A comparison of different diagnostic criteria of acute kidney injury in critically ill patients

Acute kidney injury (AKI) is very common, especially in the intensive care unit (ICU). It is also associated with increased mortality and a longer stay in the hospital [1‐7]. There have been many definitions, such as acute renal failure and renal impairment, and this has made it difficult to compare results across studies. In 2004, the Acute Dialysis Quality Initiative group proposed a classification for AKI: the Risk, Injury, Failure, Loss of Kidney Function, and End-stage Kidney Disease (RIFLE) classification, the first evidence-based consensus [8]. The classification includes three grades of severity of AKI (risk, injury, and failure) according to relative changes in serum creatinine (SCr) and urine output and two outcomes (loss of kidney function and end-stage kidney disease, or ESKD). It has been evaluated in many studies of critically ill patients with AKI and has shown good relevance for diagnosing and classifying the severity of AKI as well as comparable predictive ability for mortality [7, 9‐13].

In 2007, the Acute Kidney Injury Network (AKIN) group proposed a modified version of the RIFLE classification, which aimed to improve the sensitivity of AKI criteria [14]. There were several changes: an absolute increase in SCr of at least 26.4 μmol/L was added to stage 1; patients starting RRT were classified as stage 3, irrespectively of SCr; and the change in glomerular filtration rate (GFR) and the two outcome classes were removed. AKI diagnosis was based on change between two creatinine values within a 48-hour period for AKIN classification and within a 1-week window for RIFLE criteria. Severity of AKI in AKIN is staged over the course of 7 days by fold-change in creatinine from baseline.

The latest classification was proposed by the Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group, was based on the previous two classifications, and had the aim of unifying the definition of AKI [15]. According to this definition, AKI was diagnosed as an increase in SCr by at least 26.4 μmol/L within 48 hours or an increase in SCr to 1.5 times baseline, which is known or presumed to have occurred within 7 days before, or a urine volume of less than 0.5 mL/kg per hour for 6 hours. For KDIGO criteria, the 26.4 μmol/L increase needs to be within 48 hours but a 1.5-fold increase can occur within 7 days to diagnose AKI; and the 1-week or 48-hour timeframe is for diagnosis of AKI, not for staging. A patient can be staged over the entire episode of AKI. Increase in SCr to 3 times baseline, or SCr of more than 4.0 mg/dL (354 μmol/L), or starting RRT were all classified as stage 3. KDIGO removes the 0.5 mg/dL increase for creatinine more than 4 mg/dL to diagnose stage 3. Besides, KDIGO explicitly states that a rolling baseline can be used over 48-hour and 7-day periods for diagnosis of AKI, but it is unclear how this is handled in RIFLE or AKIN. The definition and difference among the three criteria are shown in Additional file 1.

Many studies have compared RIFLE with AKIN in critically ill patients, but only a few have compared KDIGO with these criteria in critically ill patients with AKI. The purposes of this study were to determine the incidence of AKI in critically ill patients according to the RIFLE, AKIN, and KDIGO criteria and to compare their predictive ability.

During the study period, 9,049 patients were consecutively admitted to one of 30 ICUs. In total, 5,942 patients were excluded; of these patients, 110 were younger than 18 years old, one received renal transplantation during the past 3 months, and 95 patients had received RRT before admission to the ICU. A further 5,725 patients were excluded because their length of stay in the ICU was less than 24 hours, and 11 were excluded because of insufficient clinical recordings. Finally, 3,107 patients were enrolled. The characteristics of the whole cohort are shown in Table 1.

Numerous studies have compared the RIFLE and AKIN criteria for AKI. However, the incidence of AKI still varied. Based on these two criteria, the KDIGO criteria were recently proposed in order to unify the definition of AKI. To date, only a few previous studies have compared the incidence and mortality of AKI in critically ill patients according to these three definitions [19‐21]. This is the first, large, multicenter study to compare these three different criteria in critically ill patients with AKI in China.

The incidence of AKI according to the KDIGO criteria was higher than that defined by RIFLE and AKIN, even after we excluded patients without known baseline creatinine. It was similar to the results of a study comparing definitions of AKI in hospitalized individuals in Boston [20] but differed from a retrospective study of patients after cardiac surgery, which concluded that incidence and outcome of AKI according to the RIFLE, AKIN, and KDIGO classification were similar [19]. The study of hospitalized patients conducted by Fujii and colleagues in Japan concluded that the rates of incidence of AKI according to RIFLE, AKIN, and KDIGO were 11%, 4.8%, and 11.6%, respectively [21]. KDIGO classified 126 (4.1%) more patients with AKI than RIFLE did, the majority of which were patients with stage 1. An in-depth analysis of these patients found that the majority (124 patients) were identified by a small increase in creatinine alone but that the remaining two patients received RRT at the same time. Firstly, we found that for some patients there was a decrease in creatinine after admission to the ICU, followed by a relative increase; these patients could be identified by KDIGO and AKIN because a rolling in-hospital baseline was used for the 48-hour rise, but not by RIFLE. Secondly, patients who received RRT would be classified as stage 3 by KDIGO and AKIN, irrespectively of SCr, but not by RIFLE.

