Developing AI for Multicultural Cancer Education
Cancer is the second leading cause of death, with almost 2 million people diagnosed with cancer each year. The best way to treat cancer is to detect it early through cancer screening. The Centers for Disease Control has recommended cancer screening guidelines for breast (81%), cervical (93%), colorectal (70%), and lung cancer (8%). Less than 50% of states made the guidelines. The system is failing to reach everyone. A primary reason behind this is the fear of cancer. Patients fear cancer screening for many reasons (e.g., fatalism, the fear of becoming a burden, fear of what they don’t know about treatment or costs). Personalized 1-on-1 education can overcome these barriers, but lacks the ability to scale. Digital tools are built to scale to large populations and could be used to create personalized cancer education experiences.
Required Qualifications:
- Experience using or training Large Language Models
Scheduling Requirements: Fellows will be expected to in-office regularly. Fellows are required to attend all scheduled events and weekly cohort meetings in-person at 50 Willey Hall.
International Health in Context
Dr. Tracy Kugler and Dr. Kathryn Grace
The IPUMS International Historical Geographic Information System (IHGIS) has worked with IPUMS Demographic and Health Surveys (DHS) to provide linkages between IPUMS DHS survey respondents and IHGIS geographic units. These linkages allow DHS users to attach contextual information from population and agricultural censuses in IHGIS to DHS records. Fellows will use these linkages to explore research questions about health in context. For example, potential research questions could address associations between crop production and children’s stunting and wasting or between community-level literacy and education and vaccination rates. Fellows will present their findings in a post for the IPUMS blog series.
Required Qualifications:
- ArcGIS and/or R
Preferred Qualifications*:
- Exposure to census and/or health survey data
Scheduling Requirements: Fellows should be comfortable working with a hybrid team. Some in-office time is appreciated. Fellows are required to attend all scheduled events and weekly cohort meetings in-person at 50 Willey Hall.
Building a Homeless Mortality Predictor Database
Dr. Katie Berry and David Van Riper
People experiencing homelessness in the US typically die nearly 30 years earlier than the average American. Despite evidence that homeless mortality differs by location, we know almost nothing about how the context in which someone experiences homelessness matters for their health and survival. Fellows will work with Dr. Katie Berry and Dave Van Riper to build a Homeless Mortality Predictor Database that incorporates data on place-based factors that may influence homeless deaths such as the scope of local homeless services, social safety net programs, local housing policies, and environmental conditions. Fellows will (1) conduct literature reviews to identify predictors of homelessness and homeless deaths; (2) search for publicly available datasets across a variety of sources, domains, and geographic levels that contain variables that map onto identified constructs; and (3) develop, test, and refine protocols that create measures for a preliminary version of the Homeless Mortality Predictor Database.
Required Qualifications:
- Experience searching for and reviewing research articles
Preferred Qualifications*:
- Experience with a statistical package (R or Stata)
- Interest in housing policy, homelessness, mortality, and/or social determinants of health
Scheduling Requirements: Fellows will be required to be in the office 2 days per week (likely Mondays and Thursdays) for in-person meetings and collaborative work. Additional work hours may be completed remotely. Fellows are required to attend all scheduled events and weekly cohort meetings in-person at 50 Willey Hall.
Building and Assessing 2020 Census Data for Optimized Block Groups
Jonathan Schroeder and Dr. Steve Manson
The U.S. Census Bureau used a new privacy-protection system for 2020 census data that injects noise into most statistics. The system involved a unique set of geographic units called optimized block groups (OBGs), which received the lowest level of noise among all small areas. As a result, OBGs may be the smallest areas for which analysts could get reliable 2020 census data, but unfortunately, the Census Bureau has never published data for OBGs. In this project, fellows will help IPUMS NHGIS (National Historical Geographic Information System, nhgis.org) to construct 2020 census summary tables for OBGs as well as GIS files that represent OBG boundaries. Fellows will also investigate the reliability of OBG data and prepare a blog post that introduces OBGs and demonstrates their utility through an example application.
Preferred Qualifications*:
- Experience with programming for data processing, preferably in R
- GIS or mapping skills
- Familiarity with census data
Scheduling Requirements: Fellows should be able to meet in person for project meetings through the first few weeks. Fellows are required to attend all scheduled events and weekly cohort meetings in-person at 50 Willey Hall.
*We encourage people to apply even if they don’t meet any of the preferred qualifications. If you meet any of the preferred qualifications, please clearly indicate this in your application materials.