BMJ Awards 2019: Digital Innovation Team of the YearBMJ 2019; 365 doi: https://doi.org/10.1136/bmj.l1519 (Published 03 April 2019) Cite this as: BMJ 2019;365:l1519
- Jacqui Wise, freelance journalist
- London, UK
Text mining patient feedback
The Friends and Family Test was created to help provide patient feedback to NHS organisations so that they can identify where improvements can be made, and it often includes an option for free text entry. Imperial College Healthcare Trust, alone, gets 20 000 patient comments through this system every month.
This patient feedback is a rich source of information, says Erik Mayer, consultant surgeon at Imperial and project lead. “But the volume of feedback is so great that it can’t be matched with the human resources to read through them all, categorise them, and use them for quality improvement.”
The team created an algorithm using Natural Language Processing to analyse free text patient experience data. They trained the algorithm by taking 6000 free text comments and manually categorising them. The coded dataset was used to develop and test the algorithm.
A series of focus groups was conducted to develop a dashboard to enable frontline staff to interact with the data. The resulting dashboard displays patterns in patient feedback over time, from which staff can drill down to access the patient’s original narrative.
Manually coding 6000 comments took four days of human labour compared with 15 minutes using the algorithm. Patient feedback data can now be processed in near real time and the dashboard can be accessed from every computer in the hospital. The trust and its staff are using the dashboard to track how patient comments are changing over time in response to changes in care delivery.
The algorithm was developed on open source software, and the team is currently assessing how it can be spread to other trusts.
Getting patients to engage with, and successfully complete, substance misuse treatment is difficult. In order to tackle …