Grant: DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)

Acute kidney injury (AKI) involves a sudden decrease in kidney function. It impacts nearly 10% of people and is associated with poor outcomes. Our understanding of AKI is limited, and we have not been able to improve kidney outcomes despite better care. The social determinants of health (SDoH) are the conditions where people live that affect health and quality of life. We know that people with lower income and education have more chronic health conditions and worse access to care. There are more cases of AKI in people with chronic disease, but we have limited information on how SDoH impacts the development and the outcomes associated with AKI. Machine learning (ML), a form of artificial intelligence, learns patterns from data to help us better predict outcomes. It has been used for studying AKI, but few approaches have tried to include the SDoH and population diversity in predicting outcomes. This is also important because ML can be resource intensive and so it is important to determine if this technology adds more helpful information than our normal approaches. Thus, we aim to use data from Alberta patients who have been hospitalized and determine which factors from the SDoH are linked with more cases of AKI and worse outcomes using machine learning. Our review of current literature has shown that there is limited information on the connection between AKI and SDoH. Therefore, we will use routine health data sets to collect SDoH, kidney, and other health factors to determine the link between these factors. We will then see if various ML approaches add to the links found using our normal approach alone. In conclusion, AKI is common and linked with poor outcomes. These outcomes are influenced by the SDoH but we do not know to what amount. We hope to explore the link between AKI and SDoH using normal approaches and advance ML approaches to improve patient care and help change current health care policies, which have negatively affected certain patient populations.