Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1887
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-19T10:41:44.781093Z
status: published
lens: g_space
sector: health
headline: Large Language Models in Chronic Obstructive Pulmonary Disease Care and Health Management: A Scoping Review
dek: Background: Chronic obstructive pulmonary disease (COPD) requires sustained patient education, self-management, medication use, pulmonary rehabilitation, symptom monitoring, and guideline-concordant clinical management. Large language models (LLMs) may expand access to conversational health information, but their value depends on more than factual accuracy. Objective: To map empirical evidence on LLMs and LLM-based generative artificial intelligence systems in COPD-related care and health management, summarize a…
gain_title: LLMs often produced clinically acceptable responses in COPD patient education and information tasks, suggesting potential as supplementary, human-supervised tools.
problem_title: (none)
trace_subject: (none)
gain_reading: LLMs often produced clinically acceptable responses in COPD patient education and information tasks, suggesting potential as supplementary, human-supervised tools.
gain_evidence: LLMs often produced clinically acceptable responses
problem_reading: (none)
problem_evidence: (none)
quick_read: A scoping review searched six databases through September 10, 2026, for empirical studies of large language models in COPD care. From 327 records, eight studies met eligibility, mostly examining patient education or health information, with single studies on pulmonary rehabilitation, medication management, and guideline-based clinical management.

The findings matter because COPD relies on sustained self-management, and conversational AI could widen access to information. But most work tests model answers rather than real patient use, so benefits for behavior, clinical outcomes, equity, and safety in routine care remain unproven.
limitation: The evidence base is small (eight studies) and dominated by model-response evaluation, with only two studies involving people with COPD, so real-world effects remain unknown.
tag: Evidence-backed gain
key_points: Of 327 records identified, 34 duplicates were removed and 293 were screened, leaving eight included studies. | Five studies addressed patient education or health information; the others covered rehabilitation, medication management, and guideline-based management. | Accuracy and information quality were the most commonly measured outcomes. | Reported weaknesses included high reading grade levels, outdated guideline references, response variability, and differences across models and task formats. | Authors recommend LLMs as supplementary, human-supervised tools, not autonomous sources of COPD management advice.
rundown: The review followed Joanna Briggs Institute guidance and PRISMA-ScR reporting, searching PubMed/MEDLINE, Embase, Web of Science Core Collection, Scopus, CINAHL, and PsycINFO. Studies were eligible if they evaluated an LLM or LLM-based generative AI system in COPD-related education, information, rehabilitation, medication management, or guideline-based management.

After title and abstract screening excluded 281 records, 12 full-text reports were assessed and four were excluded. The remaining eight studies were synthesized descriptively and thematically, with accuracy and information quality the dominant outcomes.

Only two studies directly involved people with COPD. One found a marked gap between favorable expert ratings and patients' ability to formulate disease-specific queries, suggesting expert-judged accuracy may overstate usefulness.

The authors call for future research on actionability, real-world usability, health literacy, equity, behavioral and clinical outcomes, and safety under routine patient use.
sources:
- peer_reviewed | Journal of Public Health and Preventive Medicine | https://doi.org/10.64904/20261004 | 2026-10-19
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

bbc38dc2c8510a158148dbec02cbc57fb621fa5dc31a6f88264237d3ceee245f

previous

0000000000000000000000000000000000000000000000000000000000000000