aThe numerator denotes the number of events (requirement for chronic dialysis) and the denominator denotes the number of patients at risk within that subgroup. Hazard ratios are adjusted for the propensity score for acute kidney injury and dialysis and participant age. Effect estimates are not shown; models did not converge.
Wald R, Quinn RR, Luo J, Li P, Scales DC, Mamdani MM, Ray JG, University of Toronto Acute Kidney Injury Research Group FT. Chronic Dialysis and Death Among Survivors of Acute Kidney Injury Requiring Dialysis. JAMA. 2009;302(11):1179-1185. doi:10.1001/jama.2009.1322
Author Affiliations: Divisions of Nephrology (Dr Wald) and General Internal Medicine (Dr Ray), St Michael's Hospital; The Keenan Research Centre, Li Ka Shing Knowledge Institute of St Michael's Hospital (Drs Wald, Mamdani, and Ray); Department of Medicine, University of Toronto (Drs Wald, Scales, Mamdani, and Ray); Division of Nephrology (Dr Quinn) and Department of Critical Care Medicine (Dr Scales), Sunnybrook Health Sciences Centre; and the Institute for Clinical Evaluative Sciences (Drs Luo, Li, Scales, Mamdani, and Ray), all in Toronto, Ontario, Canada, and Faculty of Medicine, King Saud University, Riyadh, Saudi Arabia (Dr Mamdani).
Context Severe acute kidney injury among hospitalized patients often necessitates initiation of short-term dialysis. Little is known about the long-term outcome of those who survive to hospital discharge.
Objective To assess the risk of chronic dialysis and all-cause mortality in individuals who experience an episode of acute kidney injury requiring dialysis.
Design, Setting, and Participants We conducted a population-based cohort study of all adult patients in Ontario, Canada, with acute kidney injury who required in-hospital dialysis and survived free of dialysis for at least 30 days after discharge between July 1, 1996, and December 31, 2006. These individuals were matched with patients without acute kidney injury or dialysis during their index hospitalization. Matching was by age plus or minus 5 years, sex, history of chronic kidney disease, receipt of mechanical ventilation during the index hospitalization, and a propensity score for developing acute kidney injury requiring dialysis. Patients were followed up until March 31, 2007.
Main Outcome Measures The primary end point was the need for chronic dialysis and the secondary outcome was all-cause mortality.
Results We identified 3769 adults with acute kidney injury requiring in-hospital dialysis and 13 598 matched controls. The mean age was 62 years and median follow-up was 3 years. The incidence rate of chronic dialysis was 2.63 per 100 person-years among individuals with acute kidney injury requiring dialysis, and 0.91 per 100 person-years among control participants (adjusted hazard ratio, 3.23; 95% confidence interval, 2.70-3.86). All-cause mortality rates were 10.10 and 10.85 per 100 person-years, respectively (adjusted hazard ratio, 0.95; 95% confidence interval, 0.89-1.02).
Conclusions Acute kidney injury necessitating in-hospital dialysis was associated with an increased risk of chronic dialysis but not all-cause mortality.
Acute kidney injury, which leads to a sudden decline in kidney function, is a common and serious complication of hospitalization in the adult population.1- 4 Many patients with severe acute kidney injury require initiation of hemodialysis or hemofiltration (termed dialysis throughout), and their in-hospital mortality rate ranges from 45% to 70%.5- 12 Among those who survive, as many as 15% require dialysis at the time of discharge.6,9
Little is known about the long-term outcomes of patients with acute kidney injury requiring in-hospital dialysis, especially once they leave the hospital and recover enough kidney function to be free of dialysis in the short term. Prior studies were conducted in single centers or within restricted geographic regions,5,8,10,13,14 and few studies observed patients for more than 1 year.8,14,15 Accordingly, little is known about a common clinical scenario in which prognostic implications include resource allocation, specialist follow-up, and dialysis planning.
The universal health care system in the Canadian province of Ontario provides a unique opportunity to study long-term outcomes and health care use across multiple study centers and among an entire population. We evaluated the long-term risk of chronic dialysis and death among hospitalized patients who sustained acute kidney injury requiring acute dialysis.
