The Timing of Specialist Evaluation in Chronic Kidney Disease and Mortality

Context In the United States, the 5-year survival of patients undergoing dialysis is 29%. Early nephrologist evaluation is associated with better outcomes, but 25% of patients first see a nephrologist within a month of beginning dialysis. Contribution Late nephrology evaluation (<4 months before start of dialysis) was most common among black men, uninsured patients, and patients with severe comorbid illness. The later the first evaluation by a nephrologist, the greater the risk for death. Clinical Implications Clinicians need a system to remind them to refer patients at an early stage of chronic renal failure, especially black men, the uninsured, or patients with severe comorbid illness. The Editors In the United States, approximately 300 000 persons have treated end-stage renal disease (ESRD) and an estimated 800 000 persons have a serum creatinine concentration of 177 mol/L (2.0 mg/dL) or greater (1, 2). Annual U.S. spending related to treatment of ESRD exceeds $15 billion (3). Yet, outcomes for patients with ESRD remain relatively poor, with a 5-year survival rate of about 29% for patients undergoing dialysis (4). Whether better care of patients with ESRD earlier in their disease course improves outcomes is under increased investigation (5, 6). Many patients with chronic kidney disease may benefit from beginning their care with primary care physicians. As in the management of other chronic diseases, primary care physicians must decide whether evaluation by a specialist might improve care and, if so, when in the disease course specialist evaluation is most appropriate. Although 39% of patients undergoing hemodialysis and 52% of those undergoing peritoneal dialysis are evaluated by a nephrologist more than 1 year before dialysis, 25% and 16%, respectively, of such patients are seen less than 1 month before dialysis (7). One argument for early evaluation by a nephrologist is that management of chronic renal insufficiency and its complications, such as anemia and renal osteodystrophy, may be improved. Early evaluation might facilitate improved patient education about dialysis; provide more time to make an informed choice about the type of dialysis; and allow timely placement of permanent vascular access, which is associated with better dialysis and fewer complications compared with temporary access (8, 9). Late evaluation is associated with a higher risk for unplanned first dialysis, more complications, higher hospital costs, and longer duration of hospitalization in the first 3 months of dialysis (10-13). Most previous studies of late evaluation were done in countries other than the United States, involved only one center, or have had relatively short follow-up (10, 11, 13-16). We sought to determine the patient factors that are associated with late evaluation by a nephrologist in the United States and the effect of late evaluation on mortality. Methods Study Design and Sample We conducted a national, concurrent, prospective cohort study as part of the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) Study. (For a list of all investigators on the CHOICE Study, see the Appendix.) Between October 1995 and June 1998, 1041 patients undergoing incident dialysis were enrolled at 81 dialysis clinics in 19 states (79 clinics affiliated with Dialysis Clinics Incorporated and 2 clinics affiliated with Beth Israel Medical System) (17). Median time from initiation of dialysis to enrollment was 45 days, and 98% of enrollment took places within 4 months of initial dialysis. Patients were excluded if they were younger than 18 years of age or did not speak English or Spanish. The Johns Hopkins University School of Medicine Institutional Review Board and the review boards of each clinical center approved the CHOICE protocol. The CHOICE Study was designed to examine the choices that patients and providers make in initiation and maintenance of renal replacement therapy, particularly the choice of hemodialysis versus peritoneal dialysis. Data Collection At enrollment, patients completed a baseline questionnaire on medical and social history and provided the month and year in which they first visited a nephrologist. Demographic data, insurance information, the assigned cause of renal failure, baseline laboratory values, and the date of initial dialysis were obtained from the Center for Medicare & Medicaid Services medical evidence form. A trained research nurse abstracted medical records to determine the Index of Disease Severity score for 19 medical conditions. The Index of Physical Impairment score, a measure of impairment in 11 areas, was assessed by clinic staff. The Index of Physical Impairment and Index of Disease Severity were combined to form the Index of Coexistent Disease, a measure of the burden and severity of comorbid disease that is scored from 0 or 1 (mild coexistent disease) to 3 (severe coexistent disease) (18-21). Information from clinic reports and the Center for