Health disparities are defined by NIH as differences in the incidence, prevalence, mortality, and burden of diseases or health conditions between specific population groups in the US. These disparities are often associated with demographic, social or economic factors such as income, race, gender, and geography (e.g. rural vs. urban), among others. Iowa is not exempted from these disparities. The latest report released by the Iowa Cancer Registry shows that compared to all other racial/ethnic groups, the Black population in Iowa has the highest mortality rate in every single major cause of death, including cancer. At the same time, machine learning (ML) and artificial intelligence (AI) algorithms are increasingly used in clinical care to improve diagnosis, treatment selection, and prognosis of numerous diseases including cancer. Several recent studies have reported that AI systems can discriminate and create unequal outcomes in different population groups. Because AI/ML models learn from historically collected data, human and structural biases present in the data can be perpetuated and even exacerbated as the models are applied to clinical care. The goal of this initiative is to identify and quantify current biases in AI/ML healthcare models and to develop interventions to correct and reduce these biases.
This jumpstart proposal seeks to: 1) assess the bias of existing AI models and compare different sociodemographic groups; 2) generate fair AI models under different metrics and compare performance with existing models; and 3) generate hypothesis-driven research projects to address health disparities targeting NIH/NSF grant programs while engaging the campus research community.
- Thomas Casavant (Biomedical Engineering, Engineering)
- Mary Charlton (Epidemiology, CPH)
- John Buatti (Radiation Oncology, CCOM)
Bringing Critical Race Theory into Public Health: An Example in Cancer Care Delivery Research
On Wednesday, October 20, 2021, Dr. Mya Roberson presented a virtual seminar to discuss her work about racial disparities in breast cancer, as well as some of her work on how to support early career Black scholars.
Kickoff Session: Addressing Health Disparities and Biases in Cancer Outcomes Using AI and Machine Learning Tools
When: Friday, September 17, 2021
Recording: Listen to the recording of the kickoff session here.
Stay in Touch
Subscribe to our email list to stay up-to-date on upcoming events and information about our project.
Funding results from Pivot*
*must be connected to UI network or VPN