Skip To Main Content
 
 Figure 5.1Percent of Medicare patients aged 66+ (a) with at least one AKI hospitalization, and (b) percent among those with an AKI hospitalization who required dialysis, and (c) percent of patients with nephrology consultation, among those with a first AKI hospitalization, by whether an intensive care unit (ICU) stay was required, 2006-2016
 Figure 5.2Percent of Optum Clinformatics™ patients aged 22+ with at least one AKI hospitalization, by year, 2006-2016
 Figure 5.3Unadjusted rates of hospitalization with AKI and AKI requiring dialysis, per 1,000 patient-years at risk, by age, 2006-2016
 Figure 5.4Unadjusted rates of hospitalization with AKI, and AKI requiring dialysis, per 1,000 patient-years at risk, by race, 2006-2016
 Figure 5.5Unadjusted rates of hospitalization with AKI, and AKI requiring dialysis, per 1,000 patient-years at risk, by CKD and DM, 2006-2016
 Table 5.1Characteristics of Medicare and Optum Clinformatics™ patients with at least one hospitalization, by age, sex, race, CKD, DM, and presence of AKI, 2016
 Table AKDIGO definition and staging of acute kidney injury
 Table 5.2Characteristics of Veterans Affairs patients aged 22+ with at least one hospitalization, by age, sex, race, CKD, DM, presence and stage of AKI, defined by serum creatinine (KDIGO criteria), FY 2016
 Figure 5.6Cumulative probability of a recurrent AKI hospitalization within two years of live discharge from first AKI hospitalization in 2014 for Medicare patients aged 66+, (a) overall, (b) by age, (c) by race, and (d) by CKD and DM
 Figure 5.7Cumulative probability of a recurrent AKI hospitalization within two years of live discharge from first AKI hospitalization in 2014 for Optum Clinformatics™ patients aged 22+, (a) overall, (b) by age, (c) by race, and (d) by CKD and DM
 Figure 5.8Cumulative probability of death-censored ESRD, death, and the composite of death or ESRD within one year of live discharge from first AKI hospitalization occurring in 2014-2015
 Figure 5.9Cumulative probability of a claim for an outpatient nephrology visit within six months of live discharge from first AKI hospitalization, overall and by CKD, DM, 2006-2015
 Table BICD-9-CM and ICD-10-CM codes for Chronic Kidney Disease (CKD) stages
 Figure 5.10 Renal status one year following discharge from AKI hospitalization in 2014-2015, among surviving patients without kidney disease prior to AKI hospitalization, by CKD stage and ESRD status
 Figure 5.11Hospital discharge status of first hospitalization for Medicare patients aged 66+ (a) with diagnosis of AKI during stay, and (b) without diagnosis of AKI during stay, 2016
Download
Search This Page
Search All
Translate

Chapter 5: Acute Kidney Injury

  • In 2016, 4.4% of Medicare fee-for-service beneficiaries experienced a hospitalization complicated by Acute Kidney Injury (AKI), double the proportion of 2.2% in 2006 (Figure 5.1).
  • Risk of AKI increases with age and in the presence of comorbidities such as chronic kidney disease (CKD) and diabetes mellitus (DM). About 1 in 5 hospitalized Medicare patients with both CKD and DM experience a hospitalization with AKI each year (Figure 5.5).
  • Among hospitalized veterans aged 22+ years, 25.4% met Kidney Disease: Improving Global Outcomes (KDIGO) guidelines for AKI as defined using serum creatinine-based criteria (Table A). This included 21.4%, 0.8%, and 3.2% of patients with Stage 1, Stage 2, and Stage 3 AKI (Table 5.2). Just over half (52.6%) of patients meeting criteria for AKI were given a diagnosis of AKI.
  • In 2014, Medicare patients aged 66+ years who were hospitalized for AKI had a 36% cumulative probability of a recurrent AKI hospitalization within one year (Figure 5.6.a). For Optum Clinformatics™ patients aged 22+ years, the probability of recurrent AKI hospitalization was 23% (Figure 5.7.a).
  • Among Medicare patients without a pre-existing diagnosis of CKD, 30.8% were given a new diagnosis of CKD in the year following an AKI hospitalization (Figure 5.10.a). In the Optum Clinformatics™ population, 33.8% of patients with an AKI hospitalization were newly classified as having CKD in the subsequent year (Figure 5.10.b). In contrast, among Medicare patients with a “new” diagnosis of CKD in 2016, 25% had an AKI hospitalization in the preceding year.
  • Among Medicare patients aged 66+ years with a first AKI hospitalization in 2016, the in-hospital mortality rate was 8.2%, or 13.2% when including discharge to hospice. Comparable mortality rates for non-AKI hospitalizations were 1.8% and 3.8%. Less than half of all patients returned to their home on discharge, as compared to two-thirds of non-AKI patients, while 30.1% were discharged to an institution such as a rehabilitation or skilled nursing facility (Figure 5.11).

