.73 m2 continues to rise (1). It has been suggested that providers might be encouraging earlier dialysis initiation as a consequence of several factors, including heavy reliance on eGFR values, misinterpretation of clinical practice guidelines, belief in a benefit of earlier initiation, desire to simplify management of CKD complications, and greater financial reimbursement to providers associated with dialysis compared with CKD care (2,4,7). Although a number of patient factors have been associated with earlier dialysis initiation, we are unaware of any studies evaluating provider factors. We aimed to determine whether provider characteristics and pre-ESRD nephrology care and vascular access are independently associated with higher eGFR at dialysis initiation.
Materials and Methods
All patients ages 20 years and older who initiated hemodialysis between January 1, 2006, and December 31, 2006, and had valid information on the Centers for Medicare and Medicaid Services (CMS) Medical Evidence Report (form CMS-2728; n=83,621) were identified in the US Renal Data System incident cohort files (Figure 1).
Cohort selection. eGFR, estimated GFR; SES, socioeconomic status.
The CMS ESRD Medical Evidence Report was the source of patient data for the reported analysis. Providers are required to complete this form for all patients within 45 days of renal replacement therapy initiation irrespective of the patient’s insurance coverage. eGFR was calculated from serum creatinine using the Modification of Diet in Renal Disease (MDRD) Study formula (8); it was expressed as categories>10 (earlier initiation) and ≤10 ml/min per 1.73 m2 as a continuous variable. Patient factors considered for the final model included age, sex, race, body mass index, smoking history, alcohol use, drug dependence, cause of ESRD, functional status (inability to ambulate, inability to transfer, need for assistance with activities of daily living, institutionalization, and employment at the time of ESRD onset), comorbid conditions (atherosclerotic heart disease, congestive heart failure, history of cerebrovascular accident, peripheral vascular disease, hypertension, diabetes mellitus, chronic obstructive pulmonary disease, and cancer), predialysis nephrology care, hemoglobin, albumin, and dialysis access at the time of dialysis initiation, and they were obtained from the Medical Evidence Report. Hemoglobin, albumin, and creatinine values were obtained within 45 days before the first dialysis treatment. A socioeconomic status score was determined for each patient based on home zip code using 2000 census data (9). We used dual eligibility for Medicaid and Medicare as another surrogate for low socioeconomic status. Rural and urban definitions were adopted from the US Census Bureau based on zip code (10).
The Universal Provider Identification Number on the Medical Evidence Report was used to identify the dialysis provider at initiation. Universal Provider Identification Numbers were then merged with American Medical Association Physician Master File data to obtain provider characteristics, including age, sex, number of years since the end of training, and graduation from a domestic or foreign medical school. The number of years since the end of training variable is presented as tertiles. The number of renal providers of ESRD care in each health service area was determined from dialysis claims submitted during 2006. Provider and patient Health Service Areas (HSAs) (11) were determined based on billing and residence zip codes, respectively. The density of renal providers per 100 ESRD patients was calculated and analyzed in quartiles, and it is presented by patient HSA.
Baseline characteristics are presented by strata of eGFR at dialysis initiation. We determined associations between provider characteristics and eGFR at dialysis initiation (eGFR>10 ml/min per 1.73 m2 was considered earlier initiation) adjusted for patient characteristics using logistic regression as the primary analysis. General linear model analyses with continuous eGFR and adjustment for patient and provider characteristics were also performed as a secondary analysis. We performed a subgroup analysis limited to patients with 6 months or more of predialysis nephrology care (n=38,070). Because the results were similar, only results for the total cohort are presented. We determined associations between patient characteristics and provider characteristics (foreign medical graduate [yes or no]; years since end of training [0–8 versus ≥9]), adjusted for patient and provider characteristics using logistic regression analysis. In addition, we performed analyses in which we adjusted the model for patient insurance coverage at initiation; this adjustment did not change the results. We also performed analyses using the clustered SEM method to adjust for clustering by providers, but it did not change the results. All analyses were performed using 9.1 SAS software.
