2022 01 11 - aasiyahrashan/PhD-Aasiyah GitHub Wiki

Background

  • PhD started Sep 2021. Funding ends Sep 2026. Hoping to finish in approx 3 years, Sep 2024.
  • CDT funding 0.5.
  • Current Wellcome funding 0.4 on self-employed basis, should move to 0.5 on contract.
  • EMAP flexible hours up to 0.2. Need to revisit.
  • Expectation is that I spend around a day supervising a CCAA team who work on the pipeline and analysis.
  • Rest of the Welcome time is spent delivering the PhD and potentially working on other grants.
  • Sent email to Karla asking for supervision.

Roadmap

Stage 1: Create an OMOP database for CCAA.

  • Collaborate with CCHIC to make sure mappings and structure align. Consider the possibility of shared software.
  • Also consult/consider EHDEN/NICE and Epimed structure, as per grant extension.
  • CCAA OMOP should contain information on:
    • Antibiotics
    • Severity
    • Organ support
    • Covid

Stage 2: Develop data quality checks for OMOP (Great Expectations project)

  • CCHIC team already considering this. We should try sharing software and packages.
  • Checks would ideally be 'good enough' for semi-real time decision making, though current CCAA development is aimed at retrospective studies only.
  • Potentially build an ML model to decide when data is good enough.

Stage 3: use the OMOP database to answer a couple of research questions. Current questions on the proposal are:

  • What effect does adhering to 'current' covid guidelines have on outcomes in UCLH and Asia?
    • Has the advantage that data is available and likely to be relatively similar between the 2 locations.
    • Can call it 'causal inference' and use TMLE.
  • An unspecified question related to the effect of starting/stopping interventions. The point is to have longitudinal data from both locations which we use to understand what cadence of data the ML/TMLE models need.
    • Antibiotics
    • Steroids
    • IMV
    • Trache

Potential alternatives.

  • Community acquired pneumonia instead of covid.
  • Build continuously updating LoS model. Not causal inference, but allows downsampling for cadence.
  • VAP

Stage 4: Make sure the questions we answer have a 'causal inference' twist

Action points

  • Aasiyah create Gantt chart for points above.
  • Revisit conversation on working hours with emphasis on EMAP time. Need to be careful not to do more than full time when Welcome contract starts.
  • Steve put Aasiyah in touch with CCHIC devs (Ricardo/Ed) to move conversation on stages 1 and 2 forward.

images/2022-01-11-whiteboard.jpeg