Introduction

Acute kidney injury (AKI) is a common complication among hospitalized patients, and is associated with substantial morbidity and mortality. Among survivors, AKI is recognized as a major risk factor for the development of chronic kidney disease (CKD). Studies have demonstrated significantly increased long-term risk of CKD and ESRD following AKI, even after initial kidney function recovery (Heung, 2012). Furthermore, this relationship is bidirectional—CKD patients are at substantially higher risk for AKI. As a result, AKI is frequently superimposed on CKD, and can contribute significantly to progression of CKD. As such, an examination into the epidemiology and outcomes of AKI is an intrinsic aspect to understanding the landscape of CKD.

This year we again present data from three sources: the Medicare 5% sample, the Optum Clinformatics™ Data Mart dataset (from OptumInsight, representing claims from a large U.S. national health insurance company), and national data from the U.S. Department of Veterans Affairs (VA) health system. Medicare and Optum Clinformatics™ administrative data do not contain clinical or biochemical data with which to identify an AKI episode using the consensus criteria that are based on changes in serum creatinine or urinary output. In these data sources, episodes of AKI were identified using ICD-9-CM and ICD-10-CM (International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification) diagnosis codes from claims. While this approach carries a high degree of specificity, an important limitation of this indirect method is poor sensitivity, generally <30%, and even lower for less severe cases of AKI. In particular, trends in AKI incidence must be interpreted with caution due to the possibility of “code creep”, whereby non-clinical factors such as changing billing thresholds or increased awareness and recognition of AKI increase the likelihood of administrative coding for AKI. Thus, a rising incidence of AKI may represent a true increase in cases, an increased likelihood to code for AKI, or a combination of both factors. In addition, a lower threshold for coding would lead to identification of less severe episodes and an apparent decrease in the rate of associated adverse outcomes.

In contrast to Medicare and Optum Clinformatics™, VA data contain clinical information to identify episodes of AKI through serum creatinine-based criteria. We present some data from the VA population to illustrate the potential gap between AKI episodes identified by administrative coding versus clinical data.

We begin this chapter by exploring trends in hospitalizations complicated by AKI, and describing the characteristics of those patients. We refer to “AKI hospitalizations” as any hospitalization during which there was a diagnosis (billing code) of AKI; the AKI diagnosis was not necessarily the primary or admitting diagnosis. We focus on hospitalizations because the occurrence of AKI exclusively in the community is uncommon and often unrecognized. Next, we explore the risk of hospital readmissions with recurrent AKI, and describe follow-up care after an episode. We end by examining the impact of AKI on outcomes, including subsequent CKD status and patient disposition after an AKI hospitalization.

Methods

Starting with the 2013 claim year, the USRDS Coordinating Center has received the Medicare 5% sample from the Medicare Chronic Conditions Warehouse, a different data source than in previous years. This has coincided with a subsequent decrease in AKI hospitalizations, and we cannot rule out that this is an artifact of the differing source of the Medicare 5% data files. Conclusions regarding trends should be made in this context.

As noted above, for administrative data (Medicare and Optum Clinformatics™) AKI episodes were identified through diagnosis codes from claims. These claims could be from any point during hospitalization and were not limited to the primary diagnosis. AKI episodes are presented both as a proportion (where denominator is either all patients or all hospitalizations), and as a rate (where denominator is patient population at risk). For VA data, AKI was defined using serum creatinine-based criteria (see Table A below), but not urine output criteria.

For the Medicare data, we often present results for those aged 66 and older. This allows a full year of Medicare eligibility (ages 65-66) for us to assess the patient’s CKD and diabetes mellitus (DM) status prior to the hospitalization within which AKI occurred.