We completed a retrospective matched cohort study using linked administrative health databases for all of Ontario. We identified all adults aged 19 years and older admitted to an acute care hospital between July 1, 1996, and December 31, 2006, for whom length of stay was less than 180 days. To identify a cohort of survivors, we focused on those individuals who lived for at least 30 days after the index hospitalization discharge date without receipt of dialysis or rehospitalization. We excluded individuals with a diagnosis of acute kidney injury, receipt of a kidney transplant, or any form of dialysis in the 5-year period preceding the index hospitalization.
As the exposed group, we selected from the previously described cohort all individuals who had both acute kidney injury and acute dialysis during the index admission. Each exposed patient was then matched with 4 unexposed hospitalized patients (without acute kidney injury or acute dialysis) according to the following matching variables: age within 5 years, sex, history of chronic kidney disease in the 5-year period prior to the index hospitalization admission date, receipt of mechanical ventilation during the index hospitalization, and logit of a propensity score for developing acute kidney injury and dialysis ±0.2 standard deviations.
The receipt of mechanical ventilation was used as a surrogate for admission to a critical care unit.16 The propensity score was used to characterize known risk factors for acute kidney injury and dialysis.17 The propensity score was derived using a multivariable logistic regression model that included the following variables: age (continuous in years), sex, and income quintile at the hospital discharge date; chronic kidney disease in the 5 years preceding the hospital admission date; number of hospitalizations or health care visits to any physician or emergency department in the 5 years preceding the hospital discharge date; receipt of coronary, carotid, or peripheral arterial revascularization, cardiac valve surgery, coronary angiography, cardiac stress testing, echocardiography, single 12-lead or ambulatory 24-hour electrocardiography, pacemaker insertion or carotid Doppler ultrasonography in the 5 years before the hospital discharge date; diagnosed liver dysfunction, heart failure, diabetes mellitus, cancer, cerebrovascular disease, cardiac dysrhythmias, myocardial infarction or angina, peripheral artery disease, major infection, alcoholism, and score on the Charlson Comorbidity Index18 in the 5 years before the hospital discharge date; quarter and year of the index hospitalization; and sepsis, receipt of mechanical ventilization, nonruptured aortic aneurysm repair, or cardiac surgery, all within the index hospitalization. Although 4 nonexposed matches were sought for each exposed patient, those with at least 1 match were included.
The study was completed using linked provincial health care administrative databases at the Institute for Clinical Evaluative Sciences. These databases have been used in numerous large-scale epidemiologic studies.19,20 Hospitalization records were obtained from the Canadian Institute for Heath Information Discharge Abstract Database (DAD), which was used to identify an acute hospitalization, the presence of acute kidney injury, and receipt of acute dialysis, as well as in-patient comorbidities and procedures (eTable). The DAD contains the encrypted health care number, age and sex of the individual, date of admission and discharge, and as many as 16 diagnoses in the International Classification of Diseases, Ninth Revision (before the year 2002) and 25 diagnoses in the International Classification of Diseases, Tenth Revision (from 2002 onward) (eTable). The Ontario Health Insurance Plan database contains records of all physician billing for outpatient and inpatient services, including a service date and a single diagnosis. The Registered Persons Database contains health care identifiers for all eligible individuals, as well as a date of death.
Income quintile was based on the 2001 Statistics Canada Census, which assigns income quintiles to each neighborhood community, with 1 representing the lowest income quintile and 5 representing the highest.
The health care databases were linked anonymously using encrypted individual health card numbers to safeguard patient confidentiality. The current study was approved by the research ethics board at Sunnybrook Health Sciences Centre.
The acute kidney injury and dialysis cohort was assembled by identifying hospitalized patients with both a diagnostic code in the DAD of acute kidney injury and either a procedure code in the DAD for dialysis or a physician claim for dialysis made through the Ontario Health Insurance Plan database during the index hospitalization (eTable). To focus our interest on acute kidney injury survivors, participants were included who survived for at least 30 days postdischarge with no evidence of dialysis treatment or the need for rehospitalization during this time.
Patients were tracked for outcomes beginning 30 days after discharge. The main study outcome was receipt of chronic dialysis for at least 90 days. The date of the first dialysis claim was considered to be the date of initiation of chronic dialysis. The secondary outcome was all-cause mortality occurring more than 30 days after hospital discharge.
Continuous variables for individuals with and without acute kidney injury and dialysis were expressed as a mean (SD) or median (interquartile range [IQR]), and compared using an unpaired t test or Kruskal-Wallis test, respectively. Categorical variables were expressed as a percentage and compared using a χ2 test.