Results
The cohort comprised 83,621 patients who initiated dialysis in 2006; 51.4% of patients were ages 65 years and older at dialysis initiation, 45% of patients were women, 29.3% of patients were African American, and 45.8% of patients had ESRD caused by diabetes. eGFR values at dialysis initiation were distributed as follows: 20,115 (24.1%), eGFR<7 ml/min per 1.73 m2; 23,521 (28.1%), eGFR=7 to <10 ml/min per 1.73 m2; 26,439 (31.6%), eGFR=10 to <15 ml/min per 1.73 m2; 13,546 (16.2%), eGFR≥15 ml/min per 1.73 m2. Overall, 39,985 patients (47.8%) initiated dialysis with eGFR>10 ml/min per 1.73 m2, which is considered earlier initiation in this analysis.
Patients who initiated dialysis earlier were more likely to be elderly (58.5% ages≥65 years) than patients who initiated later (44.9% ages≥65 years). Patients who initiated earlier were less likely to be women (41.4% versus 48.3%), more likely to be white (68.4% versus 63.3%), and more likely to have diabetes as the cause of ESRD (51.3% versus 40.7%); prevalence of comorbid conditions, except for cancer, was higher, and functional impairments were more likely (Table 1).
Patient and provider characteristics at dialysis initiation
After multivariable adjustment, nephrologist care in the predialysis period was associated with greater odds of earlier dialysis initiation (adjusted odds ratio [AOR], 1.17; 95% confidence interval [95% CI], 1.13 to 1.22 for 0–12 months of predialysis nephrologist care and AOR, 1.05; 95% CI, 1.01 to 1.10 for more than 12 months of predialysis nephrologist care) compared with no nephrologist care. Compared with patients who initiated dialysis with a catheter, patients who initiated with a functional arteriovenous fistula (AOR, 1.07; 95% CI, 1.02 to 1.12) or arteriovenous graft (AOR, 1.18; 95% CI, 1.10 to 1.26) were more likely to initiate earlier (Table 2). Similar results were observed in the linear regression analysis.
Associations of provider and predialysis care characteristics with estimated GFR at dialysis initiation
After multivariable adjustment, several provider characteristics were independently associated with earlier dialysis initiation. Odds of earlier initiation were greater for patients of less experienced providers. Compared with odds for patients of providers in practice for 0–8 years since the end of training, odds of earlier initiation were 8% lower for patients of providers in practice for 9–21 years (AOR, 0.92; 95% CI, 0.89 to 0.95) and 14% lower for patients of providers in practice for ≥22 years (AOR, 0.86; 95% CI, 0.83 to 0.90). Patients whose providers were foreign medical graduates were more likely to have had dialysis initiated earlier (AOR, 1.16; 95% CI, 1.12 to 1.19). Patients living in HSAs with lower or higher nephrologist density were less likely to initiate earlier than patients living in HSAs with 1–1.5 nephrologists per 100 ESRD patients (Table 2). Similar findings were observed in the linear regression analysis (Table 2).
Several patient characteristics were independently associated with earlier dialysis initiation (Table 3).
Associations of patient characteristics with estimated GFR at dialysis initiation
Several patient characteristics independently associated with greater odds of being cared for by a foreign medical graduate were also associated with higher eGFR at dialysis initiation: older age, diabetes as a cause of kidney disease, history of congestive heart failure, inability to ambulate, eligibility for Medicaid, and initiation of dialysis through an arteriovenous graft (Supplemental Table 1). Fewer patient characteristics are independently associated with the number of years since the physician finished training (Supplemental Table 1).