To supplement the Medicare data, we also present data on patients aged 22+ years from the commercial insurance plans of a large national U.S. health insurance company, as included in the Optum Clinformatics™ Data Mart from OptumInsight. These data represent mainly working-age people and their minor dependents. For the prevalence of CKD and related conditions among these patients, see Volume 1, Chapter 2, Identification and Care of Patients with CKD, Table 2.1 for demographic characteristics of the Optum Clinformatics™ population (all ages) and Table 2.2 (ages 22-64). Additionally, Table 5.2 of this chapter uses data from all patients hospitalized at a VA hospital during fiscal year 2016, to show AKI as defined by serum creatinine measurements and staged as outlined in the KDIGO clinical practice guideline for AKI (KDIGO, 2012). Note that urine output data was not available, so identification of AKI episodes did not include the KDIGO criteria related to urine output.

Age is a major risk factor for AKI. Each of the included datasets had interactions between sex and age that are important to keep in mind when comparing differences in AKI by sex. Within both Optum Clinformatics™ and the VA data, women were younger on average than men. In Optum Clinformatics™, 55.6% of women were between the ages of 22 and 39, compared to only 19.8% of men. Conversely, women in the Medicare 5% sample were older, on average: women had a mean age of 77.1 years while for men it was 75.4 years, and a higher proportion of women (20.4%) than men (13.3%) were aged 85 and older.

Note that the analyses for all figures, except Figure 5.11, were based on all beneficiaries meeting the specified inclusion criteria. In Figure 5.11, we excluded those beneficiaries who were admitted from a long-term care facility to the inpatient setting where the AKI hospitalization occurred. Therefore, the category of institution in this figure includes only those newly readmitted following a hospitalization.

Details of this data are described in the Data Sources section of the CKD Analytical Methods chapter. Also see the CKD Analytical Methods section of the CKD Analytical Methods chapter for an explanation of the analytical methods used to generate the study cohorts, figures, and tables in this chapter. Microsoft Excel and PowerPoint files containing the data and graphics for these figures and tables are available to download from the USRDS website.

Further details of the data utilized for this chapter are described in the Data Sources section of the CKD Analytical Methods chapter. For an explanation of the analytical methods used to generate the study cohorts, figures, and tables in this chapter, see the section on Chapter 5 within the CKD Analytical Methods chapter. Downloadable Microsoft Excel and PowerPoint files containing the data and graphics for these figures and tables are available from the USRDS website.

Characteristics of Patients with Acute Kidney Injury

The percentage of Medicare fee-for-service patients with an AKI hospitalization has doubled over the past decade (Figure 5.1.a). However, the rate of AKI with an intensive care unit (ICU) stay has been relatively stable since 2010, and the increase has been in patients who did not require ICU stay during their hospitalization. Over the same period, the proportion of AKI patients requiring inpatient dialysis declined. Not surprisingly, a higher proportion of patients with an ICU stay had AKI requiring dialysis, compared to patients without an ICU stay (Figure 5.1.b). The proportion of patients with an AKI hospitalization who had a nephrology consultation has also fallen over the past decade, from 42.1% in 2006 to 25.2% in 2016 (Figure 5.1.c). Together, these findings seem to support the notion of “code creep”, in which there may be greater identification in billing codes of less severe cases of AKI, including those occurring outside the ICU and those that are managed without nephrology input.

Figure 5.2 reveals a similar rising trend of AKI in the Optum Clinformatics™ population, although the overall percentage of patients with an AKI hospitalization was far lower for these younger patients, at 0.3% in 2016.

Figure 5.1 Percent of Medicare patients aged 66+ (a) with at least one AKI hospitalization, and (b) percent among those with an AKI hospitalization who required dialysis, and (c) percent of patients with nephrology consultation, among those with a first AKI hospitalization, by whether an intensive care unit (ICU) stay was required, 2006-2016

Figure 5.2 Percent of Optum Clinformatics™ patients aged 22+ with at least one AKI hospitalization, by year, 2006-2016

As shown in Figure 5.3, rates of AKI were strongly influenced by age. Among fee-for-service Medicare patients in 2016, the rate of AKI for those aged 66-69 was 23.0 per 1,000 patient years, increasing to 31.3, 44.2, 62.9, and 95.7 for those aged 70-74, 75-79, 80-84, and 85 years and older. Unadjusted rates of AKI have risen in all age groups over the past decade, although the rate of rise seems to have slowed since 2011 in patients younger than 80 years. The rates of AKI requiring dialysis have remained fairly consistent across all age groups over the past decade. Among Optum Clinformatics™ patients, the overall group AKI rate increased over time, peaking at 3.8 per 1,000 patient years in 2016. For the subgroup aged 66 and older, the 2011 rate was 23.7 per 1,000 patient-years and remained somewhat stable at 21.8 per 1,000 in 2016.

