ELSIcon2022 • Paper • May 27, 2022
Daniel Chavez-Yenter, Xiangying Chu, Yuyu Chen, Melody S. Goodman, Kimberly Kaphingst
Health system-wide approaches are needed to identify patients eligible for cancer genetic services prior to diagnosis with cancer. We developed an algorithm to identify such patients based on cancer family history information available in the electronic health record (EHR). However, underlying disparities in the family history information available in the EHR could lead to bias in algorithm outputs due to missing data. We examined the availability and comprehensiveness of family history information in EHRs among 144,484 primary care patients in the University of Utah Health system and 645,264 primary care patients in the New York University Langone Health system. All included patients were between the ages of 25-60 and had had a primary care appointment within the last three years. We found significant differences in both availability of any family history information and availability of cancer family history by race, ethnicity, gender, and language preference in both health systems, with less availability for minority racial and ethnic groups, Latinx patients, men, and Spanish-speaking patients. Even when cancer family history information was available, we found that the records were less comprehensive, meaning that the records were less likely to have specific information on type of relative affected, age at diagnosis, and/or specific type of cancer. These findings indicate that patients from certain demographic groups are less likely to be identified using algorithms that require complete family history information in the EHR. Therefore, effective intervention approaches to address disparities in the collection of family history information are critically needed.