Chronic on Acute Renal Failure: Title and subTitle BreakLong-term Implications of Severe Acute Kidney Injury

Every year more than 1 million hospitalizations in the United States are complicated by acute kidney injury, accounting for an estimated $10 billion in excess costs to the health care system.^{1 - 4} Acute kidney injury has been shown to be a potent predictor of excess length of stay, morbidity, and mortality in a number of clinical settings. The incidence of acute kidney injury has increased more than 4-fold since 1988 and is estimated to have a yearly population incidence of more than 500 per 100 000 population^{2 - 3} —higher than the yearly incidence of stroke.^{5 - 6} Survival from an episode of acute kidney injury may be increasing by virtue of advances in critical care medicine and dialysis technologies. In short, more hospitalized patients are being discharged alive after an episode of acute kidney injury.

The report by Wald and colleagues⁷ in this issue of JAMA provides valuable insights into the complex complications faced by survivors of an episode of severe acute kidney injury. Using linked administrative health databases covering the entire province of Ontario, Canada, the authors addressed the long-term risks of death and dialysis dependence among individuals who developed acute kidney injury requiring acute temporary dialysis during hospitalization. During a 10-year period between 1996 and 2006, they identified 18 551 individuals with acute kidney injury requiring dialysis, which corresponds to an approximate yearly incidence of 19 per 100 000 population—lower than the estimate of 24.4 per 100 000 population reported in Northern California between 1996 and 2003.² After excluding 3321 individuals who had previous acute kidney injury, dialysis, or kidney transplantation in the preceding 5 years, and 202 who had extreme lengths of hospital stay, the authors identified 15 028 patients with a first hospitalization for acute kidney injury requiring dialysis. More than 40% of these individuals died during hospitalization, in keeping with previous reports of the grave implications of severe acute kidney injury. Nearly half of these patients recovered kidney function for at least 30 days following hospitalization, attesting to the remarkable ability of the kidneys to repair and regenerate even after severe, dialysis-requiring injury. Another 23% of patients required further dialysis within 30 days of discharge, but it is not reported how many of those required chronic dialysis.

The final study cohort included 4066 survivors, 3769 (92.7%) of whom were matched to control patients and observed for a median of 3 years after discharge. Even among this selected cohort of survivors, mortality rates exceeded 10% per year. One of every 12 survivors of acute kidney injury requiring acute dialysis required subsequent initiation of chronic dialysis despite being dialysis-free at 30 days after discharge.⁷

These findings are noteworthy even without considering the next step in the analysis, which was to compare this incidence rate against that of matched individuals without acute kidney injury. From the perspective of a clinician caring for an individual with severe acute kidney injury, the findings by Wald et al⁷ provide an important quantitative estimate that can be shared with affected patients and their families: even in the best of circumstances—meaning survival during hospitalization and recovery of kidney function sufficient to stop dialysis for a month—there is almost a 10% chance of requiring chronic dialysis in the next few years.

The chronic dialysis incidence rate reported by Wald et al⁷ is 72 times higher than that reported for the general population in the United States in 2006 (366/1 million person-years).⁸ This finding has important implications for the postdischarge care of patients successfully treated with acute temporary dialysis: follow-up care with a nephrologist for secondary prevention is absolutely necessary. These findings also highlight the magnitude of the problem of acute kidney injury as a cause of end-stage renal disease (ESRD): extrapolating from the data of Wald et al,⁷ a rough estimate of the yearly incidence of ESRD due to acute kidney injury is 0.3 per 100 000 population, which is approximately one-third of the incidence of ESRD secondary to cystic kidney disease. The true magnitude is even higher because this estimate does not consider the 3481 individuals excluded from the final cohort because of the need for dialysis during the 30 days following hospitalization. If only 30% of those individuals developed ESRD by the definition used by Wald et al,⁷ then the yearly incidence would be 1.0 per 100 000, accounting for approximately 3% of the overall yearly incidence of ESRD in the United States.

Acute kidney injury is therefore a non-negligible cause of ESRD—the reason why is a more difficult question to answer. Does acute kidney injury lead to an acceleration of the normal age-related decline in glomerular filtration rate or is acute kidney injury a marker for other factors that are causally related to the development of kidney failure? Several lines of basic science evidence provide pathophysiological rationale for a causal relationship between acute kidney injury and eventual kidney failure.^{9 - 11} The epidemiological evidence provided by Wald et al⁷ is less convincing on this count and could, in fact, underestimate the actual association between acute kidney injury and progressive chronic kidney disease.

