Can large language models help in the assessment of people with pain?
TL;DR
It is concluded that LLMs should serve as clinical companions rather than substitutes, streamlining qualitative data processing to enhance the clinician-patient interaction and facilitate personalized care.
Abstract
Chronic pain (CP) affects an important proportion of the global population, imposing a significant socio-economic burden and serving as a leading cause of disability. While the biopsychosocial approach is the gold standard for treatment, many patients fail to receive adequate care due to healthcare resource constraints. Effective management requires a comprehensive assessment. However, traditional questionnaires often miss the subjective and multidimensional depth of the pain experience. Differently, written narratives (WNs) empower patients to describe their pain in their own words. Nevertheless, qualitative analysis, such as WN, is often time-consuming for clinicians.This paper explores how Large Language Models (LLMs) can facilitate clinicians in analyzing WNs. Evidence from the AINarratives project and other emerging studies indicates that LLMs can evaluate pain narratives with accuracy and utility comparable to human experts, identifying biopsychosocial themes that correlate with standardized clinical scores.We conclude that LLMs should serve as clinical companions rather than substitutes, streamlining qualitative data processing to enhance the clinician-patient interaction and facilitate personalized care.