A new pilot tool reads through doctors' free-text notes to flag how much pain an osteoarthritis patient is really in.
Researchers built two tools, SPaDe and PLeDO, that mine primary care Electronic Medical Records to classify osteoarthritis patients into mild or moderate-to-severe pain categories. SPaDe works off pain-related language in unstructured chart notes alone, using information extraction, natural language processing, and machine learning. PLeDO layers on structured data too, adding medication records and pain-scale information from the same notes. The team validated both tools against human-labeled gold-standard data to check how well they categorized pain severity.
Chart notes are messy by design: the researchers themselves note that pain expressions are subjective, objective, and shaped by cultural background and demography, which makes consistent scoring genuinely hard. A tool that can parse that noise could help primary care practices flag undertreated pain without new paperwork, feeding straight into diagnosis and treatment decisions.
It's a small pilot, not a deployed system, so the real test is whether it holds up on messier records outside a research dataset.