Acute Kidney Injury: Definition, Risk Prediction, and Management

Acute Kidney Injury Definition

Acute Kidney Injury (AKI) is a clinical syndrome characterized by an abrupt decrease in the glomerular filtration rate (GFR) over a short period of time (hours or days). It can result from multiple etiologies, and its common presentation is an increase in serum levels of nitrogenous waste products, which may or may not be accompanied by a decrease in urine output (in two-thirds of cases). In this document, we use the term acute kidney injury (AKI), in accordance with KDIGO.

The syndromic concept of AKI is well-defined, and its detection is based on increases in serum creatinine (SCr). However, over the years, there has been significant disparity in establishing precise defining criteria. Bellomo et al. proposed the first classification, known as the RIFLE system. Although this classification provided numerous advantages, certain shortcomings became evident over time, such as the underdiagnosis of AKI and the inclusion of estimated GFR (eGFR) in the criteria. Since the estimation of GFR is not valid in acute processes, it was removed from the initial version. For these reasons, the Acute Kidney Injury Network (AKIN) classification was developed in 2007. It is a modification of the RIFLE system that adds an absolute increase in SCr of ≥0.3 mg/dL within a 48-h interval (a cutoff value associated with increased mortality). In 2012, the National Kidney Foundation, through the Kidney Disease: Improving Global Outcomes (KDIGO) workgroup, published the third consensus on the definition and classification of AKI. This new classification merges criteria from both RIFLE and AKIN and is the one currently recommended for use.

All these classifications are functional ones that allow for the diagnosis and severity staging of AKI. Nevertheless, they have significant limitations, the main one being the use of SCr as the parameter to assess renal function. SCr is known to be a suboptimal marker as it can be influenced by numerous factors such as muscle mass—an important aspect in cases like burn patients who experience muscle loss. Furthermore, previous classifications use the baseline SCr value for diagnosis (defined as the highest value in the last 3 months), which is often unknown. KDIGO guidelines suggest using the lowest SCr value during hospitalization or the value corresponding to a GFR ≥ 75 mL/min/1.73 m² for patients with no prior information.

Using SCr as a marker for AKI can lead to a diagnostic delay since its elevation usually occurs after the drop in GFR has taken place. For this reason, efforts are being made to incorporate biomarkers that allow for the detection of tubular damage, which typically precedes the fall in GFR (see section 2.1).

The incidence of AKI in the general population is estimated at 8.3% (community-acquired AKI), increasing to 21.6% in hospitalized patients and up to 57% in critically ill patients. Given the lack of a universal definition, mortality rates vary widely from 30% to 67% in critically ill patients, being higher among those requiring renal replacement therapy (RRT). Approximately 5%–15% of patients who develop AKI require RRT, though literature values vary depending on the clinical scenario. To improve early detection, alert systems can be implemented in the diagnostic process, either as part of hospital information systems or clinical decision-support systems.

Acute Kidney Injury Risk Prediction Injury and Risk Biomarkers

Due to the inherent limitations of SCr, special interest has been focused on new biomarkers that allow for earlier detection of AKI with improved sensitivity and specificity.

Various biomarkers in both blood and urine have been identified that could be useful for early AKI detection, severity stratification, and prognostic assessment. The most promising biomarkers to date include: Kidney Injury Molecule 1 (KIM-1), Neutrophil Gelatinase-Associated Lipocalin (NGAL), Liver-Fatty Acid Binding Protein (L-FABP), Cystatin C, hemojuvelin, N-Acetyl-Glucosaminidase (NAG), netrin-1, gamma-glutamyl transferase (GGT), glutathione S-transferase (GST), Tissue inhibitor of metalloproteinases-2 (TIMP-2), and Insulin-like growth factor-binding protein 7 (IGFBP-7), among others. These molecules participate in different physiological processes altered during AKI, such as renal parenchyma cell death, inflammatory processes, and increased oxidative stress.

Most of these biomarkers are not currently used for AKI diagnosis in routine clinical practice. These biomarkers denote kidney injury, unlike SCr, which denotes a reduction in kidney function. In at-risk patients, the elevation of biomarkers usually occurs before the increase in SCr. Furthermore, cases may occur where biomarkers increase without a corresponding rise in SCr. This situation defines what is known as subclinical AKI. The importance of subclinical AKI is that it may be associated with a worse prognosis. Although KDIGO guidelines have suggested that biomarkers could provide added value to SCr determination—thereby improving early AKI diagnosis and prognosis—they have not yet been implemented in clinical practice. Numerous studies indicate these biomarkers could be used in combination with clinical markers to improve risk prediction. In this regard, since 2012, the U.S. Food and Drug Administration (FDA) has approved the use of the TIMP-2/IGFBP-7 ratio as a biomarker for AKI risk stratification, and it has been recommended in clinical guidelines for cardiac surgery. In addition to TIMP-2 and IGFBP-7, other biomarkers such as Cystatin C, hemojuvelin, NGAL, and KIM-1 might also be useful in estimating the risk of AKI progression. However, this association is not as clear, as contradictory studies exist or suggest that the relevance of these other biomarkers is limited. This disparity in results is related to the high heterogeneity of AKI types, the specific AKI definition used, and the sample size of the studies. Therefore, while great progress has been made in validating and applying new biomarkers for AKI diagnosis and prognosis, further research is needed to improve AKI diagnostics. Because current diagnosis remains based on SCr and not on biomarkers, the term AKI is used rather than “Acute Kidney Lesion”.

