By Jonathan D. Gold, MD MHA MSc FAMIA FHIMSS
March 11, 2024
Challenge yourself:
Why are unstructured medical notes often overlooked yet vital for comprehensive patient care?
How does natural language processing aid in extracting insights from unstructured medical notes?
What role does temporal information play in organizing narratives for health timelines?
How do health timelines derived from natural language processing assist in clinical decision support and research?
Patient problem lists, lab results, medications, and procedures are just some examples of data that need to be managed and analyzed by the provider. One type of data that often gets overlooked when scanning a chart is unstructured medical notes. Text fields appearing throughout the medical record, contain essential narratives describing health events. Unlike formatted data that easily lend themselves to tables and codification, medical notes tend to be freeform and allow the author great independence in describing patients and their health status.
Unstructured medical notes are a treasure trove of valuable information. They contain detailed descriptions of patient symptoms, observations, and treatment plans. However, due to their unstructured nature, extracting meaningful insights from these notes can be daunting. One technique used to harness the power of unstructured medical notes is natural language processing (NLP).
NLP focuses on interpreting human language—organizing text into chunks of classifiable data. By using NLP algorithms, we can analyze unstructured notes and extract key information such as diagnoses, medications, and treatment outcomes. These, in turn, may be codified. NLP engines can also highlight negation and capture the lack of symptoms and signs or list diagnoses that have been ruled out.
Text in medical records narrate key events in patient histories. Those events are constructs that include both elements (like symptoms, risk factors, lab results, procedures, etc.) and time frames. When unstructured text includes time references, it is possible to link elements and provide event descriptions. Including temporal qualities, natural language processing can highlight and organize narratives to assemble a health timeline.
Timelines enable a synoptic view of key health events for a patient, comparison with peers and identification of highly similar patients. Additionally, review of a patient’s entire medical record (including unstructured text) can enable the construction of a longitudinal medical record and a computer-readable comprehensive index of many health events and sequences.
Conclusion
Determining when an event occurred is of premiere importance in contextualizing the history and supporting data analytics and predictive modeling. Natural language processing can identify temporal concepts which, in turn, can provide a means to determine when events occurred. Plotting events on a health timeline is key when comparing health histories and identifying very similar patients to assist in clinical decision support. Other uses for temporal knowledge include precision medicine initiatives, medical and pharma research, and population heath studies.
Comments