SMARTT - Neonatal

Summary

Extremely premature infants routinely require invasive breathing support with a breathing tube and a ventilator, as they grow most will eventually be able to support their own breathing when mature enough.

There is limited good quality evidence to predict who will be able breath for themselves and when, and who will become too tired and need the support of a ventilator again. Being supported by a ventilator for a longer period leads to poorer clinical outcomes, while having to be supported by a ventilator for a second time requires a breathing tube to be re-inserted is a high-risk procedure, which could have been avoided.

Emerging evidence suggests that analysis of patient characteristics and observations using machine learning may provide a better tool to separate these two groups then current best practice, reducing unwarranted variability in care, and allowing management to be individualised to each baby.
We plan to adapt an existing approach in analysing routinely collected intensive care data to ascertain the feasibility of this approach using data at St Michael's Hospital; if successful this will be the basis for further research grant funding from sources such as the NIHR. The proposed project has been supported by the Topol Digital Fellowship, which is approved and funding Dr Roberson's time working on the project. Further funding is requested to support specialist consultation on the development and evaluation of a machine learning model from the University of Bristol.