(36) Validation of a Claims-Operable Immune-Related Adverse Event (irAE) Case Definition in a Real-World Lung Cancer Population

Introduction: Seventy percent of immune checkpoint inhibitor (ICI) users with advanced non small-cell lung cancer experience some degree of immune-related adverse events (IRAEs), including ~10% with life-threatening toxicity. Predictors of increased risk are the subject of active inquiry. Use of real-world data to examine irAE incidence and risk factors is limited by measurement error when it comes to irAEs. The objective of this study is to conduct validation of two claims-operable case definitions for irAE in a Johns Hopkins Lung Cancer Registry.

Hypothesis: We hypothesize that our updated claims-operable algorithm will have increased positive and negative predictive value relative to the published claims-operable case definition, when validated against gold standard chart review. 

Methods:

Data source: The Johns Hopkins Thoracic Oncology Precision Medicine Center of Excellence (PMCOE) Registry, a primarily electronic health record (EHR)-derived database of Johns Hopkins cancer patients.

Study population: Thoracic Oncology PMCOE Registry patients concurrently enrolled in the Johns Hopkins Lung Immunotherapy Database Registry. The registry includes adults with lung cancer eligible to receive immunotherapy in a clinical trial setting at six Johns Hopkins hospitals. Patients had irAE status verified through chart review by Johns Hopkins clinicians.

Algorithm development: We will convert a published claims-operable irAE algorithm into a format that can be used in the PMCOE data using data fields common to the PMCOE and insurance claims data. The published criteria are (1) an inpatient admission with (2) an associated diagnosis of pneumonitis or myositis along with (3) initiation of ASCO guideline-recommended irAE treatment. We will augment this algorithm by requiring (3*) at least one refill or second administration of the irAE treatment and (4) no evidence of ICI utilization subsequent to suspected irAE.  We will use data fields from inpatient and emergency department encounters, including ICD-10 codes, HCPCS/CPT codes, and RxNorm and NDC codes.

Validation: We will classify the irAE status of Lung Immunotherapy Database registry participants using the original and augmented algorithms, respectively, and compare the algorithms’ performance against the irAE status ascertained through chart review. We will quantify the test characteristics and explore performance of the algorithms in relevant patient subgroups.

Results: At present the still-active Lung Immunotherapy Registry includes ~850 individuals with a primary diagnosis of lung cancer. ~300 of the patients have had chart review-confirmed irAE, and ~550 have not. Results of the validation study are pending.

Conclusion: The validation of claims-operable irAE case definitions is key to minimize as well as to inform methods to adjust for measurement error in claims-based studies of irAE.