Early Outcomes among Those Initiating Chronic Dialysis in the United States

Discussion

We followed more than 300,000 incident American ESRD patients receiving more than 35 million chronic dialysis treatments during their first year of renal replacement therapy. We found a more than twofold increase in the risk for death and hospitalization in the first 90 days of dialytic therapy, where the most substantial risk occurred in the first 2 weeks. Overall, these first weeks of chronic renal replacement therapy warrant increased attention, because dialysis patients appear most vulnerable during this time period.

Disease registries have significantly contributed to understanding the natural history, prevalence, and determinants of medical disease (22). In the dialysis population, the USRDS (2), for example, follows Medicare patients on chronic renal replacement therapy and provides the nephrology community with a series of vital statistics that describe the state of the ESRD population as a whole; however, most USRDS analyses exclude the first 90 days when registration and Medicare data collection does not occur in the full U.S. dialysis population (2). For example, patients younger than 65 and not medically disabled must wait at least 90 days after their first ESRD service date before they become Medicare eligible. The few studies of incident dialysis patients in these initial 90 days have reported variable early mortality rates between 24 and 50 deaths per 100 patient-years (12–14,19,20,23). Such inconsistencies between studies may occur because incident patients may be unintentionally censored prematurely from disease registries, which may obscure important outcomes that occur in the first 3 months of dialysis. For example, The Centers for Medicare & Medicaid Services, the USRDS, and renal research communities rely on the ESRD Medical Evidence Form (2728) to identify new chronic dialysis patients. This form can be submitted by the provider up to 45 days from ESRD initiation (24). Indeed, many incident dialysis patients may die within these 45 days before the Medical Evidence Form can be even completed.

In this study, we overcame such limitations in patient registration and premature censorship through integrated clinical and billing information systems that mandated charting with each treatment session in all facilities of a large dialysis provider. Consequently, we were able to identify all 303,289 patients initiating chronic renal replacement therapy and followed these patients through every single treatment (three times per week on average). We were able to partition and analyze our data by each week, to uncover large variations in death and hospitalization risks during the first year, where the highest risks occurred in the first 14 days of dialysis. These high risks for death became attenuated by the 90th day of dialysis. For example, the weekly risk for death peaked on the second week and then decreased by more than 50% once a patient reached their 90th day on dialysis. In contrast, a previous study on a subset of Medicare patients reported that the largest mortality occurred later, in the second month of treatment (25). Another study reported only a 26% increased risk of death in the first 120 days when compared with the period after those 120 days (14). Again, such discrepancies between these studies and our study may be partially explained by the delayed identification and recruitment of study subjects (survivorship bias), which has been avoided in this study.

We also conducted risk factor analyses that revealed which patients were prone to early death. In the initial 14 days, those with a catheter vascular access (versus fistula), a preventable risk factor, demonstrated a twofold mortality increase, even after accounting for common comorbidities and other potential confounders linked to mortality. In contrast, a patient with an arteriovenous fistula or on peritoneal dialysis had improved early (14 and 90 days) dialysis survival.

Despite these high morbidity and mortality risks in the first few weeks of dialysis, we observed reductions in 90- and 365-day mortality from 1997 to 2009, which may be attributed to improvements in process-related measures of quality and better disease management (10,13,26–28). Similarly, increasing pre-ESRD programs (29,30) that provide education, emotional support, improved nutrition, and reduce catheters in patients before they start dialysis have the most potential to improve 2-week mortality rates going forward.

Our study cohort was representative of the U.S. dialysis population and spanned a considerable spectrum of socioeconomic, practice patterns, and patient characteristics. When compared with U.S. Census statistics, we found distinct differences between the typical patient starting dialysis and the American population at large (21). A new chronic renal replacement therapy patient was approximately four times more likely to be ≥65 years old, twice as likely to be African American, and three times more likely to be unemployed. The patients lived in communities where only half of residents were high school graduates and the annual household income was lower than the national average. Only 14% of patients started dialysis with an arteriovenous fistula, which is the recommended form of access by the National Kidney Foundation's Kidney Disease Outcomes Quality Initiative (31). Despite these disparities, we found that such socioeconomic factors were only minimally associated with patient mortality, whereas modifiable factors such as vascular access were more strongly associated with good outcomes.

Our study has limitations. We captured and followed all patients from the first day of outpatient dialysis, but did not have records of antecedent hospital and clinic encounters during chronic kidney disease. The risk of early death may be higher if the study entry point was redefined from the point of first ever dialysis or by GFR. Such a definition would include subjects with very high mortality such as acute renal failure patients who initiate dialysis in a hospital and ESRD patients who refuse dialysis. Indeed, residual confounding may explain the statistically significant relationship between risk factor and mortality. Finally, it is possible that the high incident mortality may reflect palliative patients who do not tolerate the rigors of dialysis and choose to withdraw from renal replacement therapy; however, we did not have the means to identify such patients in the dataset.

In conclusion, mortality in the initial 90 days of chronic dialysis is high, yet it is declining in the setting of continuous process improvements and medical risk factor management (10,13,26–28). Nevertheless, additional efforts are needed for this critical period, particularly in the first 2 weeks, for further gains in incident patient outcomes. Such efforts may include more emphasis on medically preparing CKD patients for impending dialysis.