Acute kidney injury and in-hospital mortality in COVID-19: a causal analysis using directed acyclic graphs.

Abstract

Background: Acute kidney injury (AKI) is a common complication in patients affected by COVID-19 and has been strongly associated with increased mortality. However, its independent contribution remains debated. This study aimed to evaluate the independent association using a directed acyclic graph-based approach.

Methods: Retrospective, multicenter cohort study in Medellín, Colombia, from June 2020 to April 2022. AKI was defined according to KDIGO criteria. A directed acyclic graph was constructed to map the hypothesized causal relationship between AKI and mortality, integrating evidence from a comprehensive literature review and expert´s consensus. A Poisson regression model with robust variance was applied to estimate adjusted incidence rate ratios (IRRs) for mortality.

Results: A total of 1722 patients were included, of whom 30.7% developed AKI. Mortality was higher among patients with AKI (58.1% vs. 19.6%). A directed acyclic graph was used to identify a minimal sufficient adjustment set for confounding control. After adjustment, the IRRs for in-hospital mortality were 1.25 (95% CI: 1.09-1.43) for stage 1, 1.62 (95% CI: 1.35-1.93) for stage 2, and 1.64 (95% CI: 1.46-1.85) for stage 3.

Conclusions: AKI is independently and significantly associated with an increased risk of mortality in COVID-19 patients. This study makes a novel contribution by applying directed acyclic graphs to enhance causal inference. Directed acyclic graphs provide a rigorous framework for identifying true confounders and avoiding inappropriate adjustment for mediators, thereby reducing bias and improving the validity of causal estimates. In clinical settings where randomized controlled trials are not feasible, the use of directed acyclic graphs represents a robust alternative for exploring causal relationships.