In this study of consecutive patients initiated on dialysis at two San Francisco outpatient dialysis units, we found there was a period of rapid decline in kidney function prior to the need for permanent dialysis for 7.6% of patients at one unit and 12.7% of patients at another. All patients who started chronic dialysis precipitated by rapid decline in kidney function started hemodialysis and all relied on catheters as access. For these patients, the "primary etiology" of ESRD recorded on the ESRD Medical Evidence (CMS 2728) Form, such as diabetes mellitus, sometimes did not capture the critical role of this acute episode in precipitating the need for ESRD therapy. Furthermore, the creatinine and hemoglobin values documented in the Medical Evidence form did not fully capture the clinical status of these patients during their "pre-ESRD" period.
Few prior studies have taken our study approach and examined relatively abrupt transitions to ESRD at the individual-patient level [4, 5]. Although our results are based on a relatively small number of patients from two hospital-affiliated dialysis units, we believe that our results have face validity and are consistent with the clinical experience of practicing nephrologists. Our results are also consistent with that reported by Bhandari and Turney, who found that from 1984 to 1995, survivors of "acute renal failure" who did not regain renal function comprised 18.4% of all new patients taken into their long-term dialysis program in England . They are also consistent with the high rates of non-recovery of renal function following dialysis requiring acute kidney injury which have been reported among patients ware critically ill  or who have pre-existing chronic kidney disease .
We believe that these findings should stimulate thinking along several lines. First, periods of rapid loss of renal function appear to be a prominent feature of the natural history of kidney disease in a sizable minority of patients who develop ESRD. For example, Patient 1 at CPMC had a decline in estimated GFR from 69 ml/min/1.73 m2 to ESRD within a two-month period due to cholesterol emboli. More research is needed to investigate whether other CKD patients (e.g., those who did not develop ESRD in the short run) also experience these periods of rapid decline in renal function as part of the natural history of their disease.
Second, patients who start chronic dialysis as a result of rapid loss of kidney function over a short period of time will naturally have little, if any, opportunity to be educated about dialysis modality choices or undergo adequate preparation for dialysis such as pre-emptive creation of an arteriovenous fistula. Hence, it is not surprising that we observed that 100% of all ESRD cases preceded by rapid decline in kidney function were initiated on hemodialysis via a catheter. These considerations suggest that national policy targets such as that set by "Fistula First" in the USA must take into account the proportion of new dialysis patients whose ESRD was precipitated by rapid decline in kidney function. For example, one of the original clinical performance measures in the Center for Medicare & Medicaid Services ESRD US Clinical Performance Measures project was: "A primary arterial venous fistula should be the access for at least 50% of all new patients initiating hemodialysis ." We would suggest that the proportion of patients who started dialysis immediately precipitated by these rapid declines in kidney function should be removed from the numerator (and denominator) of such targets. In other words, the fact that some patients suffer rapid loss of kidney function and transition abruptly from mild-to-moderate CKD to dialysis-dependency must be taken into consideration when establishing appropriate clinical performance measures for incident dialysis patients. These patients will also likely not be evaluated for pre-emptive kidney transplantation.
Third, results of this study should inform our interpretation of the data in the US Renal Data System registry. The "primary etiology" of ESRD listed on the ESRD Medical Evidence (CMS 2728) Form may not convey the critical role of these episodes of rapid decline in kidney function in precipitating the need for ESRD therapy--e.g., in the example of Patient 3 at CPMC when there was severe acute tubular necrosis superimposed upon CKD due to diabetes mellitus. The traditional approach of ascribing only one single cause for each ESRD case in the American and other national registries does not capture the complexity of clinical reality. In addition, because of the existing data collection process, current documentation in the US Renal Data System may not accurately reflect the "pre-dialysis" health status in a number of patients. The hemoglobin and albumin at the start of dialysis as documented in the Medical Evidence form has been used in prior studies to assess quality of CKD care in the USA and to infer decisions that are made by patients and physicians regarding timing of initiation of dialysis [9, 10]. These analyses assume that all dialysis patients had a slowly progressive CKD course and that there were opportunities available to treat the low hemoglobin level or to initiate dialysis before hypoalbuminemia developed. This assumption naturally does not hold in the cases where there are superimposed episodes of rapid decline in kidney function.
The strengths of this study include its emphasis on a hitherto underappreciated aspect of the natural history of transition from CKD to ESRD. We studied this issue in two distinct health care settings which serve different patient populations and this strengthened the external validity of our conclusions. The clinical and public health importance of highlighting this "cause" of ESRD--affecting perhaps 5-10% of incident ESRD patients--should be put into context of other ESRD etiologies. By comparison, according to US Renal Data System, 3% of ESRD are due to "cystic/heriditary/congential diseases" and 2% to "secondary glomerulonephritis" (such as lupus nephritis) .
Limitations of this study relate to its small sample size which is partly due to the labor-intensive process of conducting chart review. However, the level of granularity achieved (down to exact day of initiation of acute dialysis which then transitioned into maintenance dialysis) may not be reliably replicated in administrative databases in which the exact start of date of dialysis may be documented with a margin of error of up to a few months. The small sample size limited our ability to conduct meaningful analyses regarding predictors of rapid decline in kidney function or whether the post-ESRD outcomes differed among patients who did and did not suffer rapid decline in kidney function. Our results are based on a two hospital affiliated dialysis units in Northern California and so may not be generalizable to other setting. The mean age and proportion of male patients in our study sample were higher than the general U.S. incident ESRD population. However, they have face validity and are consistent with the clinical experience of practicing nephrologists. Owing to the limited sample sizes, numerous comparisons (Tables 1 and 2) did not achieve statistical significance, although it seems unlikely that differences observed were merely chance findings (e.g., the pre-test probability is high that patients who develop acute kidney injury with non-recovery of renal function would not have mature dialysis fistulas in place). We were not able to access information documented on the Medical Evidence (CMS 2728) Form for patients at one of the two dialysis units. We had limited information about the setting in which the rapid declines in kidney function occurred and about the care provided to these patients (e.g. timing of nephrology involvement). Our definition of "rapid decline in kidney function" was arbitrary. We did not use consensus definitions for acute kidney injury that were developed more for hospitalized patients (especially critically ill inpatients)  as we wanted to capture a broader spectrum of rapid decline in renal function. But we believe that our estimation of the magnitude of the problem is conservative since we required documentation within three months of ESRD of a relatively high prior estimated GFR value of 30 ml/min/1.73 m2. Thus, a patient who has had stable GFR at 25 ml/min/1.73 m2 for several years before developing ESRD abruptly as a result of severe acute tubular necrosis during a hospitalization would not have been counted; neither would a patient who presented with rapidly progressive glomerulonephritis without known prior serum creatinine measurements in the hospital systems.
In summary, it appears that numerous cases of ESRD are preceded by rapid declines in kidney function. The importance of these periods of rapid decline in kidney function has been under-appreciated in prior studies of the natural history of CKD and ESRD.