Discussion
We found that predialysis care was associated with eGFR at dialysis initiation. Patients followed by a nephrologist before dialysis initiation were more likely to initiate dialysis with higher eGFRs than patients without nephrology follow-up, and patients with arteriovenous fistulas or grafts were more likely to initiate with higher eGFRs than patients with catheters, independent of sociodemographic characteristics, comorbid conditions, or nutritional markers. In addition, certain provider characteristics, such as fewer years of experience and graduation from a foreign medical school, were associated with earlier dialysis initiation in their patients. These findings support the hypothesis that, in addition to patient characteristics, provider factors contribute to earlier dialysis initiation (2,7).
Patients who received predialysis nephrology care were more likely to initiate dialysis earlier, and presence of a functional graft or fistula was associated with earlier initiation. Several potential explanations account for these findings. Possibly, patients who receive no pre-ESRD nephrology care are more likely to present late with low eGFR and initiate dialysis using a catheter. Our finding that odds of initiating dialysis at higher eGFRs are lower for patients with a history of longer (>12 months) follow-up by a nephrologist than for patients with shorter (<12 months) follow-up support this hypothesis. Alternatively, providers might initiate their patients earlier after access, particularly an arteriovenous graft, is in place. Although lack of adequate preparation for dialysis and late referral were cited as common factors that delayed a planned start of renal replacement therapy in Europe (12), patients with functioning arteriovenous access with minimal indications may be preferentially initiated on dialysis compared with patients who are unprepared for dialysis. Guidelines for vascular access placement (13) might have created unintended consequences; after the access is ready for use, providers may have lower thresholds for initiating dialysis. Many retrospective cohort studies have suggested a survival benefit during the first 1 year of dialysis for patients who receive predialysis nephrology care and patients who initiate dialysis with an arteriovenous fistula or graft (13–22). Analyses that evaluate the influence of predialysis nephrology care and dialysis access at initiation on outcomes may suffer from lead time bias; nephrologists may accelerate dialysis start dates for well-prepared patients under their care, and survival analyses that begin with dialysis initiation may show better survival in well-prepared patients.
We did not find an association between provider sex and timing of dialysis initiation as determined by eGFR. This finding contrasts with the findings of a European study that found that women nephrologists aimed to start dialysis at higher eGFRs (12). Although 35% of the nephrologists who responded to the questionnaires in the European study were women, only 15% of the patients in our cohort were cared for by women nephrologists.
Less experienced providers were more likely to initiate their patients on dialysis earlier. Several potential explanations account for this finding. The change in practice patterns during the last decade coincided with release of clinical practice guidelines that recommend considering dialysis initiation in patients with eGFR<15 ml/min per 1.73 m2 and symptoms of uremia and some patients with eGFR>15 ml/min per 1.73 m2 who are symptomatic from kidney failure (23). In addition, CKD staging that defined stage 5 CKD and ESRD as eGFR<15 ml/min per 1.73 m2, which was commonly used during the past decade, might have been misinterpreted as an indication to initiate dialysis after stage 5 CKD occurred (24). Providers with less experience were trained during that time and not surprisingly, initiated dialysis in their patients earlier. In addition, clinical experience and comfort making decisions based on patient symptoms rather than numbers might have contributed to more experienced providers’ decisions to initiate their patients later.
