Emergency department triage data - Queens University Hospital

[martintoreilly]

@vollmersj has been working on an earlier Turing project with Queens University Hospital (part of the Barking, Havering and Redbridge University Hospitals NHS Trust). This project uses emergency department admission data. It arose as a follow up to an initial Data Study Group (DSG) "hack week" at the Turing and most recently Queens have been developing an "emergency department triage app" with Health Foundation funding.

Nik Haliasos at the Trust is keen to collaborate and sees value in being able to use synthetic data in the development of the triage app. This would count as "primary use", which is important as the triage project's existing ethics approval only covers such primary use (i.e. analysis on unconsented patient data that used to improve patient care).

[martintoreilly]

Key points from initial email exchange between Martin O and Nik H on 19 Oct 2020:

• Original ethic approval for the "emergency department triage app" projects relies on unconsented patient data access for analysis used to improve patient care. other uses fall under "secondary use" and would not be covered by the triage project's ethics approval and therefore require a lengthy separate ethics approval process at the NHS Trust.

• Nik thinks that being able to evaluate how effectively synthetic data could be used to train the triage app would be analysis that would have a direct improvement to the triage app and so fall under the scope of their existing project approval.

• Work on synthetic data under the existing triage project approval will be restricted to the non-text portion of the patient data. If we want to use the text part, which Nik thinks is very rich and interesting but challenging from our research point of view, he will have to apply to National CAG. Nik is willing to do this, but it will take time. At the moment, I think QUiPP is focussing on non-text data in general, so let's cross this bridge when we come to it and stick with the non-text triage data for now.

• The dataset would need to be analysed within a safe haven with no internet access. Nik would like to retain approval rights for granting access to the safe haven to researchers.

[martintoreilly]

Notes from call with Nik H, Martin O, Catherine L, Amber R, Khanisa R on 19 Nov 2020:

Notes

• The original tirage project has built models on the real data. We could use synthetic data to generate these models and see if how their performance compares to those trained on the real data. That would indicate if the synthetic data could be shared more widely to help develop/improve the models. This would directly benefit the triage project and Nick H and Martin O both therefore believe this should be covered by the existing agreement
• The original data sharing agreement between the NHS Trust and the Turing finishes February 2020, therefore needs renewing anyway if the data is to be used going forward.
• The original project with Queens, with Catalina V and Sebastian V, now has the final bit of work being done in the hospital, with data held at the hospital, not using the ‘safe haven’, with the analyst employed by the hospital (funded via a Health Foundation award)
• The Health Foundation funding finishes in Feb and the analyst role finishes in Feb – Nick H, Sebastian V and Catalina V are still be keen to work together on the project
• The safe haven that data held in at Turing originally is a very old instantiation of a safe haven – was set up before Turing special project to create a more useable secure safe haven. Nik H's experience of using it is poor – Catherine L and Martin O confirmed there had been a huge amount of development work since that time and the Turing Safe Haven was now a very different environment from a usability perspective.

Actions

1. The existing agreement needs to be checked on several fronts:
   • Martin O, Sebastian V and Nik H to write up details of this ‘new’ project and data use, and how it directly links to the original project for which the Turing-Queens data sharing agreement was signed
   • This description of data use is then to be reviewed internally by colleagues at Turing to ensure that Partnerships, Legal team and DPO-designate are all content that current agreement would indeed cover this
   • Confirmation also required from Nik H and medical director (who is also the Caldicott Guardian) at Queens that they also agree this is an agreeable project extension
   • QUiPP project should not use the Queens data until both these confirmations are received
   • If colleagues on either side are not content that QUiPP project would be covered by the existing agreement, then a new agreement will be needed before the project can use the data
   • Martin O, Amber R and Katrina P to review existing agreement to check that it would cover the data being stored and used in the new Turing Safe Haven
   • Martin O and Nik H to agree whether the data can in principal be processed in a safe haven shared with other sensitive data sets used by the project in future (subject to the approval of all data providers).
2. Martin O and Nik H to go through the Turing’s new data and project classification process, which will indicate which Tier of Turing Safe Haven is required (data ingress subject to 1) a. and 1) b. above)
3. Katrina P, Amber R and Martin O to consider if any additional agreements need to be in place for any Turing Fellows taking part in the QUIPP project to cover any differences between Turing employee contracts and Turing Fellow agreements. If not, then the Leeds Co-Is on the QUiPP project team will not be able to access the safe haven without these additional agreements.
4. Katrina P, Martin O and Amber R to start the ball rolling on new data sharing agreement needed from February, working with NH (who stated they would be interested to renew for 2- years)
5. Martin O/Nik H to share with Amber R and Katrina P details of the information governance training that team members accessing the data need to undertake (Amber R to keep records of when completed)