Mining patient feedback and complaints for themes
Most health services in Australia are sitting on several years of patient comments that nobody has read from start to finish. The Likert scores get tabulated and put in the quarterly pack. The free text gets skimmed by whoever is assembling that pack, and the rest is archived. This post is part of our Practical AI in Health series, and if you are looking for a first AI project that is cheap,…
Most Australian health services have amassed years of patient comments without fully analyzing their content. Existing survey data provides a wealth of free-text feedback that can be processed using artificial intelligence. The 12-item AHPEQS question set collected by Australian hospitals includes a free-text field where patients can leave comments.
An implementation study analyzed 86,180 surveys from 36 private hospitals over 18 months, which produced 86 adult and 35 pediatric topics from patient comments. Topic modeling on complaints revealed themes such as long wait times, staff attitudes, and appointment booking issues. Closed survey items can only measure what patients thought to ask, while free-text comments provide more comprehensive insights.
The Commonwealth's Safe and Responsible AI in Healthcare legislation review found that Australia's current regulatory system is not fit for purpose and called for mandatory guardrails for high-risk settings. Analyzing patient feedback about waiting times and communication issues does not fall into this high-risk category. The Commonwealth requires hospitals to seek patient and caregiver feedback regularly and report complaints to their governing body.
Analyzing patient feedback can help meet these requirements and provide evidence for improving safety and quality. The Healthcare Complaints Analysis Tool (Gillespie and Reader, BMJ Qual Saf 2016) can be used as a starting point for analyzing complaints. It classifies complaints into seven problem categories: quality, safety, environment, institutional processes, listening, communication, and respect and patient rights.
A taxonomy such as HCAT can improve the accuracy of theme mining projects and allow for comparison with other sites and over time. The architecture for a theme mining project involves ingesting data from surveys, complaints registers, and feedback inboxes into a staging store, de-identifying the data, classifying comments against HCAT labels, running unsupervised clustering for themes not captured by the taxonomy, and aggregating the results into theme counts by site, ward, and quarter. Low-confidence and high-severity items should be routed to a human review queue.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