Time-to-event analyses were performed until March 31, 2007. For the chronic dialysis end point, an individual was censored at death or at the end of the study period. Incidence rates of both chronic dialysis and all-cause mortality were determined for participants with and without acute kidney injury and dialysis. Crude hazard ratios (HRs) and 95% confidence intervals (CIs) were derived from Cox proportional hazards models, and matched individuals without acute kidney injury and dialysis were the reference group. We further adjusted for age (continuous in years) and the propensity score for acute kidney injury and dialysis in the multivariable models, in case there was residual confounding beyond the matching process. Survival curves for chronic dialysis and all-cause mortality were generated from unadjusted Cox models. Adjusted HRs for receipt of chronic dialysis were further stratified according to participant characteristics.
Based on the actual number of participants enrolled in the current study, we had greater than 95% statistical power to detect an HR of 2.00 or more for the primary study outcome at a conventional P value of .05. All P values were 2-sided, and the significance level was set at .05. Analyses were performed using SAS 9.1 for UNIX (SAS Institute Inc, Cary, North Carolina).
We identified 4066 individuals with acute kidney injury requiring dialysis who survived at least 30 days after hospital discharge without receiving dialysis or further hospitalization (Figure 1). At least 1 match was identified for 3769 (92.7%) of the exposed patients (Table 1). Hence, 3769 individuals with acute kidney injury and dialysis and 13 598 matches with neither acute kidney injury nor dialysis were included in this study.
The mean age of the enrolled participants was 62 years, approximately 40% were women, and the mean Charlson Comorbidity Index score was 2.7 (Table 1). Slightly more than 25% of participants in each group had documented chronic kidney disease in the 5-year period before the index hospital admission. During their hospitalization, 47% of patients with acute kidney injury and dialysis received mechanical ventilation, compared with 42% of those without acute kidney injury and dialysis. The respective rates for cardiac surgery were 11.5% and 10.0%. Approximately 15% of participants in each group had sepsis. The rates of cardiac and noncardiac arteriography were also comparable between groups (Table 1).
After a median (IQR) follow-up of 3 (1-5) years, the incidence rates of chronic dialysis were 2.63 and 0.91 per 100 person-years among those with and without acute kidney injury and dialysis in the index hospitalization, respectively, corresponding to a crude HR of 3.26 (95% CI, 2.73-3.89) (Table 2; Figure 2A). After adjusting for age and the propensity score, the HR changed little (3.23; 95% CI, 2.70-3.86; Table 2).
The increased risk of chronic dialysis following a hospitalization complicated by acute kidney injury and dialysis was consistent across patient subgroups (Figure 3). This was especially so among patients without a preexisting diagnosis of chronic kidney disease (adjusted HR, 15.54; 95% CI, 9.65-25.03).
The incidence rate for all-cause mortality was 10.10 per 100 person-years among patients with acute kidney injury and dialysis, which was slightly lower than among those without in-hospital acute kidney injury (Figure 2B). After adjusting for age and the propensity score, the future risk of death was not significantly different between groups (HR, 0.95; 95% CI, 0.89-1.02; Table 2).
There were 297 (7.3%) participants with acute kidney injury and dialysis who could not be matched to a suitable control. Their respective incidence rates of chronic dialysis and all-cause mortality were 4.89 per 100 person-years and 14.18 per 100 person-years, which were higher than rates for the 3769 patients who were successfully matched (Table 2).
We analyzed all patients admitted to Ontario hospitals over a 10-year period and found that survivors of a hospitalization complicated by acute kidney injury requiring dialysis were approximately 3 times more likely to require chronic dialysis compared with those without acute kidney injury. However, no difference was observed between these groups for long-term mortality.
This study was designed to evaluate long-term renal health among survivors of acute kidney injury receiving dialysis in a large population with universal access to health care. Patients with acute kidney injury have a high acuity level during their hospitalization and a wide array of preexisting illness.11 Accordingly, we identified matched controls based on a series of key characteristics including the propensity to experience an episode of acute kidney injury requiring dialysis. Because we sought to evaluate long-term outcomes among survivors of severe acute kidney injury, and to reduce the possibility of survivor-treatment bias, we required a 30-day period following hospital discharge in which a participant survived and was free of dialysis.