To identify a well-matched cohort against which to compare postdischarge outcomes, Wald et al⁷ relied solely on administrative data, as laboratory data were not included in the Ontario database. Patients with acute kidney injury were matched to hospitalized controls who did not sustain acute kidney injury or require dialysis. Matching was performed according to age, history of chronic kidney disease (as identified by administrative codes), need for mechanical ventilation during hospitalization, and propensity score for developing acute kidney injury requiring dialysis. The latter score was in turn derived by a number of demographic and comorbidity variables as well as an array of procedure and diagnostic codes spanning the 5 years before hospital discharge. Comparing outcomes in the acute kidney injury and nonacute kidney injury groups, the investigators found that the risk of needing chronic dialysis was increased more than 3-fold in the acute kidney injury group, whereas the risk of death appeared slightly lower (although not reaching statistical significance) in the acute kidney injury group.

The major shortcoming of using administrative data as the primary basis for matching is misclassification of exposure status, specifically in this study for the presence or absence of chronic kidney disease. Chronic kidney disease is poorly identified by administrative data, with reported sensitivities of less than 25%.¹² While the 2 groups had comparable proportions of chronic kidney disease (approximately one-fourth of each group), it is not possible to determine if actual renal function was comparable. Differential misclassification by administrative data to identify clinical diagnoses has been reported previously; for example, billing codes for acute renal failure were found to be more sensitive (ie, more likely to be recorded) in acute renal failure patients who died during hospitalization than in those who survived.¹³ Furthermore, chronic kidney disease is a heterogeneous condition with a spectrum of severity that may not be accurately captured by administrative codes. The 2 main variables used to define chronic kidney disease, estimated glomerular filtration rate and proteinuria, are missing from the Ontario databases. Both variables are strong predictors of the development of severe acute kidney injury requiring dialysis, even stronger than diagnosed hypertension and diabetes mellitus.

Hsu et al¹⁴ have found that the presence of proteinuria confers a nearly 3-fold increased risk of acute kidney injury requiring dialysis, while chronic kidney disease stage 4 (estimated glomerular filtration rate, 15-29 mL / min/1.73 m²) confers a 20-fold increased risk. Baseline estimated glomerular filtration rate and proteinuria are also, in turn, strong predictors of the need for chronic dialysis. Hallan et al¹⁵ reported a nearly 66-fold increased risk of end-stage renal disease in patients with chronic kidney disease stage 4, and a 13-fold increased risk with micro-albuminuria. To be confident in relative risk estimates between matched groups requires confidence in the matching algorithm, which in this study omitted the most relevant laboratory variables. Residual confounding from baseline renal function, which is highly associated with outcome (chronic dialysis) and exposure (acute kidney injury requiring dialysis), is also likely in the study by Wald et al.⁷

The findings of an increased risk of chronic dialysis following an episode of severe acute kidney injury are biologically plausible, and confirmation in other epidemiologic studies has begun. Lo et al¹⁶ showed that acute kidney injury occurring in patients with normal or near-normal baseline renal function (as assessed by serum creatinine) led to a 28-fold increase in the risk of ESRD (using the Kaiser Permanente Database of Northern California, which includes laboratory data and administrative codes). Ishani et al¹⁷ have also shown, using administrative data, that acute kidney injury markedly increases the risk of ESRD in elderly individuals.¹⁰ Both of these studies^{16 - 17} reported an increased risk of long-term mortality after an episode of acute kidney injury, which Wald et al⁷ did not observe.

How can the contradictory findings of a higher risk of dialysis, but a slightly lower risk of death be reconciled, particularly since chronic dialysis is associated with a markedly increased risk of death? One possibility is that the matching algorithm and reliance solely on administrative data led to an excess of actual comorbidities or severity of illness in the nonacute kidney injury group. The higher risk of death in the nonacute kidney injury group could be due, in part, to worse baseline renal function (eg, lower estimated glomerular filtration rate and more proteinuria), which is a strong predictor of long-term mortality.¹⁸ If this was indeed the case, then the estimates provided by Wald et al⁷ in their careful analysis are quite possibly an underestimate of the actual magnitude of the risk of chronic dialysis and death conferred by an episode of severe acute kidney injury.

Based on the available evidence from administrative and laboratory-based databases, severe acute kidney injury seems to increase the risk of progressive chronic kidney disease and may increase the risk of death. Given the extraordinarily high rates of morbidity and mortality observed in chronic kidney disease patients and acute kidney injury patients, the complex interconnection between them, and increasing incidence of both, kidney disease prevention and treatment should be a major public health priority.