Over the past decades, a significant increase in the incidence of acute kidney injury (AKI) has been witnessed, with the number of cases, including those of AKI requiring kidney replacement therapy, more than doubling among hospitalized patients.1 Simultaneously, in high-income countries, the spectrum of the disease has changed due to spectacular advances in medicine, with AKI becoming a disease of the elderly and multimorbid individuals, who are much more vulnerable to acute insults. Nowadays, AKI develops in up to 20% of hospitalized patients in high-income countries, which constitutes approximately 50% of adult patients in intensive care.2

In recent years, critically ill patients with COVID-19 have emerged as a new population with high incidence of AKI, with 15% to 30% of cases requiring kidney replacement therapy.3,4 In a meta-analysis of 15 017 patients hospitalized for COVID-19, mainly from China and the United States, the prevalence of AKI was as high as 11.6%,5 while in another meta-analysis of 49 692 patients with COVID-19 from several countries, the incidence of AKI ranged from 5.4% in individuals with nonsevere COVID-19 to 22.1% in those with a severe disease course.6 The reported mortality rate among patients with COVID-19–related AKI was 54.8%.7

Although similar incidence of AKI has been reported for patients hospitalized for COVID-19 and those with other respiratory viral illnesses, COVID-19 has been associated with more severe AKI, higher incidence of chronic kidney disease (CKD), and higher mortality.8 In a recent multicenter cohort study of 9624 hospitalized patients with AKI, Aklilu et al9 indicated that individuals with COVID-19–related AKI who survived hospitalization experienced lower rates of major adverse kidney events (defined as a composite of worsened kidney function [estimated glomerular filtration rate decline by ≥25%], or the International Classification of Diseases, 10th revision code for end-stage kidney disease, or death), long-term kidney function decline, and mortality, as compared with patients suffering from AKI associated with other illnesses. In a multinational prospective cohort study conducted among hospitalized adults with confirmed COVID-19 consecutively admitted to 40 hospitals across 23 countries, the prevalence of AKI was higher in the African than in the non-African cohort (12.7% vs 7.2%, respectively), and AKI was strongly associated with overall mortality in the African cohort.10 In a retrospective study involving 1521 hospitalized COVID-19 patients in Malaysia, AKI was a risk predictor of prolonged hospitalization and higher mortality rates.11 In a recent retrospective case-control study, Klimkiewicz et al12 investigated the clinical significance of AKI, as well as previous CKD, among hospitalized patients with severe COVID-19. AKI was diagnosed in 34.4% of the patients, and was associated with more frequent transfers to an intensive care unit (ICU), increased markers of severe disease, and higher morbidity. It was also identified as a risk factor for death. In both univariable and gradual backward multivariable logistic regression analyses, the authors found that hospitalization in the ICU was the only variable independently associated with the occurrence of AKI. After elimination of the ICU stay, serum albumin and sodium levels independently correlated with AKI. Lastly, leukocyte count and serum triglyceride level were associated with the occurrence of AKI. Two recent studies from the United Kingdom (UK) explored the relationship between in-hospital mortality and AKI.13,14 In a retrospective cohort study analyzing 749 844 admissions in 337 029 adult patients with laboratory-confirmed COVID-19 in the UK, the incidence of AKI was 30.3%.13 In a multivariable logistic regression analysis, older age, mixed and black ethnicity, emergency admission, transfers from other providers, ICU care, and increasing burden of comorbidities were associated with greater odds of developing AKI. The above factors, along with the occurrence of AKI, were also associated with greater likelihood of in-hospital death. In the second study, involving 576 patients with AKI, increasingly complex analyses, from logistic regressions to competing-risk and multistate models, have provided insights into AKI progression dynamics associated with polymerase chain reaction–confirmed COVID-19 acquisition and death.14 Rapid progression in severity, prolonged hospital stay, and high mortality among patients with AKI requiring kidney replacement therapy were significantly exacerbated by COVID-19.

In their recent study published in Polish Archives of Internal Medicine, Krzanowska et al15 proposed a predictive model for the occurrence of AKI in hospitalized patients with COVID-19. Based on data extracted from electronic medical records of 5806 patients admitted between March 2020 and January 2022 to the emergency department at the University Hospital in Kraków and hospitalized for at least 7 days, a total of 4630 patients were included in the final analysis. The authors randomly split the data into a training (n = 3462) and test (n = 1168) cohorts. They used the random forest algorithm with the “ranger” package selected as the model engine. Nested 3-fold cross-validation was further conducted to assess model performance during model development, with final validation on the test set. During validation of the model-building procedure, they employed an out-of-time sample. Finally, using sophisticated statistical analysis with machine learning, they developed a model called CRACoV-AKI, which aims to serve as a robust tool capable of forecasting AKI. The tool is electronically accessible at https://kalkulator-covid.su.krakow.pl/kalkulator-ryzyka. In the model, the need for respiratory support, CKD, and procalcitonin levels were among the most important variables in permutation tests. In addition, the authors underlined that pre-existing kidney disease was the most important predictor of AKI in COVID-19 patients. However, it should be stressed that in previous papers different approaches were taken, and the focus was mainly on the predictors of mortality in patients with AKI.9-14

This particular study aimed to develop a tool to predict AKI, which is an important complication of COVID-19. Successful machine learning–based approaches to predict AKI may be valuable tools in other settings (non–COVID-19 forms of AKI). As AKI is highly heterogenous and imposes a great financial burden on the health care system, identification of high-risk patients may positively impact clinical practice. Early diagnosis and treatment / elimination of the underlying causes, such as volume depletion, hypotension, use of selected drugs, or urinary tract obstruction, may reduce the risk of any potential insults. This may minimize additional injury and enable timely implementation of supportive measures and / or kidney replacement therapy to maintain optimal fluid, acid-base, and electrolyte balance, resulting in improved kidney outcomes and survival. As we suggested previously, AKI survivors, especially those with hospital-acquired disease, should be closely followed, since they are at a substantial risk of relapse and subsequent development of end-stage CKD as well as other adverse outcomes, including hypertension and cardiovascular disease.1