We completed our study using administrative data, so there may have been a misclassification of the study exposure (acute kidney injury and dialysis) or other adjustment variables. In a recent validation study, the combination of a diagnostic code for acute kidney injury and a procedure code for dialysis had both sensitivity and specificity of more than 90% for capturing the combination of acute kidney injury and dialysis.21 However, given the low sensitivity of an acute kidney injury diagnosis code in isolation (28%-35%), it is possible that some participants classified as not having acute kidney injury actually did, leading to an underestimation of outcome differences between groups. In addition, we could not match approximately 7% of study patients with acute kidney injury and dialysis to unaffected controls. Since unmatched individuals had higher rates of chronic dialysis and mortality than those with at least 1 match, their exclusion attenuated the reported effect sizes. Finally, we cannot exclude the possibility of residual confounding despite our rigorous matching and adjustment on important variables.
We could not obtain measures of kidney function, such as serum creatinine or urinary protein excretion, neither at discharge nor during postdischarge follow-up, and were limited to measuring the outcome of chronic dialysis, for which reliable claims-based data exist. Indeed, many patients who survived acute kidney injury and dialysis may have had persistent or worsening renal dysfunction after hospital discharge, but not so severe as to warrant dialysis. We also used a widely accepted definition for chronic dialysis—the receipt of dialysis for at least 90 days. A shortcoming of this approach is that it misclassifies individuals who truly developed end-stage renal disease, but who died in the first 90 days after commencing chronic dialysis. However, since the risk of death was not significantly higher in individuals with acute kidney injury and dialysis than matched patients with no acute kidney injury, it is unlikely that this potential misclassification had an effect on the risk of chronic dialysis.
Small studies have previously evaluated the outcomes of survivors of acute kidney injury requiring dialysis. Schiffl studied 425 patients with acute tubular necrosis who required acute dialysis at a single institution.10,14 Among the 225 participants alive at discharge, survival was 66% and 53% at 1 and 5 years, respectively. Morgera et al used follow-up data for 89% of the 301 patients discharged from their institution following acute kidney injury and receipt of continuous dialysis.8 Survival at 6 and 12 months was 77% and 69%, respectively. Information about renal function, available for about one-third of survivors, revealed that 10% required chronic dialysis and 40% had some degree of persistent renal insufficiency. Bagshaw et al observed a 1-year survival rate of 90% and a chronic dialysis rate of 22% among 97 Albertans discharged alive following dialysis for acute kidney injury.13 While these studies clearly shed some light on the health outcomes of acute kidney injury survivors, most data were derived from single centers or local health regions and were based on small cohorts. The absence of a comparator group in these studies also prevents one from determining the degree to which acute kidney injury and dialysis is a prognosticator for long-term health status in excess of global patient comorbidity before and during the index hospitalization. Finally, since survivorship was assessed starting from the hospital discharge date, patients who died shortly following discharge, and who may not have been candidates for future dialysis, may have been inappropriately included.
Ishani et al previously used American Medicare administrative data to identify approximately 7000 patients older than 67 years with hospital-associated acute kidney injury who survived to discharge.22 Postdischarge mortality was doubled and the risk of chronic dialysis was increased by nearly 7 times among survivors of acute kidney injury. These risks were accentuated in patients with preexisting chronic kidney disease. However, their cohort study22 was limited to individuals older than 67 years, and they did not distinguish between patients with acute kidney injury who required dialysis and those who did not. This study used a fairly limited set of covariates, and thus, increased the likelihood of confounding within their risk estimates. Among individuals in our study with preexisting chronic kidney disease, an episode of acute kidney injury and dialysis was associated with an approximate 2-fold higher risk of a chronic dialysis, which rose to about 15 times higher among patients without previous chronic kidney disease (Figure 3). Comparing our findings with those of Ishani et al22 is challenging, since the Ishani study included controls with neither acute kidney injury nor prior chronic kidney disease. In contrast, more than one quarter of our control cohort had a history of preexisting chronic kidney disease.
Injury to the kidney, especially that necessitating acute dialysis, seems to independently predispose a patient to long-term dialysis, even among those who are free of dialysis at hospital discharge. This parallels basic science data that demonstrate the long-term effects of ischemia-reperfusion injury on the loss of the renal vasculature and progressive renal fibrosis.23,24 Acute kidney injury may induce irreversible chronic kidney damage in patients with previously healthy kidneys, and accelerate kidney disease progression among patients with preexisting chronic kidney disease.
