Restricted Research - Award List, Note/Discussion Page
Fiscal Year: 2023
67 University of North Texas (141955)
Principal Investigator: Niranjan,Suman
Total Amount of Contract, Award, or Gift (Annual before 2011): $ 837,907
Exceeds $250,000 (Is it flagged?): Yes
Start and End Dates: 9/17/22 - 9/16/23
Restricted Research: YES
Academic Discipline: Logistics & Operations Mgmt
Department, Center, School, or Institute: Ryan College of Business
Title of Contract, Award, or Gift: Evaluating Bias in Predictive and Explainable ML Algorithms Among Older Adults with Cancer
Name of Granting or Contracting Agency/Entity:
University of North Texas Health Science at Forth Worth
CFDA: 93.31
Program Title: none
Note:
There is an urgent need to investigate the feasibility of integrating RWD for AI/ML applications and evaluating the potential bias in AI/ML in cancer outcomes research using RWD. The proposed pilot study will purse the following objectives. 1) Harmonize data across RWDs, create common data elements (CDE), define repeatable and reusable steps for data processing across multiple RWDs using the FAIR (Findability, Accessibility, Interoperability, and Reuse) principles; 2) Test the feasibility of implementing various ML algorithms including deep learning within a federated data ecosystem and 3) Develop pre-trained models to show proof of concept in evaluating bias in data and algorithms. We propose to use a nationwide (Surveillance, Epidemiology, and End Results Program) covering a diverse population with cancer and state-wide cancer registry (Pennsylvania cancer registry) data covering a large rural population. The cancer registries will be linked to publicly available external data to derive SDoH variables at multiple levels (e.g., individual, provider, zip code, county). The proposed study will inform the A-CC to better understand the infrastructure challenges and solutions when harmonizing diverse data hosted by different institutions. The results of this study will inform future research and funding efforts in the feasibility of applying AI/ML bias-audit tools in evaluating “fairness of data and algorithms” to improve cancer care and reduce cancer disparities.
Discussion: No discussion notes