Natural Language Processing Reveals Shift in Palliative Care Documentation in Metastatic Cancer

Strategic Insights -

Researchers from the University of California San Francisco (UCSF) have analyzed inpatient clinical notes to understand how documentation around palliative care has changed in relation to metastatic cancer. Using unsupervised language models, the team found that while metastatic cancer and palliative care terms were initially used together in similar contexts, this relationship weakened over time. The study suggests that text in clinical notes can offer valuable insights into how medical providers document palliative care in patients with advanced malignancies.

Key Takeaways:

  • The study used unsupervised language models to analyze inpatient clinical notes from the University of California, San Francisco system.
  • The researchers found that metastatic cancer and palliative care terms appeared in similar contexts in clinical notes each year, suggesting a close relationship in documentation.
  • However, over time, this relationship weakened, with these terms becoming less commonly used together as measured by cosine similarities.
  • The study suggests that natural language processing can be used to understand trends in clinical documentation and how medical providers document palliative care.
  • The research was funded by the Hellman Foundation, Ucsf Noyce Initiative For Digital Transformation in Computational Biology & Health, and Ucsf Department of Anesthesia Seed Grant.
  • The study's findings have implications for understanding how medical providers document palliative care in patients with advanced malignancies.

Statistics:

  • The study analyzed inpatient clinical notes from 10 years (2009-2019).
  • The researchers used word2vec to model language numerically as vectors, and cosine similarity to measure relational data between vectors.
  • The study found that the relationship between metastatic cancer and palliative care terms weakened by 30% over the 10-year period.
  • The study's sensitivity analysis showed similar trends when the models were retrained just on patients with a diagnosis code for metastatic cancer.

Sources:

  • Lexical associations can characterize clinical documentation trends related to palliative care and metastatic cancer. Scientific Reports, 2025,15(1):1-9.
  • University of California San Francisco (UCSF)
  • Hellman Foundation
  • Ucsf Noyce Initiative For Digital Transformation in Computational Biology & Health
  • Ucsf Department of Anesthesia Seed Grant
  • NewsRx LLC