Medicare and Medicaid Services were used to determine the date of death. The time between first evaluation by a nephrologist and the date of first dialysis is referred to as the time of evaluation and was categorized as late (<4 months), intermediate (4 to 12 months), or early (>1 year). The 4-month cutoff for late evaluation has been used in other studies (13, 22). The 12-month cutoff for early evaluation was chosen because some experts recommend 1 year as the minimal time necessary to prepare a patient adequately for dialysis (23). We reviewed available medical records for patients having late evaluation and a 10% sample of remaining charts. In 13 cases, time of evaluation was adjusted from late to intermediate or early evaluation. Three hundred thirty-four patients did not answer the question on time of evaluation. On the basis of medical record review, 70 of these patients were categorized as having early evaluation and 51 as having intermediate evaluation. For 182 of the 334 patients, the medical records showed no definite evidence to indicate evaluation by a nephrologist more than 4 months before dialysis. These patients, along with 31 patients without available medical records, were categorized as having missing data. The total sample comprises 828 patients. Statistical Analysis Characteristics of the sample stratified by time of evaluation were compared by using chi-square tests and analysis of variance, as appropriate. Distributions of patients according to the time of evaluation were compared by using the Wilcoxon rank-sum test for two-category comparisons and the KruskalWallis test for multiple comparisons. Unadjusted percentages of patients having late evaluation were calculated for different characteristics. Multivariate logistic regression was performed to determine the presence, magnitude, and independence of the association between patient characteristics and late evaluation. We considered the potential effect of multiple centers on our analyses (24). Given the possibility of confounding by clinic, all logistic regression analyses were conditioned on clinic (25). Patient characteristics in the multivariate conditional logistic regression model that were significantly associated (P < 0.10) with late evaluation in univariate analysis were considered potential confounders. Adjusted percentages were calculated on the basis of the adjusted odds ratios derived from logistic regression (26) and the relevant unadjusted frequencies of reference group in each analysis. A HosmerLemeshow test was used to assess model adequacy. We used Cox proportional-hazards regression to test the presence, strength, and independence of the association between time of evaluation and mortality. Survival time was calculated from the date of first dialysis. Patients were considered to be under observation from time of enrollment until death or 30 April 2000. Patients were censored if they received a transplant, changed to a non-CHOICE clinic, or declined to participate further in the study. To account for the possibility that differing standards of care at the various clinics explained differences in survival, all proportional hazards models were stratified by clinic (27). We first included nonmodifiable risk factors in the regression models (demographic characteristics and socioeconomic status) and then added potentially modifiable risk factors (smoking, exercise, comorbid conditions and disease severity, and laboratory values) to examine whether these factors explained associations between time of evaluation and mortality. Factors included in the final Cox proportional-hazards model were significantly associated with mortality (P < 0.10) in univariate analysis or had been shown in the literature to have a clinically important association with mortality. Sex and type of dialysis were included a priori. Some modifiable factors may appear in the causal pathway between late evaluation and mortality. Their addition to the model might explain any observed associations. Thus, hematocrit less than 0.3 and hypoalbuminemia (serum albumin level < 36 g/L) were added to the model as potentially modifiable risk factors because previous studies have shown a relationship between these factors and ESRD mortality (28, 29). The glomerular filtration rate, which was calculated according to the Modification of Diet in Renal Disease equation (30), was included to adjust for renal function. We performed sensitivity analyses to explore the effect of assumptions about missing data on time of evaluation and the effect of alternative categorizations of evaluation time. Statistical analyses were performed by using Stata software, version 6.0 (Stata Corp., College Station, Texas). Role of the Funding Source The project was funded by a grant from the Agency for Healthcare Research and Quality (Dr. Powe, principal investigator), the Robert Wood Johnson Clinical Scholars Program (Dr. Kinchen), and the National Institute of Diabetes and Digestive and Kidney Diseases (Drs. Powe and Klag). N