College of Nursing
60 Developing an Algorithm to Detect irAE-related Treatment Discontinuation
Megan Rose; Djin Tay; and Malek Alnajar
Faculty Mentor: Djin Tay (Nursing, University of Utah)
Background
Immune-related adverse events (irAEs) are significant complications associated with immune checkpoint inhibitors (ICIs) used in cancer treatment. Depending on the severity of the irAE, some ICIs must be discontinued altogether to ensure patient safety and to manage adverse effects effectively. Using real-world data from the Flatiron Health Database, this study aims to develop an algorithm that identifies treatment discontinuations related to irAEs in cancer patients.
Methods
This study used national data from the electronic health record derived Flatiron Health Database. Data from patients diagnosed with advanced melanoma, lung, bladder, colorectal, and head and neck cancers were analyzed using R software. Algorithm 1 included patients who were diagnosed with an irAE. Algorithm 2 expanded upon this by adding subsequent steroid use. Algorithm 3 incorporated cases where patients received no subsequent ICI. Algorithm 4 combined irAE diagnoses, subsequent steroid use, and no subsequent ICI. Diagnostic codes were used to identify irAEs. Steroids assessed included Dexamethasone, Methylprednisolone, Prednisone, Hydrocortisone, Budesonide, Triamcinolone, Fludrocortisone, Clobetasol, Betamethasone, Fluorometholone, Prednisolone, Mometasone Furoate, Beclomethasone, Loteprednol, Difluprednate, Fluocinolone Acetonide, Oxandrolone, and various unnamed clinical study drugs. Cross tabulations were produced to compare the distribution of patients in each algorithm category with patients for whom treatment cancellation was flagged and patients who died within 30 days of their first irAE diagnosis.
Results
With all cancers combined, there were 104,615 total patients. Algorithm 1 identified 34.28% of patients (n = 61,181), Algorithm 2 identified 37.39% of patients (n = 18,081), Algorithm 3 identified 42.73% of patients (n = 20,024), and Algorithm 4 identified 43.54% of patients (n = 5,329) with treatment cancellations. Algorithm 1 identified 5.67% of patients, Algorithm 2 identified 3.69% of patients, Algorithm 3 identified 1.86% of patients, and Algorithm 4 identified 1.01% who had mortality rates within 30 days of the first irAE.
Discussion
These findings highlight the utility of using diagnostic codes and medication administration data to identify treatment discontinuations related to irAEs. Findings support that combining diagnostic codes associated with irAEs with subsequent steroid use and ICI treatment data identifies approximately 4 in 10 patients who had treatment cancellations. Future directions include identifying patient and practice characteristics.