Pulmonology

LLMs largely follow GOLD-2025 COPD guidance but performance varies by format

Six large language models answered a 90-question GOLD-2025 COPD dataset generating 3,780 responses over seven days. For binary questions, Kimi K1.5 reached 97.1% accuracy (408/420) and Gemini 3 was lowest at 89.5% (376/420). For open-ended clinical reasoning, Gemini 3 scored highest (96.5%) while DeepSeek-V3.2 trailed (91.8%). Authors conclude LLMs may supplement clinician decision-making under expert supervision but need further validation for generalisability. 1

BREATHE: a validated, flexible education package to optimise breathlessness care in primary care

The BREATHE program combines a webinar, three courses, two podcasts, and five case studies with a seven-point mnemonic validated against 33 guidelines. Thirteen clinicians judged the package acceptable and useful for remembering care elements. Time constraints were a common barrier and respondents recommended flexible, modular delivery. Authors propose BREATHE as a practical tool to help primary care optimise chronic respiratory disease management. 2

SUITS trial protocol: hybrid home-monitoring versus standard clinic visits for pulmonary fibrosis

The SUITS multicenter RCT will randomize people with pulmonary fibrosis to standard 3-monthly clinic visits or hybrid care with weekly home spirometry and pulse-oximetry plus video visits. Primary outcome is change in patient self-management (PAM) over 12 months with regular PROM collection. Hybrid care alternately replaces clinic visits with remote monitoring and video consultations. This protocol tests feasibility and effectiveness of replacing some outpatient visits with home monitoring. 3