Mining patient feedback and complaints for themes

Short summary

A detailed guide to using AI for theme-mining patient feedback and complaints in Australian healthcare, leveraging existing survey corpora (AHPEQS) and the Healthcare Complaints Analysis Tool (HCAT) taxonomy. The article argues for supervised classification over unsupervised clustering, noting complaints average 1.94 problems each requiring multi-label models. A 2025 study compared GPT-3.5, GPT-4o mini, and Claude 3.5 Sonnet for HCAT coding. The project satisfies NSQHS accreditation requirements as a by-product while staying outside high-risk regulatory boundaries.

• •Patient feedback free-text is an untapped, low-risk AI project with accreditation payoffs
• •Use HCAT taxonomy for supervised multi-label classification, not unsupervised clustering
• •GPT-4o mini and Claude 3.5 Sonnet benchmarked against human-coded complaints in 2025 study

Generated with AI, which can make mistakes.