New technology is being developed at UCLH which could help ensure all patients receive the right follow up care after their appointment.
The software, called CogStack, uses a type of artificial intelligence called natural language processing (NLP) to interpret free text notes found in a patient’s record.
The software is being developed by a team from the BRC-supported Clinical and Research Informatics Unit, in conjunction with the Gastrointestinal (GI) Clinic at UCLH.
The challenge
The GI clinic identified a problem where, occasionally, the ‘next clinical action’ is not recorded in a patient’s record. These follow up actions include referrals onto other services, requests for bloods and imaging procedures, and should be created by the clinician or clinical support staff upon the completion of a patient consultation.
In a small number of cases, due to human error, these orders are not placed and after a period of time the patients can become lost to follow up.
The solution
Where next clinical actions have not been recorded in a patient’s record, it is usually possible to infer what orders should have been created for a patient, since a patient’s next clinical steps are almost always documented by the clinician in the patient’s notes.
This is what CogStack is designed to do – through use of NLP. By interpreting free text found in a patient’s medical record, it can highlight who the clinic may need to follow up with and how.
The CogStack system makes use of a trained machine learning model (BERT model) that has been trained to detect various intents/orders related to the GI clinic. The model has been pre-trained in a supervised training session where GI operational staff trained the model to accurately detect various intents [1].
The UCLH CogStack team have developed an alerting dashboard that would allow GI operational staff to act on what CogStack identifies as a follow up action it believes should have been created for a patient based on what it has picked up on in the patient’s notes.
The next stage of the research is to trial the system in the GI clinic.