Figure 5.3 Unadjusted rates of hospitalization with AKI and AKI requiring dialysis, per 1,000 patient-years at risk, by age, 2006-2016

Figure 5.4 highlights differences in AKI rates by race. In 2016, among fee-for-service Medicare patients aged 66 and older, the incidence rate for those of Black race was 71.6 per 1,000 patient-years at risk compared to 44.7 and 35.8, in Whites and individuals of other races. A similar relationship was observed in the Optum Clinformatics™ population, albeit at much lower rates: 5.6, 4.0, and 3.0 per 1,000 patient-years at risk in Blacks, Whites, and individuals of other races. Rates of AKI rose across all race subgroups between 2006 and 2016. However, the rate of AKI requiring dialysis appears to have remained stable.

Figure 5.4 Unadjusted rates of hospitalization with AKI, and AKI requiring dialysis, per 1,000 patient-years at risk, by race, 2006-2016

As shown in Figure 5.5, incidence rates for AKI also varied substantially by underlying comorbidity. In 2016, Medicare patients with DM, but no known CKD, had an AKI incidence rate of 54.1 per 1,000 patient-years, compared to 27.0 per 1,000 patient-years in non-diabetic, non-CKD patients. Non-diabetic patients with CKD experienced an AKI incidence rate of 141.7 per 1,000 patient-years, while the rate in patients with both DM and CKD was 207.4 per 1,000.The overall rate of hospitalization with AKI appears to be stable between 2010 and 2016. However, the rate of AKI requiring dialysis has declined in patients with CKD and those with both CKD and DM.

The Optum Clinformatics™ population showed similar relationships. Patients with both CKD and DM experienced the highest rates of AKI hospitalization at 100.9 per 1,000 patient-years. However, their overall rates were much lower than among the Medicare population, presumably reflecting the younger age range in this population.

Figure 5.5 Unadjusted rates of hospitalization with AKI, and AKI requiring dialysis, per 1,000 patient-years at risk, by CKD and DM, 2006-2016

Table 5.1 presents characteristics of hospitalized Medicare and Optum Clinformatics™ patients in 2016, along with their demographic and comorbidity characteristics by whether or not AKI occurred (defined as an inpatient stay with any diagnosis for AKI during any hospitalization during the year). AKI occurs commonly in older adults, impacting nearly 25% of Medicare patients aged 66 or older who have at least one hospitalization, and the incidence rises with age. Persons over age 80 accounted for 44% of all hospitalizations and 51% of hospitalizations with AKI. Although males appear to be more likely to develop AKI than females, it is important to remember that this does not account for differences in age distribution, although an age adjustment would tend to exacerbate the gender differential. In both the Medicare and ClinformaticsTM populations, a higher proportion of Black/African American patients had AKI compared to Whites or Asians. Diabetes and pre-existing CKD are recognized as two major risk factors for AKI; at least one of these risk factors was present in 57.2% of Medicare patients with an AKI hospitalization and 23.4% of patients had both. Even in the younger Optum Clinformatics™ population, about 40.2% of patients with an AKI hospitalization had either DM, CKD, or both.

Table 5.1 Characteristics of Medicare and Optum Clinformatics™ patients with at least one hospitalization, by age, sex, race, CKD, DM, and presence of AKI, 2016

Table 5.2 presents characteristics of hospitalized VA patients who had an AKI hospitalization in fiscal year 2016. Here, AKI was defined using serum creatinine-based criteria per the KDIGO guidelines (Table A). The incidence of AKI generally increased with age, and among race/ethnicity groups the highest proportion of AKI was again observed among non-Hispanic Black patients. For VA patients with diabetes, about 26.0% had an AKI hospitalization as defined by KDIGO criteria. Although this proportion appears similar to that observed in the Medicare population, direct comparison is not possible due to unaccounted for differences in patient characteristics as well as differences in methodology to identify AKI episodes (i.e. clinical vs claims data). The percentage of VA patients with an AKI hospitalization increased to 42.4% among CKD patients, and 53.7% among patients with both DM and CKD. Of note, among VA patients with an AKI hospitalization as defined by KDIGO serum creatinine-based criteria, only 52.6% were given a diagnosis of AKI.