In contrast to other studies,22 we found that a prior history of acute kidney injury and dialysis was not independently associated with long-term mortality (HR, 0.95; 95% CI, 0.89-1.02). Indeed, there was a surprising trend that suggested a lower associated risk of death in patients with prior acute kidney injury and dialysis. There are 2 likely explanations for this finding. First, although our comparator group comprised individuals without acute kidney injury (in whom acute dialysis was not indicated), those with acute kidney injury who received dialysis may have been a selected group of individuals who were more likely to survive. The tendency to administer life-saving therapies to those most likely to have a favorable prognosis has been demonstrated in other settings. For example, patients at lowest risk for cardiovascular disease and the highest likelihood of survival were prescribed efficacious drug therapies more commonly than those at higher risk.25 Second, as mentioned previously, approximately 7% of the 4066 survivors with acute kidney injury and dialysis were excluded from the data analyses due to a lack of at least 1 appropriate match. These individuals exhibited a higher mortality rate over time. Had these individuals been included in the final analyses, the small nonsignificant trend toward reduced mortality would likely have been further diminished.
Our findings expand on prior knowledge to provide clinicians with new information about the long-term effect of acute kidney injury that arises during a hospitalization. First, if affected patients survive to hospital discharge, then they remain at high risk of needing dialysis over the next 3 to 5 years. Patients who survive a hospitalization complicated by acute kidney injury requiring dialysis may benefit from specialized care to address complications of chronic kidney disease, and also from concerted efforts to prevent progression to chronic dialysis. At the same time, their high mortality rate is similar to hospitalized patients without acute kidney injury or need for dialysis. Hence, an episode of acute kidney injury requiring in-hospital dialysis may not be an independent contributing factor to long-term survival.
Corresponding Author: Ron Wald, MDCM, MPH, FRCPC, Division of Nephrology, St Michael's Hospital, 61 Queen St East, 9-140, Toronto, ON M5C 2T2, Canada (firstname.lastname@example.org).
Author Contributions: Drs Luo and Li had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Wald, Quinn, Scales, Mamdani, Ray.
Analysis and interpretation of data: Wald, Luo, Li.
Drafting of the manuscript: Wald, Ray.
Critical revision of the manuscript for important intellectual content: Wald, Quinn, Luo, Li, Scales, Mamdani.
Statistical analysis: Wald, Luo, Li, Scales, Ray.
Obtained funding: Wald.
Administrative, technical, or material support: Wald.
Study supervision: Mamdani, Ray.
Financial Disclosures: None reported.
Funding/Support: This study was supported by the Institute for Clinical Evaluative Sciences, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care; and a New Staff Dean's Grant from the University of Toronto Faculty of Medicine. Dr Scales reports support from a Canadian Institutes of Health Research New Investigator Award. Dr Wald reports receipt of an RCT Mentoring Program Award from the Canadian Institutes of Health Research and an unrestricted grant from Amgen. Dr Quinn reports support from a Canadian Institutes of Health Research Institute of Health Services and Policy Research Fellowship. Dr Ray reports support from a Canadian Institutes of Health Research New Investigator Award.
Role of the Sponsor: The Ontario Ministry of Health and Long-Term Care and the University of Toronto Faculty of Medicine had no involvement in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Other members of the University of Toronto Acute Kidney Injury Research Group: Stephen Lapinsky, MD (Mt Sinai Hospital); Karen Burns, MD, MSc, and Jan Friedrich, MD, MSc, DPhil (St Michael's Hospital); Neill Adhikari, MDCM, MSc, Michelle Hladunewich, MD, MSc, and Sheldon Tobe, MD (Sunnybrook Health Sciences Centre); and Robert Richardson, MD (University Health Network), Toronto, Ontario, Canada.
Disclaimer: The opinions, results, and conclusions reported in this article are those of the authors and are independent from the funding sources. No endorsement by the Institute for Clinical Evaluative Sciences or the Ontario Ministry of Health and Long-Term Care is intended or should be inferred.
Previous Presentations: Presented in part at the 2008 American Society of Nephrology Renal Week (Philadelphia, Pennsylvania).
Additional Contributions: We are grateful for the thoughtful input provided by Ahmed Bayoumi, MD, Tara Gomes, MHSc, Michael Paterson, MSc, and Jan Hux, MD. They received no compensation for their contributions.