🤖 AI Summary
This study addresses the detection of implicit stigmatizing and biased language in psychiatric clinical notes—expressions that may erode patient trust, exacerbate clinician–patient cognitive biases, and perpetuate healthcare inequities. We propose the first dual-perspective sentiment classification framework grounded in reader stance (clinician vs. patient/public), leveraging prompt engineering and in-context learning to adapt GPT-3.5, Llama 2, and Mistral for fine-grained bias detection on real-world clinical notes. Results reveal systematic perspective alignment biases across models: GPT-3.5 aligns more closely with clinician perspectives, whereas Mistral better captures non-clinician viewpoints. All models reliably distinguish divergent affective valences of identical statements across reader groups, achieving AUCs of 0.82–0.89. This work establishes a novel, interpretable, and scalable paradigm for bias assessment in clinical text, advancing fairness-aware natural language processing in mental health informatics.
📝 Abstract
In psychiatry, negative patient descriptions and stigmatizing language can contribute to healthcare disparities in two ways: (1) read by patients they can harm their trust and engagement with the medical center; (2) read by future providers they may negatively influence the future perspective of a patient. By leveraging large language models, this work aims to identify the sentiment expressed in psychiatric clinical notes based on the reader's point of view. Extracting sentences from the Mount Sinai Health System's large and diverse clinical notes, we used prompts and in-context learning to adapt three large language models (GPT-3.5, Llama 2, Mistral) to classify the sentiment conveyed by the sentences according to the provider or non-provider point of view. Results showed that GPT-3.5 aligns best to provider point of view, whereas Mistral aligns best to non-provider point of view.