Drug-induced acute kidney injury in intensive care patients Leveraging electronic health records and causal inference to improve medication safety

Acute kidney injury (AKI) is a frequent problem in patients admitted to the intensive care unit (ICU). Drugs are important risk factors for AKI, and pharmacotherapy may be modifiable to prevent drug-induced AKI (DAKI). However, this would require answering questions about causal relationships between drugs and AKI. The aim of this thesis was to contribute to the improvement of medication safety in the ICU with respect to DAKI by answering causal questions.
First, we estimated the associations between 44 commonly administered potentially nephrotoxic drug groups and AKI in the ICU. After adjustment for confounding bias, we found 14 groups to be associated with a higher risk of AKI. Second, for one specific drug – vancomycin – we presented stronger evidence for a causal relationship by applying target trial emulation (TTE). Third, we reviewed previous studies that developed models for the diagnosis or prognosis of adverse drug events (ADEs) in hospitalized patients. This work showed that most studies did not address the crucial aspect of ADEs: a causal relationship between a drug exposure and an adverse event. We proposed two potential routes for future studies to address this important issue. Lastly, we developed a novel causal model-based approach for the diagnosis of DAKI through a combination of TTE and individualized treatment effect estimation with machine learning.
This thesis is relevant for clinicians and researchers as it presents novel etiological insights and directions for future research on DAKI in the ICU.