Wenxin Xu: AI models can accurately impute performance status from clinical notes

Wenxin Xu, Genitourinary oncologist and clinical cancer researcher at Dana-Farber Cancer Institute and Assistant Professor of Medicine at Harvard Medical School, shared on their X/Twitter:

”Out now in JCO Clinical Cancer Informatics! AI models can accurately impute performance status from clinical notes (AUROC 0.95), even when it’s not explicitly documented. Read further.

During training, our models ‘learned’ context clues associated with poor PS. Imputed poor performance status was strongly associated with mortality (HR 11.9) in cancer patients.

Performance status is used by oncologists to determine eligibility for treatment and clinical trials. The ability to predict performance status from electronic records might help to expedite cancer research using large datasets.”

Source: Wenxin Xu/X