The SEER PI meeting was held on March 16-17 in Albuquerque, NM and was sponsored by the New Mexico Tumor Registry. The meeting began with presentations by the NM tumor registry staff highlighting the unique characteristics of the region (e.g., racial/ethnic diversity, 5th largest state but 36th most populated) and summarizing current research projects, many of which are health disparities- related. The remainder of the meeting focused primarily on electronic data capture and alternate methods to complete cancer registration, including natural language processing, deep machine learning, and incorporation of medical billing data.
Among the many thoughtful presentations were some nuggets of information that we found particularly interesting and relevant to our work in HDRP:
- Alissa Greenbaum from University of New Mexico reported preliminary data that rural residents with colorectal cancer are less likely to get KRAS testing and thus benefit from use of targeted therapies.
- Paul Fern from Fred Hutch Cancer Center talked about use of NLP and how computers treat text like a “bag of words” and look for relationships. “The dog ate the bird and it died” is an example of text that neither humans nor computers can easily decipher. Unfortunately, this is a common problem in pathology reports.
- Georgia Tourassi from Oak Ridge National Laboratory presented her work using web crawlers to extract information from obituaries, which often contain information on who, where, when, age, and sometimes cause of death.
We were also reminded that identifying and capturing cancer recurrence and progression is very difficult to do using claims data. The shift to electronic health records might help but it’s unclear when or if it will be realistic to extract these data. Finally, we were encouraged to see progress being made towards implementing a virtual cancer registry that will increase efficiency of access for research. In fact, 47 registries nationally have agreed to participate in Camp Lejeune study to look at the effects of poisoned water consumed for decades by military and families living on base.
Overall, this was a stimulating two days that highlighted the need for electronic data capture and processing of data using NLP to further research in cancer.