Table A KDIGO definition and staging of acute kidney injury

Table 5.2 Characteristics of Veterans Affairs patients aged 22+ with at least one hospitalization, by age, sex, race, CKD, DM, presence and stage of AKI, defined by serum creatinine (KDIGO criteria), FY 2016

Readmission Associated with Acute Kidney Injury

Figures 5.6 and 5.7 show the probability of a patient’s recurrent AKI hospitalization after live discharge from an initial AKI hospitalization. Among 2014 Medicare patients aged 66 and older the overall probability of a recurrent AKI event was 0.36 in the next 12 months and 0.49 by 24 months, as shown in Figure 5.6.a. Among Optum Clinformatics™ patients, these probabilities were 0.23 and 0.31. In contrast to first episodes, the rate of recurrent AKI was relatively similar across age groups in the fee-for-service Medicare population (Figure 5.6.b). Interpretation of this finding is limited, however, because of the effect of death censoring, which was higher in older age groups. In Optum Clinformatics™ patients, who represent a wider range of ages, older patients appeared to have a higher probability for recurrent AKI (Figure 5.7.b).

In both the Medicare and Optum Clinformatics™ populations, Blacks had a higher probability of recurrent AKI compared to Whites or individuals of other races (Figures 5.6.c and 5.7.c). Similarly, having either DM or CKD was associated with an increased probability for recurrent AKI compared to having neither (see Figures 5.6.d and 5.7.d). The highest probability for recurrent AKI was for patients with both DM and CKD, reaching 0.59 by 24 months among Medicare patients and 0.38 among Optum Clinformatics™ patients. In contrast, Medicare patients with neither comorbidity had a cumulative probability for recurrent AKI hospitalization of 0.33 by 24 months, while their Optum Clinformatics™ counterparts had a probability of 0.17 by 24 months.

Siew et al. (2016) examined recurrent AKI for VA patients in 2003 and 2010 who survived their first AKI hospitalization (n=11,683). Of these, 8.5% had a second AKI episode within 30 days, 14.6% within 90 days, 19.5% within 180 days, and 25.3% with 12 months. AKI was defined according to KDIGO criteria using serum creatinine.

Figure 5.6 Cumulative probability of a recurrent AKI hospitalization within two years of live discharge from first AKI hospitalization in 2014 for Medicare patients aged 66+, (a) overall, (b) by age, (c) by race, and (d) by CKD and DM

Figure 5.7 Cumulative probability of a recurrent AKI hospitalization within two years of live discharge from first AKI hospitalization in 2014 for Optum Clinformatics™ patients aged 22+, (a) overall, (b) by age, (c) by race, and (d) by CKD and DM

Patient Care and Outcomes

Poor short-term outcomes for AKI, including hospital mortality, are well recognized. However, survivors of an AKI hospitalization continue to be at risk for significant adverse outcomes. Figure 5.8 illustrates that, among survivors of an AKI hospitalization in 2014-2015, the overall probability of developing ESRD in the following year was about 2% in the Medicare fee-for-service population aged 66 and older, and 5% in the Optum Clinformatics™ population. The seemingly paradoxical higher risk for ESRD in the younger Optum Clinformatics™ population may be due to higher competing risk of death in the Medicare population: in this same period, the probability of death was 40.6% and 11.7% in the Medicare and Optum Clinformatics™ populations, respectively.

