STEMJazz Talklet with Kenya Andrews
Abstract
Kenya recently gave a STEMJazz talk highlighting new research on how the language in medical records can influence patient care. While we often hope medical documentation is purely objective, the way a doctor describes a patient’s symptoms can carry subtle, unintended meanings.
The core of the issue lies in word choice. For example, writing that a patient “reported” shooting pains sounds different than writing that they “complained of” shooting pains. The latter can carry a slight tone of skepticism. Since these notes are read by various doctors and nurses throughout a person’s treatment, biased or judgmental language can inadvertently shift a patient’s care plan, sometimes leading to less effective treatment or misdiagnosis.
Kenya’s research focuses on what is known as algorithmic justice. She uses large language models to analyze thousands of medical records, looking for linguistic patterns that correlate with a patient’s race, age, or gender. Her work involves organizing these findings into causal graphs to see how specific types of language—like stigmatizing or negative descriptors—connect to different demographic groups.
The ultimate goal of this work is to improve health outcomes by identifying potentially harmful language in real-time. By using AI to flag these patterns, the system could nudge writers toward more neutral, fact-based language that ensures parity in care.
This research is still in its early stages. While using AI to identify patterns in data is one step, the next phase involves the complex task of testing how doctors respond to these changes in practice. Navigating medical privacy laws and designing experiments to see if these “nudges” actually lead to better clinical decisions will be essential for the work to have a broad impact.
If you are interested in the technical details, you can read the full paper here: https://aclanthology.org/2023.clinicalnlp-1.39.pdf
Thank you to Kenya for sharing such an important perspective on how we can use data science to make healthcare more equitable!
