Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1762
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-18T01:09:18.279914Z
status: published
lens: g_space
sector: health
headline: Extraction of Pain Severity and Functional Interference From Clinical Narratives Using Domain-Informed Large Language Models: Protocol for a Development and Validation Study
dek: Background Chronic pain is a leading cause of disability and requires multidimensional assessment of pain intensity and functioning, yet electronic health records rarely capture these measures systematically. By contrast, surveys collecting patient-reported outcomes can assess pain over multiple dimensions but remain resource-intensive and difficult to scale for continuous population-level monitoring. Objective The objective of this study is to develop and validate a domain-informed natural language processing f…
gain_title: Language models may let health systems track pain severity and functional interference from existing clinical notes instead of relying on resource-intensive patient surveys, though this is not yet validated.
problem_title: (none)
trace_subject: (none)
gain_reading: Language models may let health systems track pain severity and functional interference from existing clinical notes instead of relying on resource-intensive patient surveys, though this is not yet validated.
gain_evidence: serve as a reliable, scalable surrogate for resource-intensive patient-reported surveys | enable health care systems to continuously monitor reports of pain-related functional interference
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers within the Veterans Health Administration are building and testing a natural language processing framework that reads unstructured clinical notes to estimate pain severity and two kinds of pain interference, with enjoyment of life and with general activities. The cohort covers 3725 Veterans with chronic musculoskeletal pain who began complementary and integrative health therapies.

The work matters because electronic records rarely capture pain and functioning systematically, while surveys are hard to scale. As of the July 2026 status report, the main validation is unfinished: expert review is in progress and final results are expected in the first half of 2027, so the approach's reliability is unproven.
limitation: This is a protocol with expert validation still in progress, and final results from the knowledge distillation and validation stages are not expected until the first half of 2027, so accuracy is not yet established.
tag: Evidence-backed gain
key_points: The cohort is 3725 Veterans with chronic musculoskeletal pain across 18 Whole Health Flagship sites from 2021 to 2023. | Survey-based patient-reported outcomes serve as the benchmark for judging the text-derived measures. | A developmental LLM screening of 6747 notes for 260 patients found at least one pain domain in 75% of notes and 99% of patients. | Six hundred candidate encounters, 200 per domain, were chosen for expert annotation as the ground truth sample. | The best of two candidate LLMs will generate annotations to train a lightweight model.
rundown: The framework targets three domains drawn from established psychometric instruments: pain severity, interference with enjoyment of life, and interference with general activities. A seed lexicon and annotation guidelines were developed with subject matter expert input to distinguish them.

Validation has three stages. The first checks how completely clinical narratives document pain interference against expert-annotated references. The second tests accuracy of the LLM annotator and the fine-tuned lightweight model at note-level classification and span-level localization. The third compares the lightweight model's output with patient-reported pain, enjoyment, and general activity scores across several time windows.

The authors describe it as the first study to validate clinical text-based pain outcome extraction against patient-reported outcomes in a nationwide longitudinal cohort. They say the aim is to support more holistic, patient-centered pain management at scale.
sources:
- peer_reviewed | JMIR Research Protocols | https://doi.org/10.2196/96346 | 2026-10-07
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

c313524431241545a374cca70de07e0e73c275d7cf2f0d2d301f0de7718b98ee

previous

0000000000000000000000000000000000000000000000000000000000000000