Why providers who graduated from foreign medical schools initiate their patients on dialysis earlier than providers who graduated from domestic medical schools is unclear. Evidence that patient outcomes and quality of care differ by whether providers graduated from domestic or foreign medical schools is inconsistent in other areas of medicine (25–27). For domestically and internationally educated physicians, learning to care for CKD and ESRD patients occurs during a nephrology fellowship in the United States. One can speculate that foreign medical graduates might be more likely to practice in underserved areas with a higher prevalence of sicker patients (28). In adjusted analysis, older patients, patients with diabetes or congestive heart failure, bedbound patients, and patients receiving Medicaid were more likely to be cared for by foreign medical graduates. Although we were able to adjust the analysis for many measured patient characteristics, unobserved confounding might, in part, explain the results. We found a weak association between density of nephrologists per 100 ESRD patients and eGFR at dialysis initiation. Patients living in HSAs with lower or higher nephrologist density were less likely than patients living in HSAs with 1–1.5 nephrologists per 100 ESRD patients to initiate dialysis earlier. Possibly, HSAs differ in their capacity to accommodate patients with ESRD, competition for patients, incentive to initiate patients at lower eGFRs, and health care intensity (29). In a study of factors that influence the decision to start dialysis among European nephrologists, presence of a waiting list in a dialysis facility was cited as a reason to delay dialysis initiation (12). Also, nephrologists from countries with high incidence of treated ESRD and nephrologists who practiced in for-profit dialysis facilities chose higher eGFR for dialysis initiation and were more likely to initiate their patients early in the presence of advanced age and chronic conditions (12). Similarly, in the United States, incidence of treated ESRD, particularly among elderly patients and patients with multiple comorbid conditions, is highest in regions with the highest intensity of end of life care (29). Associations of the geographic variation in many medical practices, including eGFR at dialysis initiation, remain to be determined.
Unfortunately, although earlier dialysis initiation contributes to increasing costs of care for ESRD patients (2,6,30), it does not improve either survival or quality of life (3–6). Timely initiation of dialysis is likely to improve patient wellbeing by decreasing the proportion of life spent undergoing dialysis and provide financial savings to Medicare. As the body of evidence related to earlier dialysis initiation and outcomes develops, greater emphasis should be placed on disseminating these findings to practicing nephrologists.
To our knowledge, this study is the first study to evaluate the association between provider and pre-ESRD care characteristics and timing of dialysis initiation as determined by eGFR. The study is large and generalizable to the overall dialysis population. However, it is limited in several ways. As an observational study, this study suffers from residual bias and cannot support claims of causality. Although we adjusted the analysis for many patient characteristics available to us, residual bias from unobserved confounders might remain. In addition, eGFR at dialysis initiation was estimated using the MDRD Study formula from serum creatinine measured in local laboratories within 45 days before dialysis initiation. Compared with measured GFR, eGFR might not be as good a measure of kidney function, especially in older or malnourished patients. We were unable to ascertain whether dialysis was initiated in the hospital or the community, and also, we could not determine the uremic symptoms that led to the decision to initiate dialysis. We used information from the CMS Medical Evidence Report to determine timing of predialysis nephrology care. There is significant disagreement between timing of predialysis nephrology care as determined from the Medical Evidence Report and Medicare physician claims (31); however, we dichotomized the timing of first predialysis nephrology care as >12 or ≤12 months, a definition with accuracy of 70% (31). Physicians were identified based on the Medical Evidence Report at the time of dialysis initiation; whether the same physician cared for the patient before or after dialysis initiation is unknown.
In conclusion, predialysis nephrology follow-up, particularly follow-up of less than 12 months, functional arteriovenous fistula or graft, and provider characteristics, such as fewer years in practice and graduation from a foreign medical school, were independently associated with earlier dialysis initiation. This study supports the hypothesis that provider factors are associated with earlier dialysis initiation in the United States. Because current evidence does not support earlier dialysis initiation to improve patient outcomes, providers should be educated about this lack of benefit. This education should focus on recent graduates and foreign medical graduates. In addition, the societies that produce guidelines regarding timing of nephrology follow-up and vascular access placement should consider possible unintended consequences of recommendations for early interventions.
Disclosures
None.
Acknowledgments
The authors thank Delaney Berrini for manuscript preparation and Nan Booth, MSW, ELS, for manuscript editing.
This study was supported by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health Grant R01DK082415.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
This article contains supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.04190413/-/DCSupplemental.
See related editorial, “Early Dialysis Initiation, a Look from the Rearview Mirror to What’s Ahead,” on pages 222–224.
- Received April 22, 2013.
- Accepted October 16, 2013.
- Copyright © 2014 by the American Society of Nephrology