
To reduce the administrative burden on GPs the use of AI systems to record and make a summary of a GP-patient conversation are being investigated. These summaries should be an accurate representation of the conversation. The CoDE laboratory performed tests on a number of different microphones used to record these GP-patient conversations from which the summary is produced. The main findings were that errors in the information were produced the further from the microphone that the conversation was being held. The graph shows the amount of correct information falls sharply as the distance increases. On the graph, the red dashed line represents 100% correct information is recorded, whereas 0% represents all information lost.
The important thing to note is that the laptop microphone was by far the worst performance, and it is recommended not use the laptop microphone but an independent stand-alone microphone. The cost of the stand-alone microphone had little effect on the loss of information with distance.