Figure 5.8 Cumulative probability of death-censored ESRD, death, and the composite of death or ESRD within one year of live discharge from first AKI hospitalization occurring in 2014-2015

Recognizing that AKI can be associated with adverse long-term renal outcomes, including CKD and ESRD, both KDIGO guidelines and HP2020 objectives recommend follow-up renal evaluation after an AKI episode. In 2015, 16% of Medicare patients discharged alive from an AKI hospitalization had outpatient nephrology follow-up within the next six months, while 17% of Optum Clinformatics™ patients had follow-up over the same period. As shown in Figure 5.9, follow-up rates varied by comorbidity. Among patients with AKI superimposed on pre-existing CKD, but without DM, 16% of Medicare and 14% of Optum Clinformatics™ patients were seen by a nephrologist within six months following discharge. For patients with both CKD and DM, these proportions rose to 24% and 21%. In contrast, just 3% of Medicare and 9% of Optum Clinformatics™ AKI patients without DM or CKD were seen by a nephrologist by six months following an AKI hospitalization.

Trends over the past decade showed a slight decrease in post-AKI hospitalization nephrology follow-up for both the Medicare and Optum Clinformatics™ populations. This may once again reflect code creep: the milder cases of AKI captured by diagnosis may have been the least likely to require nephrology referral.

Figure 5.9 Cumulative probability of a claim for an outpatient nephrology visit within six months of live discharge from first AKI hospitalization, overall and by CKD, DM, 2006-2015

Changes in CKD Status after Acute Kidney Injury

CKD status changed significantly in the year following an AKI hospitalization, as shown in Figure 5.10. Among Medicare patients without baseline CKD, 30.8% were reclassified as having some degree of CKD, including 0.2% being declared ESRD. In the Optum Clinformatics™ population, about 33.8% of patients with an AKI hospitalization were newly classified as having CKD in the subsequent year, and 2.6% were given a diagnosis of ESRD. Although the percent of patients with ESRD was markedly higher in the younger Optum Clinformatics™ population as compared to Medicare patients, it is important to note that these were proportions of surviving patients only. Table B shows the ICD-9-CM diagnosis codes used to define stages of CKD for Figure 5.10.

Table B ICD-9-CM and ICD-10-CM codes for Chronic Kidney Disease (CKD) stages

Figure 5.10 Renal status one year following discharge from AKI hospitalization in 2014-2015, among surviving patients without kidney disease prior to AKI hospitalization, by CKD stage and ESRD status

In Figure 5.11, we examined the status and disposition of 2016 Medicare AKI patients once they were discharged from the hospital. We excluded patients admitted from a skilled nursing facility (SNF; n=1,942), leaving 57,256 AKI discharges. Among AKI patients aged 66 and older about 49.1% were discharged directly to their home. Mortality (including those discharged to hospice) was 13.2%, while 30.1% of patients were discharged to institutions such as short-term SNFs, rehabilitation hospitals, or long-term care facilities. By comparison, among hospitalized Medicare patients without a diagnosis of AKI (excluding those admitted from a SNF, n= 2,837, leaving 174,193 discharges), 68.8% returned home and approximately 22.7% were discharged to institutions. It is worth noting that, due to data limitations, we cannot fully ascertain and exclude admissions from residential facilities; therefore the high rate of “long-term care facility” in the discharge status could be a reflection of a higher rate of admissions from these facilities.

Figure 5.11 Hospital discharge status of first hospitalization for Medicare patients aged 66+ (a) with diagnosis of AKI during stay, and (b) without diagnosis of AKI during stay, 2016

References

Chawla LS, Eggers PW, Star RA, Kimmel PL. Acute kidney injury and chronic kidney disease as interconnected syndromes. N Engl J Med 2014;371:58-66.

Grams ME, Waikar SS, MacMahon B, Whelton S, Ballew SH, Coresh J. Performance and limitations of administrative data in the identification of AKI. Clin J Am Soc Nephrol 2014;9:682-689.

Heung M, Chawla LS. Predicting progression to chronic kidney disease after recovery from acute kidney injury. Curr Opin Nephrol Hypertens 2012;21:628–634.

Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Workgroup. KDIGO clinical practice guideline for acute kidney injury. Kidney Int 2012;2:1-138.

Siew ED, Parr SK, Abdel-Kader K, Eden SK, Peterson JF, Bansal N, Hung AM, Fly J, Speroff T, Ikizler RA, Matheny EW. Predictors of Recurrent AKI. J Am Soc Nephrol 2016;27: 1190-1200.

Waikar SS, Wald R, Chertow GM, Curhan GC, Winkelmayer WC, Liangos O, Sosa MA, Jaber BL. Validity of international classification of diseases, ninth revision, clinical modification codes for acute renal failure. J Am Soc Nephrol 2006;17:1688–1694.