E-Alerts and Prediction Models

AKI is one of the most serious and frequent complications in hospitalized patients, entailing high costs and poor global outcomes. Despite its prevalence, impact, and the fact that it is potentially preventable, the onset of AKI often goes unnoticed and is frequently not even recorded in clinical histories or discharge reports. Recognizing the importance of early diagnosis to implement actions aimed at minimizing kidney injury, the introduction of electronic alert systems (e-alerts) within medical record systems has been evaluated in recent years as part of routine clinical practice. The actual effectiveness of these alerts depends on a combination of patient-specific factors, the underlying disease and type of kidney injury developed, the clinical setting, and, above all, the intervention triggered by this early diagnosis.

Perhaps the most representative reports on the utility of these e-alerts are those from the United Kingdom. At the beginning of the last decade, an AKI electronic alert system was mandatorily implemented in inpatient care within the National Health Service (NHS)-UK, which was subsequently extended to Primary Care in a planned manner. These e-alerts are based on the algorithm described in the National Institute for Health and Care Excellence (NICE) guidelines and define the severity of AKI using the KDIGO classification. Laboratory systems automatically calculate the AKI stage based on SCr levels, using the patient's results from the previous year as the baseline creatinine.

The e-alert itself, in addition to identifying the patient, directs the physician toward a decision-support manual. Under this philosophy, each institution opted for different design and implementation modalities for their e-alerts, resulting in significant variability in both the reporting method (email, pop-up windows in the patient's electronic health record, mobile text messages, etc.) and the various clinical action algorithms. Numerous publications have analyzed the impact following their implementation in the United Kingdom, yielding contradictory results. Consequently, in the absence of careful studies regarding both their efficacy and potential adverse effects, various opinions have emerged suggesting the need to moderate the enthusiasm for this type of e-alert.

In most published studies, no significant impact has been demonstrated regarding short-term mortality or the requirement for dialysis. Nonetheless, positive results have been found concerning a decrease in the prevalence of hospital-acquired AKI, improvements in clinical management, and a reduction in the mean length of hospital stay. Results from recent meta-analyses show a high degree of variability in the design of e-alert systems and indicate that the e-alert per se does not improve outcomes unless it is associated with complementary care measures. In such cases, improvements translate into a shorter time for the modification of nephrotoxic drugs, a more rational application of fluid therapy, diuretics, or vasopressors, and more frequent nephrology consultations; this reduces the rate of severe AKI and increases the proportion of patients who achieve renal function recovery. Recently, results from a large hospital in Birmingham (UK) have demonstrated that, after two years of follow-up post-implementation of e-alerts, the progression of AKI has decreased, and therefore, it is likely that long-term survival will improve. Readmissions to emergency departments following hospital discharge also decreased, a fact the authors attribute to the reduced use of nephrotoxic agents in these patients. The authors emphasize that even minimal changes in patient management can have significant repercussions on long-term outcomes.

More recently, the NHS has implemented AKI alerts in Primary Care. These experiences have reported an increase in community AKI detection, improvements in follow-up, shorter times to hospital admission, and higher rates of renal injury recovery. The pros and cons of using e-alerts are summarized in Table 1.

SCr: serum creatinine; AKI: acute kidney injury.

AKI Prediction Models Using Artificial Intelligence

At the consensus conference of the Acute Disease Quality Initiative (ADQI) in 2015, AKI was recognized as an ideal disease state for the application of machine learning and big data. Since then, artificial intelligence has been used to develop AKI risk scales that allow for the implementation of measures in patients at risk of AKI or with early-stage kidney injury. These Machine Learning (ML) models automatically include many variables and allow for the identification of patients at higher risk of adverse outcomes and the discrimination of different kidney injury subgroups. Models published in different AKI settings lack external validation; therefore, the results are not generalizable to other populations. Furthermore, they predict the risk of AKI at a single point in time rather than continuously. On the other hand, there is significant variability among the analyzed cohorts, which, in most cases, are retrospective. Consequently, while the predictive potential of machine learning algorithms is recognized, they still require improvement. Additionally, these models have demonstrated the ability to predict AKI, but not to prevent its occurrence.

AKI Bundles Fluid Therapy

One of the key points in AKI management is maintaining an appropriate hydration status. Currently, a wide variety of solutions are available for volume replacement. However, few studies exist, most of them conducted in the critically ill patient setting, that allow for an evaluation of which of these solutions is the most suitable.

The objective of fluid therapy in critically ill patients, and especially in those with septic shock, is to increase preload in order to augment cardiac output. The challenge lies in maintaining adequate t