Kia ora,
I’m looking to collaborate with those in the informatics space who might have access to non-identifiable inpatient vital signs data and would be interested in a model validation study to compare deterioration risk algorithms.
Some background: last year, I developed a clinical deterioration monitoring approach for acute inpatients based on time-to-event analysis. The gist is that this predicts the timing of deterioration rather than whether it will happen, and therefore is useful for prioritising scarce clinical resources (like doctor or nurse manager time).
The manuscript is here:
Essentially the project would involve taking a few different algorithms and applying them to new datasets outside of their development setting, assessing their generalisable performance (discrimination and calibration). In the longer term, the goal is to progress towards randomised trials of these algorithms to assess efficacy and cost-effectiveness.
Thanks!