Our study demonstrates that AKI occurs in almost a quarter of emergency admissions in a large single centre university teaching hospital. This is even higher (38.1%) in the subgroup of patients where true baseline renal function was known at the time of admission. However the incidence of AKI varies depending on which scoring system was used. The differences between the less severe stages when using RIFLE or AKIN are small, and this is consistent with what has been seen in previous validation studies for both systems. The time criteria used in the definition for each system, with RIFLE using a change in serum creatinine or eGFR over 7 days as opposed to 48 hours is the most likely explanation for this.
The AKIB system, which does not include urine output, gives a lower incidence of total AKI at 20.8%. This suggests that urine output may be more sensitive than serum creatinine as an early indicator of renal dysfunction (although probably much more difficult to assess accurately).
Whilst the overall incidence of AKI is higher at 38.1% in the group when baseline renal function was known compared to 25.4% in the whole group the breakdown across RIFLE or AKI stages 1/2/3 was almost identical. RIFLE R 60.5% (whole cohort) 61.3% (known baseline cohort), RIFLE I 27.9% (whole cohort) 26.9% (known baseline cohort), RIFLE F 11.6% (whole cohort) 11.8% (known baseline cohort). This is consistent with what has been seen in the Zeng US study with a split of AKI 1 70.9%, AKI 2 17.2%, AKI 3 12.0%.
Irrespective of the definition used, in both the total cohort and sub group analysis, we have shown AKI has a significant impact on length of hospital stay, use of critical care beds and mortality, even at the lower end of the severity of injury, which has also been seen in large studies in Italy and the US
[8, 20]. In our study AKI is associated with increased length of hospital stay, increased likelihood of admission to critical care (odds = 5.2, 2.3 to 12.7), and increased risk of death (odds = 3.7, 1.9 to 7.3). The presence of AKI and its severity correlated positively with both length of stay and hospital mortality. Median length of stay in the total cohort more than doubles from 4.0 days without AKI, up to 10.0 days with AKI (subgroup 5.0 days without AKI, up to 9.0 days with AKI). Although the more severe AKI episodes were small in number, so limiting the statistical power, the odds ratio of death increases from 2.0 with stage 1 to 9.0 for stage 3 (RIFLE-F). The figures in our study are comparable to other studies which have shown odds ratio of 2.0 for death in stage 1 and 10.1 in stage 3
Previous reports demonstrate that patients receiving RRT make up a small proportion of those reaching RIFLE-F in an ICU, yet the hospital mortality rate is greater than five times higher than that of the same ICU population without AKI
[20, 31–33]. In our study we also find that hospital mortality in critical care patients is nearly five times higher in patients with AKI (of any stage) compared with those patients that do not develop an AKI (8.1 vs 1.7% respectively). In the past this has been attributed to AKI simply being an indicator of illness severity, but this is now being challenged
. The increased morbidity and mortality seen with increasing severity of AKI is associated with an increased risk of “non-renal” complications such as bleeding and sepsis, but AKI may also influence remote organ function
. In the total cohort and subgroup, approximately one third of patients reaching RIFLE F required RRT. This has significant cost and resource implications, and if chronic dialysis patients’ admissions are included this equates to 3% of our total admissions.
Although the weeks studied were not consecutive we found no statistical differences between the patients in each week, and the variation in admission profile in the overall number of admissions, demographics, gender, age which could have affected the incidence of AKI were not different. Although the study was conducted in a single academic centre, only 9% of the admitted patients were under tertiary specialties and only 7% of patients were already known to the renal services, we feel our data is generally applicable to all but the smallest secondary care centres. Although the study is not as numerically large as some others, its major strength is that all admissions were captured and each case scrutinised by a nephrologist making it unlikely that any cases were missed, as can occur when relying only on electronic records as other larger studies have done
[8, 20, 32]. Furthermore our study population included very few patients (4.3%) who did not have any serum creatinine values compared to nearly 24% in Zeng’s study
A limitation which affects most studies in AKI, including ours, is the absence of a baseline creatinine for many patients hindering accurate assessment of baseline renal function. In our study 45% of patients had a known baseline, and so an estimated baseline was required in the remainder. In these cases we assumed a baseline eGFR of 75 mls/min per 1.73 m
 as recommended by the Acute Dialysis Quality Initiative, and used a serum creatinine of 60 umol/L for females and 80 umol/L for males (as the midpoint of our laboratory normal range). Recently published data from Siew et al. suggests that using many multiple imputation methods to calculate baseline function improve AKI misclassification but this has yet to be commonly accepted
. Other authors have encountered similar difficulties so that Hoste et al. used a ‘back calculated creatinine’ via MDRD using an assumed eGFR of 75 mls/min in approximately 50% of their cohort and Zeng et al. used an estimated baseline in 24.6% of their cohort. Back calculation via eGFR has its own problems since it is well recognised that estimated GFR is wildly inaccurate when serum creatinine is normal or near normal (+30% from true GFR), and serum creatinine needs to have been stable for at least 4 days before it approximates to true GFR.
Using RIFLE our AKI incidence of 23.6% is higher than that previously reported. The recent study by Zeng et al., demonstrated an incidence of 16.1% (using RIFLE) and Uchino 18.0%, both studies using a cohort including all hospitalisations. Our sub group analysis using the population with a known baseline suggested an even higher incidence, between 32.4 – 35.4% depending on the AKI definition used. Interestingly, the Zeng study also found an increased incidence when analysing only the sub-group with a known baseline creatinine value. Their variation from 18 to 33% depending on the definition of AKI used, and also depending on how estimation of unknown baseline renal function was calculated, gives an incidence comparable to our results, and higher than previous studies. The higher overall incidence of AKI at 67.4% in the Hoste study most likely reflects that it was studying a very different group of patients using an ICU population only
The potential risk factors assessed in the study are all commonly found co-morbidities in any hospital population. The significant effects of age, sepsis, diuretic use and pre-existing CKD highlight the importance of these factors in the development of AKI - aspects which may be under-recognized by clinicians. Educating clinicians to identify ‘at risk’ patients, both at the time of admission and also prior to this, will become increasingly important as we deal with an ageing population and its associated increased incidence of CKD. ‘Classical’ risk factors for AKI such as angiotensin blockade and radio-contrast use seem to play a smaller role than anticipated as risk factors for development of AKI in this and other studies. This may be due to already increased awareness of these modifiable factors. NSAIDs were not shown to be a significant risk factor despite being a well-recognised risk factor. The increased awareness of their association with renal dysfunction in an acute and chronic setting means has reduced their use in the elderly, CKD and heart failure patients – all groups which are at higher risk of AKI.
In summary, this study demonstrates that AKI is very common in acute unselected hospital admissions. Risk factors present at the time of admission suggest that it may be predictable, and therefore in some cases avoidable. If so, we calculate that a modest reduction of 10% in the incidence of AKI could save around 3,000 bed days per annum in similarly sized acute hospitals with 900–1000 beds. Whilst this has significant financial implications, of even greater importance is the benefit to the individual patient in terms of reduced morbidity, length of stay, long term renal outcome, and in some cases likelihood of death. To achieve such a reduction in the incidence of AKI would require a reliable method of early identification of the ‘at risk’ patient, and in particular those in whom AKI may be avoidable.