A Non-Interventional Pragmatic Clinical Trial of NLP Models for the Detection of Immune-Related Adverse Events

Summary

Immunotherapies have improved cancer outcomes, but have a unique profile of immune-related adverse events (irAEs). Biorepositories have been established to collect data and samples to help improve our understanding of irAEs, however identifying patients who are eligible for these biorepositories in a timely fashion can be challenging. The goal of this study is to determine if an informatics system for automated irAE detection can improve registration to a prospective irAE biorepository (NCT04242095). The informatics system automatically "reads" participants' electronic health records (EHRs) and determines whether that patient may be experiencing an irAE. The main questions it aims to answer are: * Is it feasible to implement an informatics system for daily analysis of EHR data to detect irAEs? * Does the automated irAE detection system improve registration rates to an irAE biorepository at our institution following an eligible irAE? Researchers will compare standard irAE monitoring to informatics-assisted irAE monitoring to see if using the informatics system increases the registration rate and improves data entry efficiency and quality. Participants will: * Be randomly assigned to standard monitoring or informatics-assisted monitoring for irAE detection. * Have their EHR reviewed to collect demographic, medical, and cancer treatment history. * Be monitored for irAEs through daily automated analysis of their EHR data for up to 12 months or until registration in the biorepository.

Detailed description

This is a single institution, randomized controlled trial of an informatics system for automated immune-related adverse event (irAE) detection from the electronic health records (EHR) for eligibility monitoring for an Alliance prospective irAE biorepository (NCT04242095). This study will enroll approximately 100 patients receiving immunotherapy. Our co-primary endpoints test the hypothesis that informatics-assisted EHR monitoring for biorepository eligibility is feasible and improves registration rates onto the biorepository following an eligible irAE. Secondary endpoints evaluate time to data entry after an irAE, and irAE capture rate based on post-hoc subject matter expert review. This study will take place at Dana-Farber/Brigham Cancer Center. Our co-primary objectives are: 1. To establish the feasibility of implementing the informatics system for daily inferencing on EHR data 2. To assess the ability of automated irAE detection to improve registration to a prospective irAE biorepository Our secondary objectives are: 1. To evaluate whether automated irAE detection reduces the time to data entry in a prospective biorepository data management system 2. To evaluate whether automated irAE detection leads to differences in irAE capture rates of patients with an eligible irAE based on clinician expert review of the EHR Subjects will be randomized to standard irAE monitoring for an eligible registration event (control arm) vs. informatics-assisted irAE monitoring for an eligible registration event (experimental arm). There will be a seamless feasibility run-in of the first 10-20 patients to establish feasibility. These subjects will be included in the full analysis.

Arms & interventions

• OtherInformatics system for eligibility monitoring

  A daily automated data query of eligible subjects will collect EHR data, and our informatics system will be run on these data behind our institution's secure firewall. The daily list of subjects identified by the informatics system as having a new irAE will be accessed by the Alliance irAE biorespository study team at Dana-Farber/Brigham Cancer Center for review.

• OtherStandard eligibility monitoring

  Reporting of an eligible event to the biorepository study staff by the treating clinician and EHR review by the study staff.

Outcome measures

Eligibility criteria