Study design:
A retrospective cohort study was conducted in two institutions in Bogotá, DC: The National University Hospital of Colombia and the San Ignacio University Hospital, which are level 4 hospitals that care for adult patients from the city of Bogotá and other regions in Colombia (a population of more than 8 million inhabitants). During the pandemic peaks, hospitals had approximately 150 intensive care beds and 100 hospital care beds for COVID-19. The patients included were hospitalized from March 1, 2020, to March 31, 2021, during the first two pandemic waves in Colombia.
Patient selection:
Patients older than 18 years with SARS-CoV-2 infection confirmed by RT‒PCR or viral antigen were included. Patients who were hospitalized for less than 48 hours in the participating institutions, hospitalized more than 72 hours at the remission hospital, with less than two creatinine measurements during the stay, pregnant women, and patients with stage 5 chronic kidney disease (CKD), solid organ transplant, or obstructive uropathy were excluded.
Sample:
A sample size calculation was performed based on the previous cohorts, with a frequency of 20% AKI associated with COVID-19, a statistical power of 80%, an alpha of 0.05, and odds ratio (OR) of association according to series 1.6. The size was estimated by different methods, with a required sample of 1626 patients.
Data collection:
A clinical research form was created in the REDCap® platform. The data were collected from a review of the medical records, and the data of each variable were extracted and recorded in REDCap®. A data quality review was performed by a researcher. Laboratory information was used to detect patients with at least two values of serum creatinine, with which the identification of AKI was performed. Subsequently, the KDIGO classification was performed (10) and a final review was made by the experts to verify that the definition and classification were correct.
Definitions:
AKI was defined as an increase or decrease in creatinine greater than or equal to 0.3 mg/dl with respect to baseline creatinine according to KDIGO in a period of 48 hours to 7 days during the 28 days following hospital admission (10). Baseline creatinine was considered if it was available in the different computer systems consulted (defined as the measurement in the three months prior to admission) (11,12). Urinary volume was not considered given the low frequency of recording these data in the clinical history. Immunosuppression was defined as a history of solid neoplasia, hematologic neoplasia, HIV infection, systemic lupus erythematosus, or ANCA-associated vasculitis, leukocytosis with a leukocyte count greater than 12 x10 ^ 3/ml and lymphopenia at a lymphocyte count less than 0.8 x10 ^ 3/ml
Ethical aspects
The project was approved by the Research and Ethics Committee (REC) of each of the participating institutions. Approval CEI-2021-05-05 of the National University Hospital and CEI -0631-21 of the San Ignacio University Hospital. Individual informed consent was waived by the approving (REC) in both institutions, as permitted by the fore-mentioned regulation, considering the retrospective nature of the data. Patient information was processed anonymously.
Statistical analysis:
A descriptive analysis of the sociodemographic and clinical characteristics of the study population was performed. Absolute and relative frequency measures were used for categorical variables. For continuous variables, averages and standard deviations were used, or medians and interquartile ranges if the distribution was not normal. Variables with data loss greater than 30% were excluded.
Bivariate analyses were performed to evaluate the relationship between the variables identified in the literature and those potentially related by biological plausibility with the LRA outcome. The differences in proportions and in the quantitative values observed in the groups were evaluated, and the differences were established by means of the χ2 test or the appropriate mean difference tests using Student’s t tests or nonparametric tests. For the analysis of adjusted risk factors, a multivariate logistic regression model was performed, in which the outcome variable was the appearance of AKI. Variables that had a value of p <0.2 and those of clinical relevance were included in the model. In addition, clinically relevant and previously defined interactions were added: 1. previous treatment with antihypertensive drugs and diuretics; 2. PaFI and invasive mechanical ventilation; 3. arterial hypertension and previous treatment with antihypertensives; 4. cardiac failure and previous treatment with antihypertensives, 5. cardiac failure and previous treatment with diuretics. The stepwise technique was used for model selection. The regression model evaluated the presence of atypical values, and the model was adjusted by evaluating the AIC and BIC values of the different identified scenarios. The models were also evaluated by the area under the curve (AUC) of the prediction of the outcome. The model with the greatest parsimony and lowest error was chosen. A p value <0.05 was considered statistically significant. The analysis was performed with the statistical package Stata® (see 15.0, StataCorp, Texas, USA).