STEMJazz Talklet with Rebecca Hubbard
Abstract
Rebecca gave our last STEMJazz Talklet of the semester yesterday and the topic was health data. Rebecca is that type of scholar who, as soon as she speaks, you sense a capacious depth of knowledge and understanding of her field. It’s a sort of ease with the topics, crystal clear explanations and thoughtful answers – perhaps especially when rummaging through that knowledge owing to a question, she sees a gap in the field understanding.
In any case, Health Data. There were two big take-aways:
We’d like to think that all the information collected by us is curated by magic and neatly stored away for each reference and retrieval for the greater good. The reality is more complicated with all manner of institutional boundaries and landmines (medical tests, notes, etc.) strewn all over the place. Part of Rebecca’s job is to piece together these large (and often labile) datasets to extract information about healthcare delivery, disease and making things better for everyone.
The healthcare records (HCR) industry is not driven by health outcomes and patient care. It is driven by billables (what insurance companies can bill and how). So taking the data at face value without understanding the undercurrent can lead to “completely wrong with high precision” conclusions. The most striking example was a seeming uptick in Alzheimer diagnoses – a sudden epidemic – that was driven by billing practices (you could bill if it were Alzheimer’s but couldn’t if not). So the context of data collection (side/structural information in my field) is vitally important for interpretation.
And there are all manner of other untanglings, disambiguations and massaging that needs to be done. What’s particularly exciting is that Rebecca’s area (with her expert preprocessing) is ready for ML and LLMs – and perhaps that CAN provide a pathway for better healthcare outcomes.
If you want to learn more about her work, please see Rebecca’s Apple Podcast: https://podcasts.apple.com/us/podcast/messy-data-real-answers/id1561779028?i=1000736271286
Thank you for a phenomenal talk Rebecca!