When compared with AKIN, KDIGO diagnosed AKI in an additional 391 patients, including 25 patients with chronic kidney disease; these patients were predominantly stage 1, followed by stage 2 and stage 3. The median creatinine level in these 391 patients on their first day of admission to the ICU was much higher than the baseline level, and this means that AKI may have been present on the day of ICU admission or even before. According to the AKIN criteria, AKI was diagnosed by two creatinine measurements within 48 hours. However, most patients did not have creatinine measured every day prior to the ICU admission: thus, when creatinine at ICU admission was used, some community-acquired AKI cases may have been missed [22‐24]. In addition, patients with a slow reduction of renal function may have been missed by the AKIN criteria [25]. The KDIGO definition reserved the baseline creatinine from RIFLE as well as a small increase in creatinine from AKIN criteria, allowing greater sensitivity than RIFLE and AKIN.

All definitions showed comparable and excellent associations with worse outcome according to increased severity of AKI. As for the predictive ability of these criteria, all were found to be significant predictors of increased mortality using multivariate analysis adjusting for age, gender, diabetes, hypertension, chronic kidney disease, chronic heart failure, and SOFA score. These findings were identical to those of previous studies [3, 4, 26, 27]. Patients missed by RIFLE but identified by KDIGO, most of which were classified as stage 1, had a longer length of ICU stay than no-AKI patients based on KDIGO. The patients diagnosed by KDIGO criteria as stage 1 but missed by RIFLE had much higher mortality than patients without AKI based on KDIGO (20.8% versus 5.6%, P < 0.001). Thus, we deduced that a small increase in creatinine might be accompanied by increased mortality. Similar results were observed in other studies [28, 29]. A study by Wilson and colleagues determined that the magnitude of the decrease in creatinine generation rate may be correlated with the severity of illness [30]. In other words, the patients with a small increase in creatinine, accompanied by increased mortality and longer hospital stay, could be identified by KDIGO but not by RIFLE. The KDIGO definitions also showed a little better predictive ability than RIFLE did, according to the AUC curve for in-hospital mortality. For patients missed by AKIN but not by KDIGO, AKI was also an independent risk factor for mortality, but of low risk; and the mortality of these patients was only a little higher than that of no-AKI patients according to the KDIGO criteria (12.8% versus 5.6%, P < 0.01). In addition, the mortality of patients with AKI based on AKIN was a little higher than those on KDIGO (32.2% versus 27.4%, P = 0.006) and this was probably because KDIGO identified more patients in a mild severity level of AKI, with a relatively low mortality rate. According to the AUC curve, there was no significant difference between KDIGO and AKIN in the predictive ability for in-hospital mortality (0.757 versus 0.746, P = 0.12). Therefore, we concluded that KDIGO and AKIN were comparable on their predictive ability for in-hospital mortality. So whether this small increase in the mortality of these patients, identified by KDIGO but missed by AKIN, is of high risk requires more research. However, the study of hospitalized patients in Japan concluded that KDIGO and RIFLE achieved similar discrimination but that the discrimination of AKIN was inferior [21]. Given that their conclusion is different from ours, maybe more study is needed.

There are some limitations to our study. First, we used the simplified MDRD formula as baseline for patients without known baseline creatinine. In a prospective observational study, a good correlation of estimated as compared with observed baseline values was found for patients without chronic kidney disease [31]. Second, we did not have any records of creatinine during hospitalization but we did have records prior to ICU admission, and this may have caused the incidence of AKI by AKIN to be underestimated. The AKIN criteria recommend applying only the urine output criteria “following adequate fluid resuscitation”, which is ambiguous. In our study, we did not adhere strictly to this recommendation. Third, we received hourly records of urine output for most patients, but for others only the total urine volume in a 6-hour period was recorded. A study by Etienne Macedo and colleagues [32] concluded that there was no significant difference between assessing urine output every hour or the total urine volume in a 6-hour period for the detection of episodes of oliguria, and the latter did not decrease their sensitivity for identifying patients with AKI. Finally, we did not have data regarding additional factors that could influence urine output, such as diuretic therapy.

The incidence of AKI in critically ill patients varied according to the criteria used. The KDIGO criteria identified more patients as AKI than RIFLE and AKIN did. Compared with the RIFLE criteria, KDIGO was more predictive for in-hospital mortality, but there was no significant difference between AKIN and KDIGO.