Acute Kidney Injury (AKI) is a common complication in hospital inpatients with the annual number of excess inpatient deaths associated with AKI in England estimated to be above 40,000 and the annual cost of AKI-related inpatient care around £1.02 billion. Oncology patients are a known high risk population for developing AKI. The prevalence of AKI in oncology populations is 7.5 to 9.5%, with AKI incidence associated with lower remission rates and higher mortality. To reduce the impact of this condition in cancer patients we aimed to create a model to predict AKI events before occurrence allowing intervention and prevention strategies to be put in place. We trained a Random Forest Model on 597,403 blood test results from 48,865 patients undergoing cancer treatment at The Christie NHS Foundation Trust, Manchester, UK between January 2017 and 3rd May 2020, to identify AKI events upcoming in the next 30 days. From this model, risk levels of very low, low, medium, high and very high were assigned to the risk of an upcoming AKI event. We tested the model and risk assessment through a prospective analysis on patients’ blood tests results collected between 1st June 2020 to 31st August 2020. The trained model gave an area under the receiver operating characteristic curve (ROC AUC) of 88.1% (95% confidence interval 87.8% - 88.3%) when assessing predictions per blood test for AKI occurrences within 30 days. Assigning risk levels and testing the model through prospective validation across the period 1st June to the 31st August identified 61.2% of AKI occurrences with a risk alert of `Medium' or higher, identifying 73.8% of patients with AKI event before at least one AKI occurrence. Our results demonstrate the exciting potential for using routinely collected blood test data to assign AKI risk levels allowing early intervention and preventative strategies to be put in place.

Citation Format: Lauren A. Scanlon, Jorge Barrusio, Alexander Garbett. Developing an agnostic risk prediction model for acute kidney injury in cancer patients using a machine learning algorithm from blood results data [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PR-11.