Key result
GPT-4 achieves optimal depression classification from patient transcripts using complex prompts at low temperature settings.
Why the study?
Clinical depression diagnosis is time-consuming and relies heavily on expert professionals amidst a critical shortage of specialized personnel, motivating the exploration of early detection tools.
Can GPT-4 accurately classify patient interviews into depressed and not depressed categories based on transcript analysis?
Population
Patient interview transcripts
Comparison
Prompt complexity and varied temperature settings in GPT-4
Design
Pilot study
Authors
Loading...
May support transcript-based depression screening; leaves open prospective validation before clinical adoption.
Can GPT-4 accurately classify patient interviews into depressed and not depressed categories based on transcript analysis?
GPT-4 shows promise for clinical depression assessment from transcripts, but requires careful calibration of prompt complexity and temperature settings for consistent performance.
Lorenzoni et al. (2024) studied Clinical depression. GPT-4 vs. Simple vs complex prompts across different temperatures was evaluated on Accuracy and F1-Score for binary classification of depression. GPT-4 classification of depression from patient transcripts achieved optimal accuracy and F1-Score using complex prompts at lower temperature settings (0.0-0.2).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: