Discussion

In this large cohort of hospitalised patients, AKI was detected in 15.2% of admissions, a proportion that agrees with other studies on general populations of inpatients in developed countries reporting the incidence ranging from 12% to 20% (7, 15, 16). The variability in the incidence may depend on adopted definitions for AKI and baseline SCr, the frequency of creatinine measurements and clinical settings. Patients with AKI were relatively old, half of them had more than 72 years, with a high burden of comorbidities. This is consistent with the existing evidence of AKI being particularly common in the elderly (17). Reduced renal reserve in older age(18) together with polypharmacy and greater susceptibility to nephrotoxic drugs, poses this population at high risk of AKI (19). Furthermore, advanced age is a risk factor for impaired recovery from AKI (20), progression to CKD and it is still not conclusively proven that elderly fully benefit from RRT (21). AKI development was related to a higher number of comorbidities, including but not limited to cardiovascular, diabetes, chronic pulmonary disease, hypertension and CKD, which confirms the conviction that AKI is a broad clinical syndrome of various etiologies more likely to occur in patients with a higher burden of comorbidities (10).
In epidemiological studies on AKI in hospital settings, the syndrome is often distinguished between community- and hospital-acquired assuming different causes and underlying pathophysiological processes. In our analysis, CA-AKI represented 38% of AKI population, which differs in comparison to previous investigations on AKI present at admission. In the study of Sawhney at el., the proportion was 27%(22); nevertheless, AKI originates in the community has been consistently found to be more common than hospital-acquired accounting for about 70% of all AKI episodes. This could be explained by the adoption of different criteria to define AKI(23), the extension of the time window for CA-AKI detection up to 48 hours (24) or defining the baseline SCr as normal when the value was unknown (25).
We observed that the risk profile did not differ much between CA- and HA-AKI(22), however, our results confirmed that pre-existing CKD, liver diseases, dementia and history of cancer are more common among patients presenting AKI at admission to the hospital(24). These patients also sustained more severe AKI than HA-AKI patients (23). One explanation could be more prevalent or possibly more severe CKD in patients with CA-AKI since the incidence and the severity of AKI increase considerably with lower levels of baseline eGFR (26). Besides, the high proportion of emergency admission and primary diagnoses in this group indicate that severe renal impairment could result indirectly from an acute condition that necessitated hospitalisation (for example severe infections) or directly from post-renal causes (urinary tract obstructions).
RRT during inpatient was required in 8.3% of CA- and 3.9% of HA-AKI and a large proportion of these patients died before discharge (30% and 50%, respectively) (data not shown) supporting the evidence that dialysis-requiring AKI is a strong predictor of mortality(17, 27). Given such poor prognosis along with the rapidly escalating incidence of dialysis-requiring AKI(28), development of new instruments such as standardised management recommendations on RRT initiation and discontinuation may be a measure to improve outcomes(29).
Overall, one in five of AKI patients died during the hospital stay which corresponds to mortality rates seen in other studies(4, 22), while in the absence of AKI, mortality runs at 3.6%. When comparing outcomes between CA- and HA-AKI, conclusions of a recent meta-analysis pointed out less severe clinical manifestation and lower mortality in CA-AKI (30). Higher risk of death in HA-AKI is commonly attributed to underlying chronic illness, specifically cardiovascular disease, increased incidence of complications during the hospital stay or iatrogenic origin of AKI (nephrotoxicity, surgeries) (24, 31), which are considered to do more harm to kidney than prerenal causes(32). In contrary, we did not find differences in in-hospital and 6th-month mortalities between CA- and HA-AKI in our cohort. These discordant conclusions cannot be explained by the time window we adopted to identify CA-AKI (first 24 hours), which differs from the criteria used by other authors; patients in whom AKI was detected in each of the first three consecutive days after admission had a comparable risk of death during hospitalisation (23.3%, 21.0 and 20% respectively). We suppose the reason for the discrepancy in findings may be dissimilarity in characteristics of underlying populations (age, prevalence of risk factors, in particular, CKD, socioeconomic status) and in settings, in which previous studies were conducted. Our hospital is a tertiary centre which provides specialist care, also for seriously ill patients transferred from smaller centres; therefore, the severity of AKI and mortality might have been affected.
Estimates of the regression model highlighted that severity of AKI, rather than its origin, is a strong and independent determinant of resource utilisation and mortality. It was particularly noticeable for AKI stage 3 but even mild episodes of kidney dysfunction also substantially impacted outcomes of the interest. It should be remembered; however, that reported causes of death are mostly related to coexisting conditions, including causes of AKI, rather than kidney injury(33, 34). This reinforces the fact that AKI is not a single condition and points to a need for a detailed characterisation of affected patients encompassing etiology, adequacies in management and the level of recovery to improve individualised patient-centred care effectively.
Noteworthy, the risk of death for AKI stage 1 was found to be significantly different in the community- and in the hospital-acquired syndrome. Arguably, this resulted from variant criteria determining stage 1 in the two groups; 15% of all AKI episodes (in patients with SCr concentration above 1.0 mg/dL) was defined by an absolute increase of 0.3 mg/dL above the reference SCr without reaching a relative increase of ≥ 50% within a week. Our observations of divergent prognosis are in line with the latest study of Sparrow et al. proving that in patients with AKI stage 1 defined as an absolute change in SCr concentration had shorter stays and were less likely to die in a hospital than those with a 50% relative increase(35).
Among HA-AKI, the highest instantaneously incidence was observed on the second day after admission. We suspect that part of these cases was community-acquired, with later manifestation due to the limited ability of SCr to timely reflect changes in kidney function. Nevertheless, whatever the origin, early risk assessment (for example, at the moment of hospital admission) and identification of high-risk patients provide the opportunity to intervene in the treatment and protect the kidneys from further damage. It is estimated that 20% of hospital-acquired AKI are avoidable(36). Given that each day more than 1% of patients in our cohort developed AKI, the risk should be re-evaluated during the whole stay. In patients with AKI present at admission, sufficiently early recognition and accurate management have shown to have a positive impact on prognosis(37).
There are some limitations to our study. We did not apply urine criteria to define AKI, as data on urine output were not available, and this might decrease the overall incidence of AKI. Limitations in recording data in structured electronic records also relate to etiologies of the syndrome; therefore, future research into the causes of AKI must include other sources of information. Baseline SCr was known for 20% of patients and missing data were estimated using multiple imputations that showed to be more accurate than commonly used surrogate methods. Finally, this study was conducted at a single tertiary medical centre, and the epidemiological profile of AKI of this population may not be generalisable to patients in other centres or lower health care level settings.
Availability of electronic medical records improves our ability to report a comprehensive depiction of AKI occurrence and related outcomes in the real-life setting. Our study provided valuable insights into the understanding of magnitude, complexity and strain of the syndrome in this population. There is a necessity for further efforts to increase the awareness of AKI among clinicians and health care professionals and to induce strategies for effective prevention, recognition and management of the syndrome.